Natural Language Processing Inside Pages 2

F.Y.
MCA
(TWO YEARS PATTERN)
SEMESTER - II (CBCS)
NATURAL LANGUAGE
PROCESSING
SUBJECT CODE: MCAE251
© UNIVERSITY OF MUMBAI
Dr. Suhas Pednekar

Vice Chancellor
University of Mumbai, Mumbai
Prof. Ravindra D. Kulkarni Dr. Prakash Mahanwar
Pro Vice-Chancellor, Director,
University of Mumbai IDOL, University of Mumbai
Programme Co-ordinator : Shri Mandar Bhanushe

Head, Faculty of Science and Technology IDOL,
Univeristy of Mumbai – 400098
Course Co-ordinator : Shyam Mohan T.

Department - MCA, IDOL
University of Mumbai- 400098
Course Writers : Mrs. Vidya Bharde

Assistant Professor,
MGMCET Kamothe, Navi Mumbai
: Mr. Krishna Kumar Sharma

Navneet College, Mumbai Central, Mumbai
: Mrs. Hema Darne

DMIMCS, Nashik
May 2022, Print - 1
Published by : Director,
Institute of Distance and Open Learning,
University of Mumbai,
Vidyanagari,Mumbai - 400 098.
DTP composed and Printed by:

Mumbai University Press
Vidyanagari, Santacruz (E), Mumbai- 400098
CONTENTS
Chapter No. Title Page No.
1. Introduction.............................................................................................................01
2. Word level analysis - I.............................................................................................22
3. Word level analysis - II...........................................................................................34
4. Syntax Analysis _ I..................................................................................................56
5. Syntax Analysis - II.................................................................................................69
6. Semantic Analysis - I..............................................................................................82
7. Semantic Analysis - II.............................................................................................95
8. Text Summarization, Text Classification...............................................................110
9. Sentiment Analysis and Opinion Mining - I.........................................................125
10. Sentiment Analysis and Opinion Mining - II........................................................139

F.Y. MCA
SEMESTER - II (CBCS)
NATURAL LANGUAGE PROCESSING
SYLLABUS
Module No Detailed Contents Hrs

1 Module: Introduction 4
History of NLP, Generic NLP system, levels of NLP ,
Knowledge in language processing , Ambiguity in Natural
language , stages in NLP, challenges of NLP ,Applications of
NLP
Self learning topics: Empirical laws
2 Module: Word Level Analysis 8
Morphology analysis –survey of English Morphology,
Inflectional morphology & Derivational morphology,
Lemmatization, Regular expression, finite automata, finite
state transducers (FST) ,Morphological parsing with FST
, Lexicon free FST Porter stemmer. N –Grams- N-gram
language model,
Self learning topics:N-gram for spelling correction
3 Module:Syntax analysis 8
Part-Of-Speech tagging( POS)- Tag set for English ( Penn
Treebank ) , Rule based POS tagging, Stochastic POS tagging,
Issues –Multiple tags & words, Unknown words. Introduction
to CFG, Sequence labeling: Hidden Markov Model (HMM),
Maximum Entropy
Self learning topics: Conditional Random Field (CRF).
4 Module:Semantic Analysis 10
Lexical Semantics, Attachment for fragment of English-
sentences, noun phrases, Verb phrases, prepositional phrases,
Relations among lexemes & their senses –Homonymy,
Polysemy, Synonymy, Hyponymy, Robust Word Sense
Disambiguation (WSD),Dictionary based approach
Self learning topics:WordNet
5 Module: Text Summarization, Text Classification 6
Text summarization- LEXRANK , Optimization based
approaches for summarization , Summarization evaluation,
Text classification
Self learning topics: NLKT , Naïve Bayes Theorem
6 Module: Sentiment Analysis and Opinion Mining 4
Sentiment Analysis introduction , Sentiment Analysis -
Affective lexicons, Learning affective lexicons, Computing with
affective lexicons, Aspect based sentiment analysis
Self learning topics: Named Entity Recognition
1
INTRODUCTION
Unit Structure
1.0 Objectives
1.1 History of NLP
1.2 Generic NLP system
1.3 Levels of NLP
1.4 Knowledge in Language Processing
1.5 Ambiguity in Natural Language
1.6 Stages in NLP
1.7 Challenges in NLP
1.8 Applications of NLP
1.9 Summary
1.10 Multiple Choice Question Answers
1.11 True or False
1.12 Sample Questions
1.13 List of References
1.0 OBJECTIVES
The objective of this module is to learn about history of Natural Language

Processing (NLP), Generic NLP system, Levels of NLP. After that we
will discuss the knowledge in language processing. This is followed by a
discussion of the Ambiguity in Natural language, Stages in NLP, and
challenges of NLP. Finally we will discuss the applications of NLP.
1.1 HISTORY OF NLP
As we all know Natural Language refers to the language spoken by people

e.g. English, Hindi, Marathi as opposed to artificial/programming
languages, like C, C++, Java, etc.
Natural Language Processing (NLP) is a field of research which
determines the way computers can be used to understand and manage
natural language text or speech to do useful things.
Natural Language Processing (NLP) a subset technique of Artificial
Intelligence which is used to narrow the communication gap between the
Computer and Human. It is originated from the idea of Machine
Translation (MT) which came to existence during the 1950. The primary
aim was to translate one human language to another human language, for
example, Russian language to English language using the brain of the
Computers but after that, the thought of conversion of human language to
computer language and vice-versa emerged, so that communication with
the machine became easy.
1
Natural Language Processing NLP is a process where input provided in a human language and converts
this input into a useful form of representation. The field of NLP is
primarily concerned with getting computers to perform interesting and
useful tasks with human languages. The field of NLP is secondarily
concerned with helping us come to a better understanding of human
language.
As stated above the idea had emerged from the need for Machine
Translation in the 1950s. Then the original language was English and
Russian. But the use of other words such as Chinese also came into
existence in the initial period of the 1960s.
In the 1960s, the NLP got a new life when the idea and need of Artificial
Intelligence emerged.
In 1978 LUNAR is developed by W.A woods; it could analyze, compare
and evaluate the chemical data on a lunar rock and soil composition that
was accumulating as a result of Apollo moon missions and can answer the
related question.
In the 1980s the area of computational grammar became a very active
field of research which was linked with the science of reasoning for
meaning and considering the user‘s beliefs and intentions.
In the period of 1990s, the pace of growth of NLP increased. Grammars,
tools and Practical resources related to NLP became available with the
parsers. Probabilistic and data-driven models had become quite standard
by then.
In 2000, Engineers had a large amount of spoken and textual data
available for creating systems. Today, large amount of work is being done
in the field of NLP using Machine Learning or Deep Neural Networks in
general, where we are able to create state-of-the-art models in text
classification, Question and Answer generation, Sentiment Classification,
etc.
1.2 GENERIC NLP SYSTEM
Natural language Processing should start with some input and ends with
effective and accurate output. The possible inputs to an NLP are quite
broad. Language can be in a variety of forms such as the paragraphs of
text, commands that are typed directly to a computer system, etc. The
input language might be given to the system a sentence at a time or it
might be multiple sentences all at once. The inputs for natural language
processor can be typed input, message text or speech. NLP systems have
some kind of pre-processor. Data preprocessing involves preparing and
cleaning text data for machines to be able to analyze it. Preprocessing puts
data in workable form and highlights features in the text that an algorithm
can work with. Preprocessing does dictionary lookup, morphological
analysis, lexical substitutions, and part-of-speech assignment.
22
There are variety of output can be generated by the system. The output Introduction
from a system that incorporates an NLP might be an answer from a
database, a command to change some data in a database, a spoken
response, Semantics, Part of speech, Morphology of word, Semantics of
the word or some other action on the part of the system. Remember these
are the output of the system as a whole, not the output of the NLP
component of the system.
Figure 1.2.1 shows a generic NLP system and its input-output variety.
Figure 1.2.2 shows a typical view of what might be inside the NLP box of
Figure 1.2.1. Each of the boxes in Figure 1.2.2 represents one of the types
of processing that make up an NLP analysis.
Figure 1.2.1 Generic NLP system.
Figure 1.2.2 Pipeline view of the components of a generic NLP system.
3
Natural Language Processing 1.3 LEVELS OF NLP
There are two components of Natural Language Processing

a. Natural Language Understanding (NLU)
• NLU takes some spoken/typed sentence and working out what
it means.
• Here different level of analysis required such as morphological
analysis, syntactic analysis, semantic analysis, discourse
analysis, …
b. Natural Language Generation (NLG)
• NLG takes some formal representation of what you want to
say and working out a way to express it in a natural (human)
language (e.g., English)
• Here different level of synthesis required: deep planning
(what to say), syntactic generation
Difference between NLU and NLG
NLU NLG
It is the process of reading and It is the process of writing or generating

interpreting language. language.
NLU explains the meaning behind NLG generates the natural language
the written text or speech in natural using machines.
language
NLU draws facts from the natural NLG uses the insights generated from
language using various tools and parsers, POS tags, etc. to generate the
technologies such as parsers, POS natural language.
taggers, etc,
NLU understands the human NLG uses the structured data

language and converts it into data and generates meaningful narratives out
of it
The NLP can broadly be divided into various levels as shown in Figure
1.3.1.
44
Introduction
Figure 1.3.1 Levels of NLP

1. Phonology:
It concerned with interpretation of speech sound within and across
words.
2. Morphology:
It deals with how words are constructed from more basic meaning
units called morphemes. A morpheme is the primitive unit of
meaning in a language. For example, “truth+ful+ness”.
3. Syntax:
It concerns how words can be put together to form correct sentences
and determines what structural role each word plays in the sentence
and what phrases are subparts of other phrases.
For example, “the dog ate my homework”
4. Semantics:
It is a study of the meaning of words and how these meaning
combine in sentences to form sentence meaning. It is study of
context- independent meaning. For example, plant: industrial plant/
living organism
Pragmatics concerns with how sentences are used in different
situations and how it affects the interpretation of the sentence.
Discourse context deals with how the immediately preceding
sentences affect the interpretation of the next sentence. For example,
interpreting pronouns and interpreting the temporal aspects of the
information.
5. Reasoning:
To produce an answer to a question which is not explicitly stored in
a database; Natural Language Interface to Database (NLIDB) carries
out reasoning based on data stored in the database. For example,
consider the database that holds the student academic information,
and user posed a query such as: ‘Which student is likely to fail in
Science subject?’ To answer the query, NLIDB needs a domain
expert to narrow down the reasoning process.
5
Natural Language Processing 1.4 KNOWLEDGE IN LANGUAGE PROCESSING
A natural language understanding system must have detailed information

about what the words mean, how words combine to form sentences, how
word meanings combine to form sentence meanings and so on.
What distinguishes these language processing applications from other data
processing systems is their use of knowledge of language.
For example consider the Unix wc program, which is used to count the
total number of bytes, words, and lines in a text file. When used to count
lines and bytes, wc is an ordinary data processing application. However,
when it is used to count the words in a file it requires information about
what it means to be a word, and thus becomes a language processing
system. Of course, wc is an extremely simple system with an extremely
limited and impoverished knowledge of language.
The different forms of knowledge required for natural language
understanding are given below:
Phonetic and Phonological Knowledge
Phonetics is the study of language at the level of sounds while phonology
is the study of combination of sounds into organized units of speech, the
formation of syllables and larger units. Phonetic and phonological
knowledge are necessary for speech based systems as they are concerned
with how words are related to the sounds that realize them.
Morphological Knowledge
Morphology concerns with word formation. It’s a study of the patterns of
formation of words by the combination of sounds into minimal distinctive
units of meaning called mophemes. Morphological knowledge deals with
how words are constructed from morphemes.
Syntactic Knowledge
Syntax is the level at which we study how words combine to form phrases,
phrases combine to form clauses and clauses join to make sentences.
Syntactic analysis concerns sentence formation. It’s concerned with how
words can be put together to form correct sentences. It also determines
what structural role each word plays in the sentence and what phrases are
subparts of what other phrases.
Semantic Knowledge
It deals with meanings of the words and sentences. This is the study of
context independent meaning that is the meaning a sentence has, no matter
in which context it is used. Defining the meaning of a sentence is very
difficult due to the ambiguities involved.
66
Pragmatic Knowledge Introduction
Pragmatics is the extension of the meaning or semantics. Pragmatics

concerned with the contextual aspects of meaning in particular situations.
It concerned with how sentences are used in different situations and how
use affects the interpretation of the sentence.
Discourse Knowledge
Discourse concerns connected sentences. It’s a study of chunks of
language which are bigger than a single sentence. Discourse language
concerned with inter-sentential links means how the immediately
preceding sentences affect the interpretation of the next sentence.
Discourse knowledge is important for interpreting pronouns and temporal
aspects of the information conveyed.
World Knowledge
Word knowledge is nothing but everyday knowledge that all speakers
share about the world. It includes the general knowledge about the
structure of the world and what each language user must know about the
other user’s beliefs and goals. This is essential to make the language
understanding much better.
1.5 AMBIGUITY IN NATURAL LANGUAGE
Ambiguity can occur at all NLP levels. It is a property of linguistic

expressions. If an expression (word/phrase/sentence) has more than one
meaning we can refer it as ambiguous.
For eg: Consider the sentence,
“The chicken is ready to eat.”
The meaning in the above phrase can be,
The chicken (food) is ready to be eaten or the chicken (bird) is ready to be
feeder
Consider another sentence,
“There was not a single man at the party.”
The meaning in this case can be Lack of men altogether or Lack of
bachelors at the party.
There are different forms of ambiguity that are relevant in natural
language processing.
1. Lexical Ambiguity: It’s the ambiguity of a single word. A word can
be ambiguous with respect to its syntactic class. Eg: book, study.
For eg: The word "silver" can be used as, an adjective, a noun or a
verb.
• She made a silver speech.

• She bagged two silver medals.
• His worries had silvered his hair.
7
Natural Language Processing Lexical ambiguity can be resolved by Lexical category
disambiguation ie, parts-of speech tagging. As many words may
belong to more than one lexical category. Part-of speech tagging is
the process of assigning a part-of-speech or lexical category such as
a noun, pronoun, verb, preposition, adverb, adjective etc. to each
word in a sentence.
Lexical Semantic Ambiguity:
The type of lexical ambiguity, which occurs when a single word is
associated with multiple interpretations. Eg: fast, bat, bank, pen,
cricket etc.
For eg: 1. The tank was full of water.
2. I saw a military tank.
Words have multiple meanings for such sentences. Consider the
sentence "I saw a bat."
Possible meaning of the words which changes the contexts of the
sentence are:
• bat flying mammal / wooden club?
• saw past tense of "see" /present tense of "saw “ (to cut with a
saw.)
The occurrence of tank in both sentences corresponds to the
syntactic category noun, but their meanings are different. Lexical
Semantic ambiguity resolved using word sense disambiguation
(WSD) techniques, where WSD aims at automatically assigning the
meaning of the word in the context in a computational manner.
2. Syntactic Ambiguity: A sentence can be parsed in different ways.
For example, “He lifted the beetle with red cap” or “Did he use cap
to lift the beetle”.
The structural ambiguities were syntactic ambiguities
Structural ambiguity is of two kinds: i) Scope Ambiguity ii)
Attachment Ambiguity
– Scope Ambiguity: Scope ambiguity involves operators and
quantifiers.
For example: Old men and women were taken to safe locations.
The scope of the adjective (i.e., the amount of text it qualifies) is
ambiguous. That is, whether the structure (old men and women) or
((old men) and women)?
The scope of quantifiers is often not clear and creates ambiguity.
Every man loves a woman
88
The meaning can be, For every man there is a woman and also it can Introduction
be there is one particular woman who is loved by every man.
• Attachment Ambiguity
A sentence has attachment ambiguity if a constituent fits more than
one position in a parse tree. Attachment ambiguity arises from
uncertainty of attaching a phrase or clause to a part of a sentence.
Consider the example:
The man saw the girl with the telescope.
It is ambiguous whether the man saw her through his telescope or he
saw a girl carrying a telescope. The meaning is dependent on
whether the preposition ‘with’ is attached to the girl or the man.
Consider the example:
Buy books for children and attach to the verb buy
Preposition Phrase ‘for children’ can be either adverbial or adjectival
and attach to the object noun books.
3. Semantic Ambiguity: This type of ambiguity is typically related to
the interpretation of sentence. Even after the syntax and the
meanings of the individual words have been resolved, there are two
ways of reading the sentence.
Consider example: "Seema loves mother and Sriya does too"
The interpretations can be Sriya loves Seema's mother Sriya likes
her own mother.
Semantic ambiguities born from the fact that generally a computer is
not in a position to distinguishing what is logical from what is not.
Consider the example: “The car hit the pole while it was moving"
The interpretations can be
• The car, while moving, hit the pole
• The car the pole while the pole was moving.

The interpretation is preferred than the second because we have
model of the world that helps distinguish what logical (or possible)
from what is not. To supply to computer model the world is not so
easy.
Consider the example: "We saw his duck”
Duck can refer person's bird or to a motion he made.
Semantic ambiguity happens when a sentence contains an
ambiguous word or phrase.
9
Natural Language Processing 4. Discourse Ambiguity: Discourse level processing needs shared
world or shared knowledge and interpretation is carried out using
this context. Anaphoric ambiguity under discourse level.
• Anaphoric Ambiguity: Anaphora's are the entities that have been

previously introduced into the discourse.
For example, The horse ran up the hill. It was very steep. It soon got
tired.
The anaphoric reference of ‘it’ in the two situation cause ambiguity.
Steep applies to surface hence ‘it’ can be hill. Tired applies to
animate object hence ‘it’ can be horse.
5. Pragmatic Ambiguity: Pragmatic ambiguity refers to a situation
where the context of a phrase gives it multiple meanings. One of the
hardest tasks in NLP. The problem involves processing user
intention, sentiment, belief world etc. all of which are highly
complex tasks.
Consider the example,
"Tourist (checking out of the hotel): Waiter, go upstairs to myroom
and see it my sandals are there; do not be late; I have to catch the
train in 15 minutes."
Waiter (running upstairs and coming back panting) Yes sir, they are
there.
Clearly, the waiter is falling short of the expectation of the tourist,
since he does not understand the pragmatics of the situation.
Pragmatic ambiguity arises when the statement is not specific, and
the context does not provide the information needed to clarify the
statement. Information is missing, and must be inferred. Consider
the example: "I love you too."
This can be interpreted as:
• I love you (as well as liking you)

• I love you (just like someone else does)
• I love you (just like you love me)
• I love you (and I love someone else)
It is a highly complex task to resolve all these kinds of ambiguities,
especially in the upper levels of NLP. The meaning of a word,
phrase, or sentence cannot be understood in isolation and contextual
knowledge is needed to interpret the meaning, pragmatic and world
knowledge is required in higher levels. It is not easy job to create a
world model for disambiguation tasks. Linguistic tools and lexical
resources are needed for the development of disambiguation
techniques. Resourceless languages are lagging behind in these
fields compared to resourceful languages in implementation of these
techniques. Rule based methods are language specific where as
10
10
stochastic or statistical methods are language independent. Introduction
Automatic resolution of all these ambiguities contains several long
standing problems but again development towards full-fledged
disambiguation techniques is required which takes care of all the
ambiguities. It is very much necessary for the accurate working of
NLP applications such as Machine Translation, Information
Retrieval, Question Answering etc.
Statistical Approaches of Ambiguity Resolution in Natural
Language Processing are:
1. Probabilistic model
2. Part of Speech Tagging
• Rule-Based Approaches
• Mak Model Approaches
• Maxman Entropy Approaches
• HMI Based Taggers
3. Machine Learning Approaches
1.6 STAGES IN NLP
There are general five steps in natural language processing
Figure 1.6.1 General five steps of NLP
11
Natural Language Processing 1) Lexical Analysis:
It is the first stage in NLP. It is also known as morphological
analysis. It consists of identifying and analyzing the structure of
words. Lexicon of a language means the collection of phrases and
words in a language. Lexical analysis is dividing the whole chunk
of text into words, sentences, and paragraphs
2) Syntactic Analysis:
Syntactic analysis consists of analysis of words in the sentence for
grammar and ordering words in a way that shows the relationship
among the words. For example the sentence such as “The school
goes to boy” is rejected by English syntactic analyzer.
3) Semantic Analysis:
Semantic analysis is a structure created by the syntactic analyzer
which assigns meanings. This component transfers linear sequences
of words into structures. It shows how the words are associated with
each other. Semantics focuses only on the literal meaning of words,
phrases, and sentences. This only draws the dictionary meaning or
the real meaning from the given text. The structures assigned by the
syntactic analyzer always have assigned meaning
The text is checked for meaningfulness. It is done by mapping
syntactic structure and objects in the task domain. E.g. “colorless
green idea”. This would be rejected by the Symantec analysis as
colorless here; green doesn’t make any sense.
4) Discourse Integration:
The meaning of any sentence depends upon the meaning of the
sentence just before it. Furthermore, it also brings about the meaning
of immediately succeeding sentence. For example, “He wanted
that”, in this sentence the word “that” depends upon the prior
discourse context.
5) Pragmatic Analysis:
Pragmatic analysis concerned with the overall communicative and
social content and its effect on interpretation. It means abstracting or
deriving the meaningful use of language in situations. In this
analysis, what was said is reinterpreted on what it truly meant. It
contains deriving those aspects of language which necessitate real
world knowledge.
E.g., “close the window?” should be interpreted as a request instead
of an order.
12
12
Introduction
Figure 1.6.2 Stages of NLP

Morphological Analysis:
Consider we have an English interface to an operating system and the
following sentence is typed: “I want to print Bill’s .init file “.
Morphological analysis must do the following things:
• Pull apart the word “Bill’s” into proper noun “Bill” and the
possessive suffix “’s”.
• Recognize the sequence “.init” as a file extension that is functioning
as an adjective in the sentence.
This process will usually assign syntactic categories to all the words in the
sentence. Consider the word “prints”. This word is either a plural noun or
a third person singular verb (he prints).
13
Natural Language Processing Syntactic analysis:
This method examines the structure of a sentence and performs detailed
analysis of the sentence and semantics of the statement. In order to
perform this, the system is expected to have through knowledge of the
grammar of the language. The basic unit of any language is sentence,
made up of group of words, having their own meanings and linked
together to present an idea or thought. Apart from having meanings,
words fall under categories called parts of speech. In English languages,
there are eight different parts of speech. They are nouns, pronoun,
adjectives, verb, adverb, prepositions, conjunction and interjections.
In English language, a sentence S is made up of a noun phrase (NP) and a
verb phrase (VP), i.e.
S=NP+VP
The given noun phrase (NP) normally can have an article or delimiter (D)
or an adjective (ADJ) and the noun (N), i.e.
NP=D+ADJ+N
Also a noun phrase may have a prepositional phrase (PP) which has a
preposition (P), a delimiter (D) and the noun (N), i.e.
PP=D+P+N
The verb phrase (VP) has a verb (V) and the object of the verb. The
object of the verb may be a noun (N) and its determiner, i.e.
VP=V+N+D
These are some of the rules of the English grammar that helps one to
construct a small parser for NLP.
Syntactic analysis must exploit the results of morphological analysis to

build a structural description of the sentence. The goal of this process,
called parsing, is to convert the flat list of words that forms the sentence
into a structure that defines the units that are represented by that flat list.
The important thing here is that a flat sentence has been converted into a
hierarchical structure and that the structures correspond to meaning units
14
14
when semantic analysis is performed. Reference markers are shown in the Introduction
parenthesis in the parse tree. Each one corresponds to some entity that has
been mentioned in the sentence.
Semantic Analysis:
The structures created by the syntactic analyzer assigned meanings. Also,
a mapping made between the syntactic structures and objects in the task
domain. Moreover, Structures for which no such mapping possible may
reject
Semantic analysis must do two important things:
It must map individual words into appropriate objects in the knowledge
base or database It must create the correct structures to correspond to the
way the meanings of the individual words combine with each other.
Discourse Integration:
The meaning of an individual sentence may depend on the sentences that
precede it. And also, may influence the meanings of the sentences that
follow it
Specifically we do not know whom the pronoun “I” or the proper noun
“Bill” refers to. To pin down these references requires an appeal to a
model of the current discourse context, from which we can learn that the
current user is USER068 and that the only person named “Bill” about
whom we could be talking is USER073.Once the correct referent for Bill
is known, we can also determine exactly which file is being referred to.
15
Natural Language Processing
Pragmatic Analysis
Moreover, The structure representing what said reinterpreted to determine
what was actually meant. The final step toward effective understanding is
to decide what to do as a results. One possible thing to do is to record what
was said as a fact and be done with it. For some sentences, whose intended
effect is clearly declarative, that is precisely correct thing to do. But for
other sentences, including this one, the intended effect is different. We can
discover this intended effect by applying a set of rules that characterize
cooperative dialogues. The final step in pragmatic processing is to
translate, from the knowledge based representation to a command to be
executed by the system.
Results of each of the main processes combine to form a natural language
system.
The results of the understanding process are lpr /ali/stuff.init. All of the
processes are important in a complete natural language understanding
system. Not all programs are written with exactly these components.
Sometimes two or more of them collapsed. Doing that usually results in a
system that is easier to build for restricted subsets of English but one that
is harder to extend to wider coverage
1.7 CHALLENGES IN NLP
There are a number of challenges of natural language processing because

natural language is ever-evolving and always somewhat ambiguous.
The major challenges around NLP that one must be aware of are as
follows:
1. Language differences.
Different languages have not only different sets of vocabulary, but
also different types of phrasing, different modes of inflection, and
16
16 different cultural norms. You can resolve this issue with the help of
“universal” models that can move at least part of what you’ve Introduction
learned to other languages. However, you have to spend time
updating your NLP system for each additional language.
2. Training data
NLP is all about studying language to better understand it. For a
person to be proficient, it is essential to spend a substantial amount
of time listening, reading, and understanding it, Even the best AI
must also spend a significant amount of time reading, listening to,
and using a language. The training data given to an NLP system
determines its capabilities. If you feed the system bad or
questionable data, it’s going to learn the wrong things, or learn in an
inefficient way.
3. Development time.
You also need to think about the development time for an NLP
system. The total time taken to develop an NLP system is higher. AI
evaluates the data points to process them and use them accordingly.
With a shared deep network and many GPUs working together, the
training time can be reduced by a few hours. The pre-existing NLP
technologies can help in developing the product from scratch.
4. Phrasing ambiguities.
Many times it’s hard even for another human being to parse out
what someone means when they say something ambiguous. There
may not be a clear, concise meaning to be found in analysis of their
words. In order to resolve this, an NLP system must be able to seek
context that can help it understand the phrasing. It may also need to
ask the user for clarity.
5. Misspellings.
It is common for humans to make spelling mistakes but it can be
easily interpreted by human being. But for a machine, misspellings
can be harder to detect it, hence it is essential to employ NLP
technology to progress and identify the misspellings.
6. Words with multiple meanings.
No language is perfect, and most languages have words that can
have multiple meanings, depending on the situation. For example, a
user who asks, “how are you?” has a totally different aim than a user
who asks something like “how do I add a new debit card?” Good
NLP tools should be able to discriminate between these utterances
with the help of context.
7. Phrases with multiple intentions.
Some words and questions actually have multiple meaning, so your
NLP system can’t oversimplify the situation by interpreting only one
of them. For example, a user may say to your chatbot with
something like, “I need to cancel my previous order and update my
card on file.” Your AI needs to be able to distinguish these
intentions separately. 17
Natural Language Processing 8. False positives and uncertainty.
A false positive occurs when an NLP notices a word that should be
intelligible or addressable, but cannot be sufficiently answered. The
idea is to develop an NLP system that can recognize its own
limitations, and use questions to clear up the ambiguity.
9. Keeping a conversation moving.
Many current NLP applications are built on communication between
a human and a machine. Accordingly, your NLP AI needs to be able
to keep the conversation moving, providing more questions to
collect more information or data and always pointing toward a
solution.
1.8 APPLICATIOS OF NLP
• Machine Translation
Machine translation is basically used to convert text or speech from
one natural language to another natural language. Machine
translation, an integral part of Natural Language Processing where
translation is done from source language to target language
preserving the meaning of the sentence. Example: Google
Translator
• Information Retrieval
It refers to the human-computer interaction (HCI) that happens when
we use a machine to search a body of information for information
objects (content) that match our search query. A Person's query is
matched against a set of documents to find a subset of 'relevant'
document. Examples: Google, Yahoo, Altavista, etc.
• Text Categorization
Text categorization (also known as text classification or topic
spotting) is the task of automatically sorting a set of documents into
categories (clusters).
Uses of Text Categorization
• Filtering of content
• Spam filtering Identification of document content
• Survey coding
• Information Extraction -
Identify specific pieces of information in unstructured or semi-structured
textual document. Transform unstructured information in a corpus of
documents or web pages into a structured database.
18
18
Applied to different types of text: Introduction
• Newspaper articles
• Web pages
• Scientific articles
• Newsgroup messages
• Classified ads
• Medical notes
• Grammar Correction-
In word processor software like MS-word, NLP techniques are widely
used for spelling correction & grammar check.
• Sentiment Analysis-
Sentiment Analysis is also known as opinion mining. It is mainly
used on the web to analyse the behaviour, attitude, and emotional
state of the sender. This application is implemented through a
combination of NLP) and statistics by assigning the values to the
text (natural, positive or negative), identify the mood of the context
(sad, happy, angry, etc.)
• Question-Answering systems-
Question Answering focuses on constructing systems that
automatically answer the questions asked by humans in a natural
language. It presents only the requested information instead of
searching full documents like search engine. The basic idea behind
the QA system is that the users just have to ask the question and the
system will retrieve the most appropriate and correct answer for that
question.
E.g.
Q. “What is the birth place of Shree Krishna?
A. Mathura
• Spam Detection
To detect unwanted e-mails getting to a user's inbox, spam detection
is used
• Chatbot
Chatbot is one of the most important applications of NLP. It is used
by many companies to provide the customer's chat services.
• Speech Recognition-
Speech recognition is used for converting spoken words into text. It
is used in applications, such as mobile, home automation, video
recovery, dictating to Microsoft Word, voice biometrics, voice user
interface, and so on.
19
Natural Language Processing • Text summarization
This task aims to create short summaries of longer documents while
retaining the core content and preserving the overall meaning of the
text.
1.9 SUMMARY
In this chapter we studied what is natural language processing (NLP),

history of NLP, generic NLP system, levels of NLP, and knowledge and
ambiguity in NLP. Here we also discussed the stages in NLP, challenges
and applications of NLP.
1.10 MULTIPLE CHOICE QUSETION ANSWERS
i. What is the field of Natural Language Processing (NLP)?

a) Computer Science
b) Artificial Intelligence
c) Linguistics
d) All of the mentioned
Ans : d
ii. What is the main challenge/s of NLP?
a) Handling Ambiguity of Sentences
b) Handling Tokenization
c) Handling POS-Tagging
Ans : a
iii. Choose form the following areas where NLP can be useful.
a) Automatic Text Summarization
b) Automatic Question-Answering Systems
c) Information Retrieval
Ans : d
iv. Which of the following includes major tasks of NLP?
a) Automatic Summarization
b) Discourse Analysis
c) Machine Translation
Ans : d
v. Natural language processing is divided into the two subfields of -
a) symbolic and numeric
b) understanding and generation
c) algorithmic and heuristic
d) None of the above.
20
20 Ans : b
1.11 TRUE OR FALSE Introduction
i. NLP is concerned with the interactions between computers and

human (natural) languages.
True
ii. Machine Translation converts human language to machine language
False
iii. Modern NLP algorithms are based on machine learning, especially
statistical machine learning.
True
1.12 SAMPLE QUESTIONS
1. What is NLP? Discuss the different stages in NLP

2. What are the different types of ambiguities present in NLP? How to
remove it
3. Explain generic NLP system with a diagram
4. State and explain the different applications and challenges in
5. Explain Components of NLP
LIST OF REFERENCES
1. https://www.guru99.com/nlp-tutorial.html
2. https://www.datascienceprophet.com/different-levels-of-nlp/
3. https://www.slideshare.net/HareemNaz/natural-languageprocessing
4. “Natural Language Processing”, Staredu Solutions.
5. Dan Jurafsky and James Martin. “Speech and Language Processing:
An Introduction to Natural Language Processing, Computational
Linguistics and Speech Recognition”, Prentice Hall, Second
Edition, 2009.
6. http://www.deepsky.com/~merovech/voynich/
voynich_manchu_reference_materials/PDFs/jurafsky_martin.pdf
7. Christopher D.Manning and Hinrich Schutze, ― Foundations of
Statistical Natural Language Processing ―, MIT Press, 1999.
8. https://www.cs.vassar.edu/~cs366/docs/
Manning_Schuetze_StatisticalNLP.pdf
9. https://slideplayer.com/slide/4211188/
 21
2
WORD LEVEL ANALYSIS - I
Unit Structure
2.0 Objectives
2.0.1 Morphology Analysis
2.0.2 Survey of English Morphology
2.0.3 Inflectional Morphology
2.0.4 Derivational Morphology
2.0.5 Difference of inflectional and derivational morphology
2.0.6 Stemming and Lemmatization
2.1 Summary
2.1 OBJECTIVES
The objective of this module is to learn about morphology, morphology

analysis, concentrating mainly on survey of English Morphology. After
that we will discuss the Inflectional Morphology & Derivational
Morphology. This is followed by a discussion of the difference of
Inflectional and Derivational Morphology. At last we will discuss the
Stemming and Lemmatization.
2.0.1 MORPHOLOGY ANALYSIS
Word is a letter or group of letters that has meaning when spoken, signed
or written. Words are orthographic tokens separated by white space.
Many languages the distinction between words and sentences is less clear.
In Chinese, Japanese languages no white space between words
Example: nowhitespace ~ no white space/no whites pace/now hit esp ace
In Turkish language words could represent a complete “sentence”
Example: uygarlastiramadiklarimizdanmissinizcasina
Morphology is a study of word formation. Some words can be divided into
two parts but still have meaning. Example: handsome (hand + some).
Many words have meaning in themselves but some words have meaning
only when used with other words. Example: sleeping, colourful. Some
words can stand alone. Example: free, cat, run tree, table, local etc. Words
parts must be combined in the correct way and systematically.
Morphology deals with the syntax of complex words and parts of words,
also called morphemes, as well as with the semantics of their lexical
meanings. Understanding how words are formed and what semantic
22
22
properties they convey through their forms enables human beings to easily Word level analysis - I
recognize individual words and their meanings in discourse.
Morpheme is defined as “minimal meaning-bearing unit in a language"
Morphology handles the formation of words by using morphemes
base form (stem , lemma), e.g., believe
affixes (suffixes, prefixes, infixes), e.g., un-, -able, -ly
Morphological parsing is the task of recognizing the morphemes inside a
word e.g., hands, foxes, children and it’s important for many tasks
machine translation, information retrieval, ,in parsing, and text
simplification, etc
2.0.2 SURVEY OF ENGLISH MORPHOLOGY
Morphology concerns the structure of words. Words are assumed to be
made up of morphemes, which are the minimal information carrying unit.
So for example the word dog consists of a single morpheme (the
morpheme dog) while the word cats consists of two: the morpheme cat
and the morpheme-s.
Consider a word like: "unhappyness". This has three parts as shown in
figure 2.1.2
Figure 2.1.2
There are three morphemes, each carrying a certain amount of
meaning. un means "not", while ness means "being in a state or
condition". Happy is a free morpheme because it can appear on its own (as
a "word" in its own right). Bound morphemes have to be attached to a free
morpheme, and so cannot be words in their own right. Thus you can't have
sentences in English such as "Jason feels very un ness today".
We can usefully divide morphemes into two broad classes of morphemes:
stems and affixes.
• The stem is the “main” morpheme of the word, supplying the main
meaning
• The affixes add “additional” meanings of various kinds.
23
Affixes are further divided into prefixes, suffixes, infixes, and

circumfixes.
• prefixes precede the stem,

• suffixes follow the stem,
• circumfixes do both,
• infixes are inserted inside the stem.
For example, the word cats is composed of a stem cat and the suffix -s.
The word undo is composed of a stem do and the prefix un-. Good
examples of circumfixes are e.g Unreachable, Unbelievable. Infixes, in
which a morpheme is inserted in the middle of a word, occur very
commonly for example in the Philipine language Tagalog. For example
the affix um, which marks the agent of an action, is infixed to the Tagalog
stem hingi “borrow” to produce humingi (ask for or to demand).
A word can have more than one affix. For example, the word rewrites has
the prefix re-, the stem write, and the suffix -s. The word unbelievably has
a stem (believe) plus three affixes (un-, -able, and -ly). There are many
ways to combine morphemes to create words.
Four of these methods are common and play important roles in speech and
language processing: inflection, derivation, compounding, and
cliticization.
Inflection is the combination of a word stem with a grammatical
morpheme, usually resulting in a word of the same class as the original
stem, and usually filling some syntactic function like agreement. For
example, English has the inflectional morpheme -s for marking the plural
on nouns, and the inflectional morpheme –ing for marking the present
tense on verbs. e.g. clean (verb), clean-ing (verb)
Derivation is the combination of a word stem with a grammatical
morpheme, usually resulting in a word of a different class, often with a
meaning hard to predict exactly. For example the verb delight can take the
24
24
derivational suffix -ful to produce the adjective delightful. e.g delight Word level analysis - I
(verb), delight-ful (adj)
Compounding is the combination of multiple words stems together. In
English compounds can be written together, for example, notebook,
bookstore, fireman etc. For example the noun doghouse is the
concatenation of the morpheme dog with the morpheme house.
Finally, cliticization is the combination of a word stem with a clitic. A
clitic is a morpheme that acts syntactically like a word, but is reduced in
form and attached (phonologically and sometimes orthographically) to
another word. For example the English morpheme’ve in the word I’ve is a
clitic. That is different words from different syntactic categories, e.g. I’ve
= I + have.
2.0.3 INFLECTIONAL MORPHOLOGY
Inflection is a morphological process that adapts existing words so that
they function effectively in sentences without changing the category of the
base morpheme.
Inflectional affixes are only suffixes which express grammatical features
such as singular/plural, past/present forms of verbs. The study of such
morphemes is called inflectional morphemes. Examples: bigger (big+er),
loves (love+s).
English has a relatively simple inflectional system; only nouns, verbs, and
sometimes adjectives can be inflected, and the number of possible
inflectional affixes is quite small.
English nouns have only two kinds of inflection: an affix that marks plural
and an affix that marks possessive. For example, many English nouns can
either appear in the singular form, or take a plural suffix. Here are
examples of the regular plural suffix -s (also spelled -es), and irregular
plurals and sometimes -x (box/boxes). Nouns ending in -y preceded by a
consonant change the -y to -i (butterfly/butterflies).
Table 2.1.3.1
Regular Nouns Irregular Nouns

Singular cat thrush mouse ox
Plural cats thrushes mice oxen
The possessive suffix is realized by apostrophe + -s for regular singular

nouns (llama’s) and plural nouns not ending in -s (children’s) and often by
a lone apostrophe after regular plural nouns (llamas’) and some names
ending in -s or -z (Euripides’ comedies).
English verbal inflection is more complicated than nominal inflection.
First, English has three kinds of verbs;
25
Natural Language Processing • main verbs, (eat, sleep, impeach),
• modal verbs (can, will, should),
• primary verbs (be, have, do).
Here we will mostly be concerned with the main and primary verbs,
because it is these that have inflectional endings. Of these verbs a large
class are regular, that is to say all verbs of this class have the same endings
marking the same functions. These regular verbs (e.g. walk, or inspect)
have four morphological forms. These verbs are called regular because
just by knowing the stem we can predict the other forms by adding one of
three predictable endings and making some regular spelling changes.
These regular verbs and forms are significant in the morphology of
English first because they cover a majority of the verbs, and second
because the regular class is productive. As discussed earlier, a productive
class is one that automatically includes any new words that enter the
language. For example the recently-created verb fax (My mom faxed me
the note from cousin Everett) takes the regular endings -ed, -ing, -es.
(Note that the -s form is spelled faxes rather than faxs).
The table 2.0.3.2 shows the morphological forms for regular verbs.
Table 2.0.3.2
Morphological Regularly Inflected Verbs

Form Classes
stem talk merge cry map
-s form talks merges cries maps
-ing participle talking merging crying mapping
Past form or -ed talked merged cried mapped
participle
The irregular verbs are those that have some more or less idiosyncratic
forms of inflection. Irregular verbs in English often have five different
forms, but can have as many as eight (e.g., the verb be) or as few as three
(e.g. cut or hit). While constituting a much smaller class of verbs estimate
there are only about 250 irregular verbs, not counting auxiliaries), this
class includes most of the very frequent verbs of the language. The table
2.4.3 shows some sample irregular forms. Note that an irregular verb can
inflect in the past form (also called the preterite) by changing its vowel
(eat/ate), or its vowel and some consonants (catch/caught), or with no
change at all (cut/cut).
The table 2.1.3.3 shows the morphological forms for irregular verbs.
Table 2.1.3.3
Morphological Form Classes Irregularly Inflected Verbs

stem eat catch put
-s form eats catches puts
-ing participle eating catching putting
Past form ate caught put
-ed participle eaten caught put
26
26
The -s form is used in the “habitual present” form to distinguish the third- Word level analysis - I
person singular ending (She jogs every Tuesday) from the other choices of
person and number (I/you/we/they jog every Tuesday). The stem form is
used in the infinitive form, and also after certain other verbs (I’d rather
walk home, I want to walk home). The -ing participle is used in the
progressive construction to mark present or ongoing activity (It is raining),
or when the verb is treated as a noun; this particular kind of nominal use
of a verb is called a gerund use: Fishing is fine if you live near water. The
-ed participle is used in the perfect construction (He’s eaten lunch already)
or the passive construction (The verdict was overturned yesterday).
In addition to noting which suffixes can be attached to which stems, we
need to capture the fact that a number of regular spelling changes occur at
these morpheme boundaries. For example, a single consonant letter is
doubled before adding the -ing and -ed suffixes (beg/begging/begged). If
the final letter is “c”, the doubling is spelled “ck”
(picnic/picnicking/picnicked). If the base ends in a silent -e, it is deleted
before adding -ing and -ed (merge/merging/merged). Just as for nouns, the
-s ending is spelled -es after verb stems ending in -s (toss/tosses), -z,
(waltz/waltzes) -sh, (wash/washes) -ch, (catch/catches) and sometimes -x
(tax/taxes). Also like nouns, verbs ending in –y preceded by a consonant
change the -y to -i (try/tries).
2.0.4 DERIVATIONAL MORPHOLOGY
Derivation is concerned with the way morphemes are connected to
existing lexical forms as affixes.
English inflection is relatively simple compared to other languages,
derivation in English is quite complex. Recall that derivation is the
combination of a word stem with a grammatical morpheme, usually
resulting in a word of a different class, often with a meaning hard to
predict exactly.
The common kind of derivation in English is the formation of new nouns,
often from verbs or adjectives. This process is called nominalization. For
example, the suffix -ation produces nouns from verbs ending often in the
suffix -ize (computer-ize → computerization).
The table 2.0.4.1 shows the formation of new nouns, often from verbs or
adjectives.
Table 2.0.4.1
Suffix Base Derived Noun

Verb/Adjective
-ation computerize (V) computerization
-ee appoint (V) appointee
-er kill (V) killer
-ness fuzzy (A) fuzziness
27
Natural Language Processing Adjectives can also be derived from nouns and verbs. The table 2.0.4.2
shows adjectives derived from the verbs or noun.
Table 2.0.4.2
Suffix Base Noun/Verb Derived Adjective

-al computation (N) computational
embrace (V) embraceable
-able
clue (N) clueless
-less
Derivation in English is more complex than inflection for a number of
reasons. One is that it is generally less productive; even a nominalizing
suffix like -ation, which can be added to almost any verb ending in -ize,
cannot be added to absolutely every verb. Thus we can’t say *eatation or
*spellation (we use an asterisk (*) to mark “non-examples” of English).
Another is that there are subtle and complex meaning differences among
nominalizing suffixes. For example sincerity has a subtle difference in
meaning from sincereness.
2.0.5 DIFFERENCE OF INFLECTIONAL AND DERIVATIONAL
MORPHOLOGY
1) Inflectional morphemes never change the grammatical category (part
of speech) of a word.
For example, tall and taller are both adjectives. The inflectional
morpheme -er (comparative marker) simply produces a different
version of the adjective tall.
derivational morphemes often change the part of speech of a word.
Thus, the verb read becomes the noun reader when we add the
derivational morpheme -er.
It is simply that read is a verb, but reader is a noun
example:-For example, such derivational prefixes as re- and un- in
English generally do not change the category of the word to which
they are attached. Thus, both happy and unhappy are adjectives, and
both fill and refill are verbs, for example. The derivational suffixes -
hood and -dom, as in neighborhood and kingdom, are also the
typical examples of derivational morphemes that do not change the
grammatical category of a word to which they are attached.
2) When a derivational suffix and an inflectional suffix are added to the
same word, they always appear in a certain relative order within the
word
That is, inflectional suffixes follow derivational suffixes. Thus, the
derivational (-er) is added to read, then the inflectional (-s) is
attached to produce readers.
Similarly, in organize– organizes the inflectional -s comes after the
derivational -ize. When an inflectional suffix is added to a verb, as
with organizes, then we cannot add any further derivational suffixes
28
28 It is impossible to have a form like organizesable, with inflectional -
s after derivational - able because inflectional morphemes occur Word level analysis - I
outside derivational morphemes and attach to the base or stem.
3) A third point worth emphasizing is that certain derivational
morphemes serve to create new base forms or new stems to which
we can attach other derivational or inflectional affixes. For example,
we use the derivational -atic to create adjectives from nouns, as in
words like systematic and problematic.
2.0.6 STEMMING AND LEMMATIZATION
In natural language processing, there may come a time when you want
your program to recognize that the words “ask” and “asked” are just
different tenses of the same verb. This is the idea of reducing different
forms of a word to a core root. Words that are derived from one another
can be mapped to a central word or symbol, especially if they have the
same core meaning.
“When we are running a search, we want to find relevant results not only
for the exact expression we typed on the search bar, but also for the other
possible forms of the words we used. For example, it’s very likely we will
want to see results containing the form “shoot” if we have typed “shoots”
in the search bar.”
This can be achieved through two possible methods: stemming and
lemmatization. The aim of both processes is the same: reducing the
inflectional forms of each word into a common base or root. However,
these two methods are not exactly the same.
Example: Affectation, affects, affections, affected, affection, affecting ->
all of these came from single root work affect.
Stemming
Stemming algorithms work by cutting off the end or the beginning of the
word, taking into account a list of common prefixes and suffixes that can
be found in an inflected word.
This indiscriminate cutting can be successful on some occasions, but not
always, and that is why we affirm that this approach presents some
limitations.
Ex: writing 🡺 writ + ing (or write + ing)
29
Natural Language Processing Lemmatization
Lemmatization usually refers to doing things properly with the use of a
vocabulary and morphological analysis of words, normally aiming to
remove inflectional endings only and to return the base or dictionary form
of a word, which is known as the lemma
Example: Lemmatization will map gone, going and went->Go.
• Output of Lemmatization is proper word

• If confronted with the token saw, stemming might return just s,
whereas lemmatization would attempt to return either see or saw
depending on whether the use of the token was as a verb or a noun.
How do they work?
• Stemming: there are different algorithms that can be used in the

stemming process, but the most common in English is Porter
stemmer. The rules contained in this algorithm are divided in five
different phases numbered from 1 to 5. The purpose of these rules is
to reduce the words to the root.
• Lemmatization: the key to this methodology is linguistics. To
extract the proper lemma, it is necessary to look at the
morphological analysis of each word. This requires having
dictionaries for every language to provide that kind of analysis.
Difference between Stemming and Lemmatization:
Sr.no Stemming Lemmatization

Stemming is used to
Lemmatization concept is used to
group words with a
1 make dictionary or WordNet kind
similar basic meaning
of dictionary
together
Accuracy is more as compared to
2 Accuracy is less
stemming.
Stemming is preferred Lemmatization would be
when the meaning of the recommended when the meaning
3 word is not important for of the word is not important for
analysis. Example: Spam analysis. Example: Question
detection Answer
30
30
Stemming usually Word level analysis - I
Lemmatization usually considers
operates on single word
4 words and the context of the word
without knowledge of the
in the sentence
context
When we convert any
word into root form then Lemmatization always gives the
5 stemming may create the dictionary meaning word while
non-existence meaning of converting into root form
word.
6 Stemming for the word Lemma for the word "studies" is
"studies" is studi study
2.1 SUMMARY
In this chapter we studied concept of morphology, morphology analysis in

detailed, and survey of English Morphology. Morphological parsing is the
process of finding the constituent morphemes in a word. We discussed the
Inflectional morphology & Derivational morphology, English mainly uses
prefixes and suffixes to express inflectional and derivational morphology.
English inflectional morphology is relatively simple and includes person
and number agreement (-s) and tense markings (-ed and -ing). English
derivational morphology is more complex and includes suffixes like -
ation, -ness, -able as well as prefixes like co- and re-. Here we also
discussed stemming and Lemmatization.
i. In linguistic morphology, _____________ is the process for

reducing inflected words to their root form.
a) Rooting
b) Stemming
c) Text-Proofing
d) Both a & b
Ans : b
ii. ________ techniques looks at the meaning of the word
a) Stemming
b) Stop Word
c) Leematization
d) Marphological Analysis
Ans : c
31
Natural Language Processing iii. _________is the study of the internal structure of words, and the
rules by which words are formed
a) Syntax
b) Semantics
c) Pragmatics
d) Morphing
Ans : d
iv. Lemma for the word "studies" is_______________
a) studies
b) ies
c) studi
d) study
Ans : d
v. Stemming for the word "studies" is_______________
a) studies
b) ies
c) studi
d) study
Ans : c
TRUE OR FALSE
i. Derivational morphemes form new lexemes conveying new
meanings
True
ii. Lemmatization involves resolving words to their dictionary form.
True
iii. Inflectional morphemes change the grammatical category (part of
speech) of a word.
False
SAMPLE QUESTIONS
1. Define morpheme, stem and affixes.
2. Explain stemming and lemmatization with example
3. Explain with examples Inflectional morphology & Derivational
morphology
4. State differences between inflectional and derivational morphemes
32
32
2.3 LIST OF REFERENCES Word level analysis - I

Edition, 2009.
5. https://cl-illc.github.io/nlp1/resources/slides/Morphology-notes.pdf
6. https://www.cs.vassar.edu/~cs395/docs/3.pdf

33
3
WORD LEVEL ANALYSIS - II
Unit Structure
3.0 Objectives
3.1 Regular expression
3.2 Finite Automata
3.3 Finite State Morphological Parsing
3.4 Building a Finite State Lexicon
3.5 Finite State Transducers (FST)
3.6 Morphological Parsing with FST
3.7 Lexicon free FST Porter Stemmer
3.8 N –Grams
3.9 N-Gram Language Model
3.10 Summary
3.12 True or False
3.13 Sample Questions
3.0 OBJECTIVES
The objective of this module is to learn about Regular expression (RE),

Finite Automata, Finite state morphological parsing, how to build finite
state lexicon, and Finite State Transducers (FST). After that we will
discuss the Morphological Parsing with FST. This is followed by a
discussion of Lexicon free FST Porter Stemmer. Finally we will discuss N
–Grams- N-gram language model.
3.1 REGULAR EXPRESSION
Regular expressions are defined as a sequence of characters that are

mainly used to find substrings in a given string. A regular expression (RE)
is a language for specifying text search strings. It helps us to match or
extract other strings or sets of strings, with the help of a specialized syntax
present in a pattern. For Example, extracting all hashtags from a tweet,
getting phone numbers, email ID etc from large unstructured text content.
Regular Expressions are used in various tasks such as Data pre-processing,
Text feature Engineering, Data Extraction, Rule-based information Mining
systems, Pattern Matching, Web scraping, etc.
34
34
Word Level Analysis - II
Properties of Regular Expressions

Some of the important properties of Regular Expressions are as follows:
1. Regular Expression (RE) is a formula in a special language, which
can be used for specifying a sequence of symbols, simple classes of
strings. In simple words, we can say that regular Expression is an
algebraic notation for characterizing a set of strings.
2. Regular expression requires two things, one is the pattern that we
want to search and the other is a corpus of text or a string from
which we need to search the pattern. For example, the Unix
command-line tool grep takes a regular expression and returns every
line of the input document that matches the expression
Regular Expression can be defined in the following manner:
1. ε is a Regular Expression, which indicates that the language is
having an empty string.
2. φ is a Regular Expression which denotes that it is an empty
language.
3. If X and Y are Regular Expressions, then the following expressions
are also regular.
• X, Y
• X.Y(Concatenation of XY)
• X+Y (Union of X and Y)
• X*, Y* (Kleen Closure of X and Y)
4. If a string is derived from the above rules then that would also be a
regular expression.
How can Regular Expressions be used in NLP?
In NLP, we can use Regular expressions at many places such as,
1. To validate data fields. For example, URLs, dates, email address,
abbreviations, etc.
2. To filter a particular text from the whole corpus. For
example, disallowed websites, spam etc.
3. To identify particular strings in a text. For example, token
boundaries
4. To convert the output of one processing component into the format
required for a second component
Basic Regular Expression Patterns
Brackets ([ ]):
They are used to specify a disjunction of characters.
35
For Examples,
/[A-Z]/ → matches an uppercase letter
/[a-z]/ → matches a lowercase letter
/[0–9]/ → matches a single digit
For Examples,
/[cC]hirag/ → Chirag or chirag
/[xyz]/ → ‘x’, ‘y’, or ‘z’
/[1234567890]/ → any digit
Here slashes represent the start and end of a particular expression.
Dash (-):
They are used to specify a range.
Caret (^):
They can be used for negation or just to mean ^.
For Examples,
/[ˆa-z]/ → not an lowercase letter
/[ˆCc]/ → neither ‘C’ nor ‘c’
/[ˆ.]/ → not a period
/[cˆ]/ → either ‘c’ or ‘ˆ’
/xˆy/ → the pattern ‘xˆy’
36
36
Question mark (?):

It marks the optionality of the previous expression.
For Examples,
/maths?/ → math or maths
/colou?r/ → color or colour
Period (.):
Used to specify any character between two expressions.
/beg.n/ → Match any character between which fits between beg and n
such as begin, begun
Anchors :
These are special characters that help us to perform string operations
either at the beginning or at the end of text input. They are used to assert
something about the string or the matching process. The most common
anchors are the caret ^ and the dollar sign $.
i) Caret character ‘^’
It specifies the start of the string. For a string to match the pattern,
the character followed by the ‘^’ in the pattern should be the first
character of the string.
For Examples,
^The: Matches a string that starts with ‘The’
ii) Dollar character ‘$’
It specifies the end of the string. For a string to match the pattern,
the character that precedes the ‘$’ in the pattern should be the last
character in the string.
For Examples,
end$: Matches a string that ends with ‘end’
^The end$’: Exact string match (starts and ends with ‘The end’)
roar: Matches a string that has the text roar in it.
37
Quantifiers:
They allow us to mention and control over how many times a specific
character(s) pattern should occur in the given text.
The most common Quantifiers are: *, +, ? and { }
For Examples,
abc*: matches a string that has 'ab' followed by zero or more 'c'.
abc+: matches 'ab' followed by one or more 'c'
abc?: matches 'ab' followed by zero or one 'c'
abc{2}: matches 'ab' followed by 2 'c'
abc{2, }: matches 'ab' followed by 2 or more 'c'
abc{2, 5}: matches 'ab' followed by 2 upto 5 'c'
a(bc)*: matches 'a' followed by zero or more copies of the sequence 'bc'
3.2 FINITE AUTOMATA
The term automata, derived from the Greek word "αὐτόματα. Automaton
may be defined as an abstract self-propelled computing device that follows
a predetermined sequence of operations automatically. An automaton
having a finite number of states is called a Finite Automaton (FA) or
Finite State automata (FSA).
The finite-state automaton is not only the mathematical device used to
implement regular expressions, but also one of the most significant tools
of computational linguistics. Variations of automata such as finite- state
transducers, Hidden Markov Models, and N-gram grammars are important
components of the speech recognition and synthesis, spell-checking, and
information-extraction applications.
Mathematically, an automaton can be represented by a 5-tuple (Q, Σ, δ,
q0, F),
where −
• Q is a finite set of states.

• Σ is a finite set of symbols, called the alphabet of the automaton.
• δ is the transition function
• q0 is the initial state from where any input is processed (q0 ∈ Q).
• F is a set of final state/states of Q (F ⊆ Q).
38
38
An RE is one way of describing finite state automata (FSA). An RE is one Word Level Analysis - II
way of characterizing a particular kind of formal language called a regular
language. Regular grammar is the equivalent ways of describing regular
languages.
Relation between Finite Automata, Regular Grammars and Regular
Expressions
• As we know that finite state automata are the theoretical foundation

of computational work and regular expressions is one way of
describing them.
• We can say that any regular expression can be implemented as FSA
and any FSA can be described with a regular expression.
• On the other hand, regular expression is a way to characterize a kind
of language called regular language. Hence, we can say that regular
language can be described with the help of both FSA and regular
expression.
• Regular grammar, a formal grammar that can be right-regular or left-
regular, is another way to characterize regular language.
Figure 3.2.1 Relation between Finite Automata, Regular languages

and Regular Expressions
Finite state automation is of two types:
• Deterministic Finite automation (DFA)

• Non-deterministic Finite Automation (NDFA)
1. Deterministic Finite automation (DFA)
It may be defined as the type of finite automation wherein, for every input
symbol we can determine the state to which the machine will move. It has
a finite number of states that is why the machine is called Deterministic
Finite Automaton (DFA).
Mathematically, a DFA can be represented by a 5-tuple (Q, Σ, δ, q0, F),
where −
• Q is a finite set of states.

• Σ is a finite set of symbols, called the alphabet of the automaton.
• δ is the transition function where δ: Q × Σ → Q .
• q0 is the initial state from where any input is processed (q0 ∈ Q).
• F is a set of final state/states of Q (F ⊆ Q).
39
Natural Language Processing Whereas graphically, a DFA can be represented by diagraphs called state
diagrams where −
The states are represented by vertices.
The transitions are shown by labeled arcs.
The initial state is represented by an empty incoming arc.
The final state is represented by a double circle.
Construct the state diagram for the finite-state automata M= (Q, Σ, δ, q0,
F),, where Q = {q0, q1, q2, q3}, Σ = {0,1}, F= {q0, q3} and the transition
function δ is given in the following table.
State 0 1
q0 q0 q1
q1 q0 q2
q2 q0 q0
q3 q2 q1
q2
2. Non-deterministic Finite Automation (NDFA)

It may be defined as the type of finite automation where for every input
symbol we cannot determine the state to which the machine will move i.e.
the machine can move to any combination of the states. It has a finite
number of states that is why the machine is called Non-deterministic
Finite Automation (NDFA).
Mathematically, NDFA can be represented by a 5-tuple (Q, Σ, δ, q0, F),
where −
● Q is a finite set of states.
● Σ is a finite set of symbols, called the alphabet of the automaton.
δ :-is the transition function where δ: Q × Σ → 2 Q.
● q0 :-is the initial state from where any input is processed (q0 ∈ Q).
● F :-is a set of final state/states of Q (F ⊆ Q).
40
40
The finite state automaton M= (Q, Σ, q0, δ, F), where Q= {a, b, c }, Σ Word Level Analysis - II
={0,1}, F={c}, and δ is depicted graphically as follows:
State 0 1
a a, b b
b c a, c
c b, c c
3.3 FINITE STATE MORPHOLOGICAL PARSING
To know the structure about a word when we perform the morphological

parsing for that word. Given a surface form (input form), e.g., “going” we
might produce the parsed form: verb-go + gerund-ing. Morphological
parsing can be done with the help of finite-state transducer. A morpheme
is a meaning bearing unit of any language. For example, cats: has two
morphemes, cat, -s.
The objective of the morphological parsing is to produce output lexicons
for single input lexicon. Our aim is to take input forms like in the first
column of Table 3.3.1 and produce the output like in second column. The
second column contains the stem of each word as well as mixed
morphological features. These features include, like N (noun), V (verb),
specify additional information about the word stem, e.g., +N means that
word is noun, +SG means singular, +PL for plural, etc.
Table 3.3.1 Mapping of input word to Morphemes.
41
Natural Language Processing To build morphological parser, we need at least following database:
1. Lexicon: List of stems, and affixes, plus additional information
about them, like +N, +V.
2. Morphotactics rules: Rules about ordering of morphemes in a word,
e.g. -ed is followed after a verb (e.g., worked, studied), un (undo)
precede a verb, for example, unlock, untie, etc.
3. Orthographic rules (spelling): For combining morphemes, e.g., city+
-s gives cities and not citys.
3.4 BUILDING A FINITE-STATE LEXICON
A lexicon is a repository for words. They are grouped according to their

main categories like noun, adjective, verb, adverb etc. They may be
divided in to sub categories like regular-nouns, irregular-singular nouns,
irregular-plural nouns. The simplest possible lexicon would consist of an
explicit list of every word of the language (every word, i.e., including
abbreviations (“AAA”) and proper names (“Joy” or “Bebo”)) as follows:
a, AAA, AA, Aachen, aardvark, aardwolf, aba, abaca, aback,
The simplest way to create a morphological parser put all possible words
in to lexicon. As it’s often impossible, for the various reasons, to list every
word in the language, computational lexicons are usually structured with a
list of each of the stems and affixes of the language together with a
representation of the morphotactics that tells us how they can fit together.
There are many ways to model morphotactics; one of the most common is
the finite state automaton. A very simple finite-state model for English
nominal inflection might look like Fig. 3.4.1.
Figure 3.4.1 A finite-state automaton for English nominal inflection.

The FSA in Fig. 3.4.1 assumes that the lexicon includes regular nouns
(reg-noun) that take the regular -s plural (e.g. dog, fox, cat, and aardvark).
The lexicon also includes irregular noun forms that don’t take -s, both
singular irreg-sg-noun (goose, mouse) and plural irreg-pl-noun (geese,
mice).
42
42
reg-noun irreg-pl-noun irreg-sg-noun plural Word Level Analysis - II
fox cat geese sheep goose sheep -s

aardvark mice mouse
A similar model for English verbal inflection might look like Fig. 3.4.2
Figure 3.4.2 A finite-state automaton for English verbal inflection

This lexicon has three stem classes (reg-verb-stem, irreg-verb-stem, and
irreg-past- verb-form), plus four more affix classes (-ed past, -ed
participle, -ing participle, and third singular -s):
reg- irreg- irreg- past past- pres- 3sg

verb- verb- past- part part
stem stem verb
walk cut caught -ed -ed -ing -s

fry speak ate
talk sing eaten
impeac sang
h
English derivational morphology is significantly more complex than
English inflectional morphology, and so automata for modeling English
derivation tend to be quite complex. An initial hypothesis might be that
adjectives can have an optional prefix (un-), an obligatory root (big, cool,
etc.) and an optional suffix (-er, -est, or -ly). This might suggest the FSA
in Fig. 3.4.3.
big, bigger, biggest, cool, cooler, coolest, coolly
happy, happier, happiest, happily red, redder, reddest
unhappy, unhappier, unhappiest, unhappily real, unreal, really
clear, clearer, clearest, clearly, unclear, unclearly
43
Figure 3.4.3 An FSA for a fragment of English adjective morphology

An FSA for another fragment of English adjective morphology is shown
in Fig. 3.4.4.
Figure 3.4.4 An FSA for another fragment of English derivational

morphology.
Figure 3.4.5 Expanded FSA for a few English nouns with their
inflection
44
44
Expanded FSA for a few English nouns with their inflection is shown in Word Level Analysis - II
Fig, 3.4.5. Note that this automaton will incorrectly accept the input foxs.
3.5 FINITE STATE TRANSDUCERS (FST)
FSAs can be used to recognise particular patterns but don’t, by

themselves, allow for any analysis of word forms. Hence for morphology,
we use finite state transducers (FSTs) which allow the surface structure to
be mapped into the list of morphemes. FSTs are useful for both analysis
and generation, because the mapping is bidirectional. This approach is
known as two-level morphology.
FST is a type of finite automaton which maps between two sets of
symbols. We can visualize an FST as a two-tape automaton which
recognizes or generates pairs of strings. We can do this by labelling each
arc in the finite-state machine with two symbol strings, one from each
tape. Fig 3.5.1 shows an example of an FST where each arc is labelled by
an input and output string, separated by a colon.
Figure 3.5.1 A Finite-state transducer

FSA defines a formal language by defining a set of strings, an FST defines
a relation between sets of strings. In short, FST is as a machine that reads
one string and generates another.
Four-fold way of thinking about transducer:

i. FST as recognizer: a transducer that takes a pair of strings as input
and outputs accept if the string-pair is in the string-pair language,
and reject if it is not.
ii. FST as generator: a machine that outputs pairs of strings of the
language. Means the output is a yes or no, and a pair of output
strings.
iii. FST as translator: a machine that reads a string and outputs another
string
iv. FST as set relater: a machine that computes relations between sets.
For morphological parsing (and for many other NLP applications), we can
apply the FST as translator metaphor, taking as input a string of letters
and producing as output a string of morphemes.
45
Natural Language Processing Formal definition an FST can be defined with 7 parameters:
Q: a finite set of N states q0, q1, . . . , qN−1
Σ: a finite set corresponding to the input alphabet
Δ: a finite set corresponding to the output alphabet
q0 ∈ Q: start state
F ⊆ Q: set of final states
δ(q, w): the transition function between states; Given a state q ∈ Q and a
string w ∈ Σ∗,
δ(q, w) returns a set of new states Q′ ∈ Q. δ is thus a function from Q × Σ∗

to 2Q (because there are 2Q possible subsets of Q). δ returns a set of
states rather than a single state because a given input may be ambiguous
in which state it maps to.
σ(q, w): the output function giving the set of possible output strings for
each state and
input. Given a state q ∈ Q and a string w ∈ Σ∗, σ(q, w) gives a set of
output
strings, each a string o ∈ Δ∗. σ is thus a function from Q ×Σ∗ to 2Δ∗
Where FSAs are isomorphic to regular languages, FSTs are isomorphic to
regular relations. Regular relations are sets of pairs of strings, a natural
extension of the regular languages, which are sets of strings. Like FSAs
and regular languages, FSTs and regular relations are closed under union,
although in general they are not closed under difference, complementation
and.
Besides union, FSTs have two additional closure properties that turn out to
be extremely useful:
● Inversion: The inversion of a transducer T (T −1) simply switches
the input and output labels. Thus if T maps from the input alphabet I
to the output alphabet O, T −1 maps from O to I.
Inversion is useful because it makes it easy to convert a FST-as-
parser into an FST- as-generator.
T = {(a, a1), (a, a2), (b, b1), (c, c1), (c, a)}
T-1 = {(a1, a), (a2, a), (b1, b), (c1, c), (a, c)
● Composition: If T1 is a transducer from I1 to O1 and T2 a
transducer from O1 to O2, then T1 ◦ T2 maps from I1 to O2.
Composition is useful because it allows us to take two transducers that run
in series and replace them with one more complex transducer.
Composition works as in algebra; applying T1 ◦ T2 to an input sequence S
is identical to applying T1 to S and then T2 to the result; thus T1 ◦ T2(S) =
T2 (T1(S)).
46
46
Fig. 3.5.2, for example, shows the composition of [a:b]+ with [b:c]+ to
produce [a:c]+.
Figure 3.5.2 The composition of [a: b]+ with [b:c]+ to produce [a:c]+.
The projection of an FST is the FSA that is produced by extracting only
one side of the relation. We can refer to the projection to the left or upper
side of the relation as the upper or first projection and the projection to the
lower or right side of the relation as the lower or second projection
3.6 MORPHOLOGICAL PARSING WITH FST
Now we see the task of morphological parsing. If we give the input cats,
we’d like to output cat +N +Pl, telling us that cat is a plural noun
In the finite-state morphology paradigm, we represent a word as a
correspondence between a lexical level (which represents a concatenation
of morphemes making up a word) and the surface level (which represents
the concatenation of letters which make up the actual spelling of the
word). Fig. 3.6.1 shows these two levels for (English) cats.
Figure 3.6.1 Schematic examples of the lexical and surface tapes

The actual transducers consist of intermediate tapes also. For finite-state
morphology it’s convenient to view an FST as having two tapes. The
lexical tape is composed from characters from one alphabet Σ. The surface
tape is composed of characters from another alphabet Δ. In the two-level
morphology, we allow each arc only to have a single symbol from each
alphabet. We can then combine the two symbol alphabets Σ and Δ to
create a new alphabet, Σ′, which makes the relationship to FSAs quite
clear. Σ′ is a finite alphabet of complex symbols. Each complex symbol is
composed of an input-output pair i : o; one symbol i from the input
alphabet Σ, and one symbol o from an output alphabet Δ, thus Σ′ ⊆ Σ×Δ.
Thus where an FSA accepts a language stated over a finite alphabet of
single symbols,
such as the alphabet of our sheep language:
Σ = {b, a, ! }
an FST defined this way accepts a language stated over pairs of symbols,
as in: 47
Natural Language Processing Σ′ = {a : a, b : b, ! : !, a : !, a : ǫ, ǫ : !}
In two-level morphology, the pairs of symbols in Σ′ are also called feasible
pairs. Thus each feasible pair symbol a : b in the transducer alphabet Σ ′
expresses how the symbol a from one tape is mapped to the symbol b on
the other tape. For example a : ǫ means that an a on the lexical tape will
correspond to nothing on the surface tape. Just as for an FSA, we can write
regular expressions in the complex alphabet Σ′. Since it’s most common
for symbols to map to themselves, in two-level morphology we call pairs
like a : a default pairs, and just refer to them by the single letter a.
We can now build an FST morphological parser out of our earlier
morphotactic FSAs and lexica by adding an extra “lexical” tape and the
appropriate morphological features. Fig. 3.6.2 shows an augmentation of
Fig. 3.4.1 with the nominal morphological features (+Sg and +Pl) that
correspond to each morpheme. The symbol ˆ indicates a morpheme
boundary, while the symbol # indicates a word boundary. The
morphological features map to the empty string ǫ or the boundary symbols
since there is no segment corresponding to them on the output tape.
Figure 3.6.2 A schematic transducer for English nominal number

inflection Tnum.
The symbols above each arc represent elements of the morphological
parse in the lexical tape; the symbols below each arc represent the surface,
using the morpheme-boundary symbol ˆ and word-boundary marker #.
The arcs need to be expanded by individual words in the lexicon.
In order to use Fig. 3.6.2 as a morphological noun parser, it needs to be
expanded with all the individual regular and irregular noun stems,
replacing the labels reg-noun etc. In order to do this we need to update the
lexicon for this transducer, so that irregular plurals like geese will parse
into the correct stem goose +N +Pl. We do this by allowing the lexicon to
also have two levels. Since surface geese maps to lexical goose, the new
lexical entry will be “g:g o:e o:e s:s e:e”. Regular forms are simpler; the
two-level entry for fox will now be “f:f o:o x:x”, but by relying on the
orthographic convention that f stands for f:f and so on, we can simply refer
to it as fox and the form for geese as “g o:e o:e s e”. Thus the lexicon will
look only slightly more complex
48
48
reg-noun irreg-pl-noun irreg-sg-noun Word Level Analysis - II
fox cat g o:e o:e se goose sheep
dog sheep mouse
m o:i u:ǫ s:c e
Figure 3.6.3 A fleshed-out English nominal inflection FST Tlex,

expanded from Tnum by replacing the three arcs with individual
word stems (only a few sample word stems are shown).
The resulting transducer, shown in Fig 3.6.3 will map plural nouns into the
stem plus the morphological marker +Pl, and singular nouns into the stem
plus the morphological marker +Sg. Thus surface cats will map to cat +N
+Pl. This can be viewed in feasible-pair format as follows:
c:c a:a t:t +N:ǫ +Pl:ˆs#
Since the output symbols include the morpheme and word boundary
markers ˆ and #, the lower labels do not correspond exactly to the surface
level. Hence we refer to tapes with these morpheme boundary markers in
Fig. 3.6.4 as intermediate tapes; the next section will show how the
boundary marker is removed.
Figure 3.6.4 A schematic view of the lexical and intermediate tapes
3.7 LEXICON FREE FST PORTER STEMMER
While building a transducer from a lexicon plus rules is the standard

algorithm for morphological parsing, there are simpler algorithms that
don’t require the large on-line lexicon demanded by this algorithm. These
are used especially in Information Retrieval (IR) tasks like web search (a
49
Natural Language Processing user needs some information, and is looking for relevant documents) User
gives the system a query with some important characteristics of documents
she desires, and the IR system retrieves what it thinks are the relevant
documents. Morphological information in IR is thus only used to
determine that two words have the same stem; the suffixes are thrown
away.
One of the most widely used such stemming algorithms are the simple and
efficient Porter (1980) algorithm, which is based on a series of simple
cascaded rewrite rules. Since cascaded rewrite rules are just the sort of
thing that could be easily implemented as an FST, we think of the Porter
algorithm as a lexicon-free FST stemmer. The algorithm contains rules
like these:
ATIONAL → ATE (e.g., relational → relate)
ING → ǫ if stem contains vowel (e.g., motoring → motor)
Do stemmers really improve the performance of information retrieval,
especially with smaller documents? Nonetheless, not all IR engines use
stemming, partly because of stemmer errors. Following kinds of errors of
omission and commission in the Porter algorithm:
Errors of Commission Errors of Omission
Organization- organ European- Europe
doing - doe analysis- analyzes
generalization- generic matrices- matrix
numerical- numerous noise- noisy
policy- police sparse- sparsity
university- universe explain- explanation
3.8 N-GRAMS
N-gram is simply a sequence of N words. For instance, a 2-gram (or

bigram) is a two-word sequence of words like “please turn”, “turn your”,
or ”your homework”, and a 3-gram (or trigram) is a three-word sequence
of words like “please turn your”, or “turn your homework.”
In Natural Language Processing, n-grams are used for a variety of things.
Some examples include like Auto completion of sentences, Auto spell
check, Speech recognition and Machine translation.
How do N-grams work?
Let us take a look at the following examples.
• San Francisco (is a 2-gram)

• The Three Musketeers (is a 3-gram)
• She stood up slowly (is a 4-gram)
50
50
Now which of these three N-grams have you seen quite frequently? Word Level Analysis - II
Probably, “San Francisco” and “The Three Musketeers”. On the other
hand, you might not have seen “She stood up slowly” that frequently.
Basically, “She stood up slowly” is an example of an N-gram that does not
occur as often in sentences as Examples 1 and 2.
Now if we assign a probability to the occurrence of an N-gram or the
probability of a word occurring next in a sequence of words, it can be very
useful. Why?
First of all, it can help in deciding which N-grams can be chunked together
to form single entities (like “San Francisco” chunked together as one
word, “high school” being chunked as one word).
It can also help make next word predictions. Say you have the partial
sentence “Please hand over your”. Then it is more likely that the next
word is going to be “test” or “assignment” or “paper” than the next word
being “school”.
It can also help to make spelling error corrections. For instance, the
sentence “drink cofee” could be corrected to “drink coffee” if you knew
that the word “coffee” had a high probability of occurrence after the word
“drink” and also the overlap of letters between “cofee” and “coffee” is
high.
3.9 N –GRAM LANGUAGE MODEL
An N-gram model is a type of Language Model (LM), that focuses on

finding the probability distribution over word sequences. The model is
built by counting how often word sequences occur in corpus text and then
estimating the probabilities. Due to the fact that a simple N-gram model
has limitations, improvements are often made through the means of
smoothing, interpolation and backoff. A model that just focuses on how
frequently a word occurs without looking at previous words is called
unigram. If a model only considers just the previous word to predict the
current word, then that model is called a bigram.
When there is a sequence of N-1 words, an N-gram model is used to
predict the most probable word that would follow that sequence. The N-
gram model is a probabilistic model that is trained on a corpus of text.
Formal grammars (e.g. regular, context free) give a hard “binary” model
of the legal sentences in a language. For NLP, a probabilistic model of a
language that gives a probability that a string is a member of a language is
more useful. To specify a correct probability distribution, the probability
of all sentences in a language must sum to 1.
A language model also supports predicting the completion of a sentence.
• Please turn off your cell______________.

• Your program does not ______________.
51
Natural Language Processing Predictive text input systems can guess what you are typing and give
choices on how to complete it.
Here N-Gram Models ,Estimate probability of each word given prior
context.
• P (phone | Please turn off your cell)

Number of parameters required grows exponentially with the number of
words of prior context. An N-gram model uses only N-1 words of prior
context.
• Unigram: P(phone)
• Bigram: P(phone | cell)
• Trigram: P(phone | your cell)
The Markov assumption is the presumption that the future behaviour of a
dynamical system only depends on its recent history. In particular, in a
kth-order Markov model, the next state only depends on the k most recent
states, therefore an N-gram model is a (N-1)-order Markov model.
N-Gram Model Formulas
Word sequences
Chain rule of probability
Bigram approximation
N-gram approximation
Estimating Probabilities
N-gram conditional probabilities can be estimated from raw text based on
the relative frequency of word sequences.
52
52
Bigram: Word Level Analysis - II
N-gram:
• To have a consistent probabilistic model, append a unique start

(<s>) and end (</s>) symbol to every sentence and treat these as
additional words.
• An N-gram model can be seen as a probabilistic automata for
generating sentences.
Initialize sentence with N-1 <s> symbols
Until </s> is generated do:
Stochastically pick the next word based on the conditional
probability of each word given the previous N -1 words.
Example:
Let’s work through an example using a mini-corpus of three sentences
<s> I am Sam </s>
<s> Sam I am </s>
<s> I do not like green eggs and ham </s>
In general, we can say that this probability is (the number of times the
previous word ‘wp’ occurs before the word ‘wn’) / (the total number of
times the previous word ‘wp’ occurs in the corpus) =
(Count (wp wn))/(Count (wp))
To find the probability of the word “I” following the word “<s>”,we can
write this as
P (I | <s>) which is a conditional probability.
This becomes equal to:
= (No. of times “I” occurs) / (No. of times “<s>” occurs)
= 2/3
=0.67
Here are the calculations for some of the bigram probabilities from this
corpus
P (Sam | <s>) = 1/ 3 = 0.33
P (am | I) =2/3=0.67
P (</s> | Sam )=1/2 = 0.5
P (Sam | am =1/2 = 0.5
P (do | I) =1/3 = 0.33 53
Natural Language Processing 3.10 SUMMARY
In this chapter we studied concept of Regular expression (RE), Finite

Automata, and Finite State Transducers (FST). Finite-state transducers are
an extension of finite-state automata that can generate output symbols.
Important operations for FSTs include composition, projection, and
intersection. Finite-state morphology and two-level morphology are
applications of finite- state transducers to morphological representation
and parsing
After that we learned the Morphological Parsing with FST, Lexicon free
FST Porter Stemmer. Finally we studied N –Grams and N-gram language
model. An N-gram model is one type of a Language Model (LM), which
is about finding the probability distribution over word sequences.
i. FST cannot work as _____________

a) recognizer
b) generator
c) translator
d) lexicon
Ans : d
ii. N-Gram language models cannot be used for -------.
a) Spelling Correction
b) Predicting the completion of a sentence
c) Removing semantic ambiguity
d) Speech Recognition
Ans : c
iii. Both _____ and finite-state automata can be used to describe regular
languages
a) Language model
b) Deterministic Finite Automata
c) Regular expressions
d) Finite State Translators (FSTs)
Ans : c
iv. _______ is as a machine that reads one string and generates another
a) FSA
b) FST
c) FSM
d) FSI
Ans : b
54
54
v. ________is used to remove the suffixes from an English word and Word Level Analysis - II
obtain its stem which becomes very useful in the field of
Information Retrieval (IR).
a) HMM Stemmer
b) Porter Stemmer
c) Markov Stemmer
d) Bert Stemmer
Ans : b
3.12 TRUE OR FALSE
i. Regular expressions are case sensitive--- True

ii. FSA define Regular Language while FST define Regular Relation--
- True
iii. FST used in semantic analysis ----False
3.13 SAMPLE QUESTIONS
1. Explain the different forms of regular expressions used in NLP. Also

state the uses of Regular expressions.
2. What is Lexicon free FST Porter stemmer?
3. Write a note on Morphological parsing with FST
4. What are bi-grams, tri-grams and N-grams? Give examples.
5. Explain Finite state Automata. What are deterministic and non-
deterministic FSA?
3.14 LIST OF REFERENCES

Edition, 2009.
5. https://www.analyticsvidhya.com/blog/2021/06/part-13-step-by-step-
guide-to-master-nlp-regular-expressions/
6. https://www.cs.vassar.edu/~cs395/docs/3.pdf

55
4
SYNTAX ANALYSIS - I
Unit Structure
4.0 Syntax analysis
4.1 Objective
4.2 Introduction
4.3 Define Syntax analysis
4.3.1 Part-Of-Speech Tagging
4.3.2 Tag set for English (Penn Treebank)
4.4 Rule based POS Tagging
4.4.1 Stochastic POS Tagging
4.5 Issues-Multiple tags & words
4.5.1 Introduction to CFG
4.6 Sequence labeling
4.6.1 Hidden Markov Model (HMM)
4.6.2 Maximum Entropy
4.0 Syntax analysis
• Check if the code is valid grammatically

• The syntactical analyser helps you to apply rules to the code
• Helps you to make sure that each opening brace has a corresponding
closing balance
• Each declaration has a type and that the type must be exists.
4.1 Objective
This chapter would make you to give the idea of the goals of NLP.
Syntax Analysis is a second phase of the compiler design process in which
the given input string is checked for the confirmation of rules and structure
of the formal grammar. It analyses the syntactical structure and checks if
the given input is in the correct syntax of the programming language or
not.
• Syntax Analysis in Compiler Design process comes after the Lexical

analysis phase. It is also known as the Parse Tree or Syntax Tree.
The Parse Tree is developed with the help of pre-defined grammar
of the language. The syntax analyser also checks whether a given
program fulfills the rules implied by a context-free grammar. If it
satisfies, the parser then creates the parse tree of that source
program. Otherwise, it will display error messages.
56
56
• Why do you need Syntax Analyser? Syntax Analysis - I
• Important Syntax Analyser Terminology

• Why do we need Parsing?
• Parsing Techniques
• Error – Recovery Methods
• Grammar:
• Notational Conventions
• Context Free Grammar
• Grammar Derivation
• Syntax vs. Lexical Analyser
• Disadvantages of using Syntax Analysers
Why do you need Syntax Analyser?
• Check if the code is valid grammatically

• Helps you to make sure that each opening brace has a corresponding
closing balance
• Each declaration has a type and that the type must be exists
Important Syntax Analyser Terminology
Important terminologies used in syntax analysis process:
• Sentence: A sentence is a group of character over some alphabet.

• Lexeme: A lexeme is the lowest level syntactic unit of a language
(e.g., total, start).
• Token: A token is just a category of lexemes.
• Keywords and reserved words – It is an identifier which is used as
a fixed part of the syntax of a statement. It is a reserved word which
you can’t use as a variable name or identifier.
• Noise words – Noise words are optional which are inserted in a
statement to enhance the readability of the sentence.
• Comments – It is a very important part of the documentation. It
mostly display by, /* */, or//Blank (spaces)
• Delimiters – It is a syntactic element which marks the start or end of
some syntactic unit. Like a statement or expression, “begin”…”end”,
or {}.
• Character set – ASCII, Unicode
• Identifiers – It is a restrictions on the length which helps you to
reduce the readability of the sentence.
• Operator symbols – + and – performs two basic arithmetic
operations.
• Syntactic elements of the Language
57
Natural Language Processing Why do we need Parsing?
A parse also checks that the input string is well-formed, and if not, reject
it.
Following are important tasks perform by the parser in compiler design:
• Helps you to detect all types of Syntax errors

• Find the position at which error has occurred
• Clear & accurate description of the error.
58
58
• Recovery from an error to continue and find further errors in the Syntax Analysis - I
code.
• Should not affect compilation of “correct” programs.
• The parse must reject invalid texts by reporting syntax errors
Parsing Techniques
Parsing techniques are divided into two different groups:
• Top-Down Parsing,
• Bottom-Up Parsing
Top-Down Parsing:
In the top-down parsing construction of the parse tree starts at the root and
then proceeds towards the leaves.
Two types of Top-down parsing are:
1. Predictive Parsing:
Predictive parse can predict which production should be used to
replace the specific input string. The predictive parser uses look-
ahead point, which points towards next input symbols. Backtracking
is not an issue with this parsing technique. It is known as LL(1)
Parser
2. Recursive Descent Parsing:
This parsing technique recursively parses the input to make a prase
tree. It consists of several small functions, one for each nonterminal
in the grammar.
Bottom-Up Parsing:
In the bottom up parsing in compiler design, the construction of the parse
tree starts with the leave, and then it processes towards its root. It is also
called as shift-reduce parsing. This type of parsing in compiler design is
created with the help of using some software tools.
Error – Recovery Methods
Common Errors that occur in Parsing in System Software
• Lexical: Name of an incorrectly typed identifier

• Syntactical: unbalanced parenthesis or a missing semicolon
• Semantical: incompatible value assignment
• Logical: Infinite loop and not reachable code
A parser should able to detect and report any error found in the program.
So, whenever an error occurred the parser. It should be able to handle it
and carry on parsing the remaining input. A program can have following
59
Natural Language Processing types of errors at various compilation process stages. There are five
common error-recovery methods which can be implemented in the parser
Statement mode recovery
• In the case when the parser encounters an error, it helps you to take
corrective steps. This allows rest of inputs and states to parse ahead.
• For example, adding a missing semicolon is comes in statement
mode recover method. However, parse designer need to be careful
while making these changes as one wrong correction may lead to an
infinite loop.
Panic-Mode recovery
• In the case when the parser encounters an error, this mode ignores
the rest of the statement and not process input from erroneous input
to delimiter, like a semi-colon. This is a simple error recovery
method.
• In this type of recovery method, the parser rejects input symbols one
by one until a single designated group of synchronizing tokens is
found. The synchronizing tokens generally using delimiters like or.
Phrase-Level Recovery:
• Compiler corrects the program by inserting or deleting tokens. This

allows it to proceed to parse from where it was. It performs
correction on the remaining input. It can replace a prefix of the
remaining input with some string this helps the parser to continue
the process.
Error Productions
● Error production recovery expands the grammar for the language
which generates the erroneous constructs. The parser then performs
error diagnostic about that construct.
Global Correction:
● The compiler should make less number of changes as possible while
processing an incorrect input string. Given incorrect input string a
and grammar c, algorithms will search for a parse tree for a related
string b. Like some insertions, deletions, and modification made of
tokens needed to transform an into b is as little as possible.
Grammar:
A grammar is a set of structural rules which describe a language.
Grammars assign structure to any sentence. This term also refers to the
study of these rules, and this file includes morphology, phonology, and
syntax. It is capable of describing many, of the syntax of programming
languages.
60
60
Rules of Form Grammar Syntax Analysis - I
• The non-terminal symbol should appear to the left of the at least one
production
• The goal symbol should never be displayed to the right of the::= of
any production
• A rule is recursive if LHS appears in its RHS
Notational Conventions
Notational conventions symbol may be indicated by enclosing the element
in square brackets. It is an arbitrary sequence of instances of the element
which can be indicated by enclosing the element in braces followed by an
asterisk symbol, { … }*.
It is a choice of the alternative which may use the symbol within the single
rule. It may be enclosed by parenthesis ([,] ) when needed.
Two types of Notational conventions area Terminal and Non-terminals
1. Terminals:
• Lower-case letters in the alphabet such as a, b, c,

• Operator symbols such as +,-, *, etc.
• Punctuation symbols such as parentheses, hash, comma
• 0, 1, …, 9 digits
• Boldface strings like id or if, anything which represents a
single terminal symbol
2.Nonterminals:
• Upper-case letters such as A, B, C

• Lower-case italic names: the expression or some
Context Free Grammar
A CFG is a left-recursive grammar that has at least one production of the
type. The rules in a context-free grammar are mainly recursive. A syntax
analyser checks that specific program satisfies all the rules of Context-free
grammar or not. If it does meet, these rules syntax analysers may create a
parse tree for that programme.
expression -> expression -+ term
expression -> expression – term
expression-> term
term -> term * factor
term -> expression/ factor
term -> factor factor
factor -> ( expression )
factor -> id
61
Natural Language Processing Grammar Derivation
Grammar derivation is a sequence of grammar rule which transforms the
start symbol into the string. A derivation proves that the string belongs to
the grammar’s language.
Left-most Derivation
When the sentential form of input is scanned and replaced in left to right
sequence, it is known as left-most derivation. The sentential form which is
derived by the left-most derivation is called the left-sentential form.
Right-most Derivation
Rightmost derivation scan and replace the input with production rules,
from right to left, sequence. It’s known as right-most derivation. The
sentential form which is derived from the rightmost derivation is known as
right-sentential form.
Syntax vs. Lexical Analyser
Syntax Analyser
The syntax analyser mainly deals with recursive constructs of the
language.
The syntax analyser works on tokens in a source program to recognize
meaningful structures in the programming language.
It receives inputs, in the form of tokens, from lexical analysers.
Lexical Analyser
The lexical analyser eases the task of the syntax analyser.
The lexical analyser recognizes the token in a source program.
It is responsible for the validity of a token supplied by the syntax analyser
Disadvantages of using Syntax Analysers
• It will never determine if a token is valid or not

• Not helps you to determine if an operation performed on a token
type is valid or not
• You can’t decide that token is declared & initialized before it is
being used
Summary
• Syntax analysis is a second phase of the compiler design process that

comes after lexical analysis
• Sentence, Lexeme, Token, Keywords and reserved words, Noise
words, Comments, Delimiters, Character set, Identifiers are some
important terms used in the Syntax Analysis in Compiler
construction
62
62
• Parse checks that the input string is well-formed, and if not, reject it Syntax Analysis - I
• Parsing techniques are divided into two different groups: Top-Down

Parsing, Bottom-Up Parsing
• Lexical, Syntactical, Semantical, and logical are some common
errors occurs during parsing method
• A grammar is a set of structural rules which describe a language
• Notational conventions symbol may be indicated by enclosing the
element in square brackets
• A CFG is a left-recursive grammar that has at least one production
of the type
• Grammar derivation is a sequence of grammar rule which transforms
the start symbol into the string
• The syntax analyser mainly deals with recursive constructs of the
language while the lexical analyser eases the task of the syntax
analyser in DBMS
• The drawback of Syntax analyser method is that it will never
determine if a token is valid or not
4.2 Introduction-The main problems of language processing will be
discussed with examples from NLP applications. This chapter will also
introduce some methodological distinctions and place the applications and
methodology like what is NLP?
Natural language processing (NLP) can be defined as the computer-based
modeling of human language. The term 'NLP' is sometimes used more
narrowly and often excludes information retrieval and sometimes even
machine translation.
Natural language processing (NLP) is the intersection of computer
science, linguistics and machine learning. The field focuses on
communication between computers and humans in natural language and
NLP is all about making computers understand and generate human
language. Applications of NLP techniques include voice assistants like
Amazon's Alexa and Apple's Siri, but also things like machine translation
and text-filtering.
NLP has heavily benefited from recent advances in machine learning,
especially from deep learning techniques. The field is divided into the
three parts:
● Speech Recognition — The translation of spoken language into text.
● Natural Language Understanding — The computer's ability to
understand what we say.
● Natural Language Generation — The generation of natural
language by a computer.
63
Natural Language Processing II. Why NLP is difficult
Human language is special for several reasons. It is specifically
constructed to convey the speaker/writer's meaning. It is a complex
system, although little children can learn it pretty quickly.
Another remarkable thing about human language is that it is all about
symbols. According to Chris Manning, a machine learning professor at
Stanford, it is a discrete, symbolic, categorical signaling system. This
means we can convey the same meaning in different ways (i.e., speech,
gesture, signs, etc.) The encoding by the human brain is a continuous
pattern of activation by which the symbols are transmitted via continuous
signals of sound and vision.
Understanding human language is considered a difficult task due to its
complexity. For example, there is an infinite number of different ways to
arrange words in a sentence. Also, words can have several meanings and
contextual information is necessary to correctly interpret sentences. Every
language is more or less unique and ambiguous. Just take a look at the
following newspaper headline "The Pope’s baby steps on gays." This
sentence clearly has two very different interpretations, which is a pretty
good example of the challenges in NLP.
Note that a perfect understanding of language by a computer would result
in an AI that can process the whole information that is available on the
internet, which in turn would probably result in artificial general
intelligence.
NLP is sometimes compared to 'computational linguistics', although NLP
is considered more applied. Nowadays, alternative terms are often
preferred, such as 'Language Technology' or 'Language Engineering'. The
term 'language' is often used in contrast to 'speech' (e.g. Speech and
Language Technology). However, I will simply refer to NLP and use the
term in a broader sense.
4.3 Define Syntax analysis
Syntax Analysis in Compiler Design process comes after the Lexical

analysis phase. It is also known as the Parse Tree or Syntax Tree. The
Parse Tree is developed with the help of pre-defined grammar of the
64
64 language. The syntax analyser also checks whether a given program
fulfills the rules implied by a context-free grammar. If it satisfies, the Syntax Analysis - I
parser then creates the parse tree of that source program. Otherwise, it will
display error messages.
Tagging is a kind of classification that may be defined as the automatic
assignment of description to the tokens. Here the descriptor is called tag,
which may represent one of the part-of-speech, semantic information and
so on.
What Is Natural Language Processing? A Gentle Introduction to NLP
An introduction to Natural Language Processing: what it is, how do we
use it, what challenges we face and what tools we have.
Natural Language Processing, or NLP is a subfield of Artificial
Intelligence research that is focused on developing models and points of
interaction between humans and computers based on natural language.
This includes text, but also speech-based systems.
Computer scientists and researchers have been studying this topic for
entire decades, but only recently has it become a hot topic again, a
situation made possible by recent breakthroughs in the research
community.
Having this said, here’s what you’ll learn by the end of this article.
• A basic overview of how Natural Language Processing works

• Who uses Natural Language Processing and for what kind of
problems
• What are the challenges of using Natural Language Processing for
your app or for your business
• What are some basic tools that can get you started working with
How Natural Language Processing works
A generally accepted truth in computer science is that every complex
problem becomes easier to solve if we break it into smaller pieces. That is
especially true in the Artificial Intelligence field. For a given problem, we
build several small, highly specialized components that are good at
solving one and only one problem. We then align all this components, we
pass our input through each component and we get our output at the end of
the line. This is what we call a pipeline.
In the NLP context, a basic problem would be that for a given paragraph,
the computer understands exactly the meaning of it and then possibly it
acts accordingly. For this to work, we need to go through a few steps.
Sentence boundary segmentation
For a given text, we need to correctly identify every sentence, so that each
sentence resulted from this will have its meaning extracted in the next
steps. It seems that extracting the meaning from every sentence in a text 65
Natural Language Processing and then putting it all together is more accurate than trying to identify the
meaning of the whole text. After all, when we speak (or write) we don’t
necessarily mean only one thing. We often tend to convey more ideas into
one and the beauty of natural language (and the curse for NLP) is that we
actually can.
A naive approach for this would be to only search for periods in a chunk
of text and define that as the end of a sentence. The problem is that periods
can also be used for other purposes(for example, abbreviations) so in
practice machine learning models have been defined to correctly identify
the punctuation marks that are used for ending sentences.
Word tokenization
This part involves taking a sentence from the previous step and breaking it
into a list of all the words (and punctuation marks) it contains. This will be
used in the next steps to perform an analysis on each word.
Part of Speech Tagging
This step involved taking each word from the previous step and classify it
as to what part of speech it represents. This is an essential step for
identifying the meaning behind a text. Identifying the nouns allows us to
figure out who or what the given text is about. Then the verbs and
adjectives let us understand what entities do, or how they are described, or
any other meaning we can get from a text. PoS Tagging is a difficult
problem but it has mostly been solved and implementations for this can be
found in most of the modern Machine Learning libraries and tools.
Named Entity Recognition — NER
This task refers to identifying the names in a sentence and correctly
classify it against a list of predefined categories. Such categories may
involve: Persons, Organizations, Locations, Time, Quantities and so on.
Lists of categories may be tailored for your own particular use case, but in
general, almost everybody needs at least these categories to be correctly
identified.
There are many implementations for this kind of problem and models
build recently have achieved near-human performance. Basically, this step
is also divided in two subtasks: correctly identify the names in a sentence
and then classify each name according to your list of categories.
There are of course many other tasks which NLP solves in order to be
used in real world applications, but the next steps are tailored to every use
case and every business need. Having all these said, the steps previously
presented are the basic steps in almost every use case we can think of.
66
66
How Natural Language Processing is used Syntax Analysis - I
NLP is used in a variety of software and various use cases have been
identified as being solvable by deploying NLP models. Some of these
examples are:
• Virtual assistants: Most modern, powerful personal assistants

employ a large suite of NLP techniques in order to help users
accomplish their tasks. Siri, the Google Assistant and many others
have become very efficient and highly skilled in helping their users
by using the latest breakthroughs in the NLP field. This means that
the companies behind them invest large amounts of resources in
further research in the race to developing the perfect virtual
assistant. While until a few years ago, these assistants were more
fancy than useful, nowadays millions of users are using them and are
able to take advantage of their solutions.
• Machine translation: Have you noticed that in the recent years,

Google Translate has become more accurate in translating texts for
you? This is thanks to latest advances achieved by the research in
this field, with Google Translate powering machine translation in
hundreds of languages for hundreds of millions of users.
• Speech to Text: In the fast-paced society that we live in today, we

often don’t have time to record in writing everything that we
discuss, be it business notes, phone calls, speeches and so on. There
are quite a handful of startups nowadays which help us with these
kinds of problems, namely taking sound as an input and providing us
with the text, from which we can carry on and take actions based on
that.
Challenges of using Natural Language Processing
There are, of course, quite a few challenges when using NLP techniques.
The most important are related to extracting context from the text.
Humans are very good at understanding the context of a sentence, but
67
Natural Language Processing computers only use statistical models to represent these sentences, so it is
very hard for them to understand exactly what we mean by our words.
For example, when we say “bush” we may refer to the plant or to the
former US President. Identifying the difference may be very easy for you,
but a computer will have to go through several steps from the pipeline
until deciding which meaning you are using.
A further challenge of NLP is related to the first one and regards finding
enough data to train our model. As with any other Machine Learning
models, training an NLP models takes a lot of data and a lot of time,
meaning there are only a handful of large enough companies that have the
resources to build truly powerful applications involving NLP. Other,
smaller companies are employing machine learning models that are highly
specialized, meaning they solve only a subset of all the NLP problems, for
which they need considerably less data.
Having considered all of these, it is important at the end of the day that we
correctly identify the value that an NLP model brings to our business or
our app. We need to see if the model we’ve managed to build can truly
help us and our customers, or if it’s just a fancy feature so that we can say
we are a machine learning company.
Natural Language Processing tools
There are quite a few libraries available for developers in order to start
learning about and developing NLP models. And the good thing is most of
them are open-source.
NLTK — Natural Language Toolkit is a Python library which is mainly
used in research and education. It includes many steps in the pipeline with
models which are ready to test and use.
Stanford CoreNLP is another Python library which provides a wide
range of tools for understanding natural language.
spaCy is yet another Python library which employs machine learning and
deep learning to help you with a lot of powerful feature.

68
68
5
SYNTAX ANALYSIS - II
Unit Structure
5.0 Part-Of-Speech Tagging
5.1 What are the parts of Speech?
5.2 Tag set for English (Penn Treebank)
5.3 Rule-based POS Tagging
5.0 Part-Of-Speech Tagging
Tagging is a kind of classification that may be defined as the automatic

assignment of description to the tokens. Here the descriptor is called tag,
which may represent one of the part-of-speech, semantic information and
so on.
Now, if we talk about Part-of-Speech (PoS) tagging, then it may be
defined as the process of assigning one of the parts of speech to the given
word. It is generally called POS tagging. In simple words, we can say that
POS tagging is a task of labelling each word in a sentence with its
appropriate part of speech. We already know that parts of speech include
nouns, verb, adverbs, adjectives, pronouns, conjunction and their sub-
categories.
Most of the POS tagging falls under Rule Base POS tagging, Stochastic
POS tagging and Transformation based tagging.
5.1 What are the parts of Speech?
POS tagging words often have more than one POS: - The back door
(adjective) - On my back (noun) - Winning back the voters (particle) -
Promised to support the bill (verb) The POS tagging task: determine the
POS tag for all tokens in a sentence. Due to ambiguity (and unknown
words), we cannot rely on a dictionary to look up the correct POS tags.
Why POS tagging?

POS Tagging is one of the first steps in the NLP pipeline (right after
tokenization, segmentation). POS tagging is traditionally considered a
prerequisite for further analysis:
Syntactic parsing: what words are in the sentence?
69
Natural Language Processing Information extraction: finding names, dates, relationships, etc. NB:
Although many neural models do not use POS tagging, it is still important
to understand what makes POS tagging difficult (or easy) and how the
basic models and algorithms work.
Creating a POS Tagger :To deal with ambiguity and coverage, POS
taggers rely on learned models. For a new language (or domain) Step 0:
Define a POS tag set Step 1: Annotate a corpus with these tags For a well-
researched language (and domain):
Step 1: Obtain a POS -tagged corpus For any language....:
Step 2: Choose a POS tagging model (e.g., an HMM) Step 3: Train your
model on your training corpus Step 4: Evaluate your model on your test
corpus
Define a tag set We need to define an inventory of labels for the word
classes (i.e., the tag set) - Most taggers are based on models that need to be
trained on annotated (tagged) corpora.
- Evaluation also requires annotated corpora. - Since annotation by
humans is expensive and time-consuming, tag sets used in a few existing
corpora are becoming the de facto standard. - Tag sets must capture
semantically or syntactically important distinctions that can be easily
made by trained human annotators.
5.2 Tag set for English (Penn Treebank)
In linguistics, a Treebank is a parsed text corpus that annotates syntactic or

semantic sentence structure. The construction of parsed corpora in the
early 1990s revolutionized computational linguistics, which benefitted
from large-scale empirical data.
70
70
What is Treebank NLP? Syntax Analysis - II
A Treebank is a collection of syntactically annotated sentences in

which the annotation has been manually checked so that the Treebank
can serve as a training corpus for natural language parsers, as a repository
for linguistic research, or as an evaluation corpus for NLP systems.
71
Natural Language Processing 5.3 Rule-based POS Tagging
One of the oldest techniques of tagging is rule-based POS tagging. Rule-

based taggers use dictionary or lexicon for getting possible tags for
tagging each word. If the word has more than one possible tag, then rule-
based taggers use hand-written rules to identify the correct tag.
Disambiguation can also be performed in rule-based tagging by analyzing
the linguistic features of a word along with its preceding as well as
following words. For example, suppose if the preceding word of a word is
article then word must be a noun.
As the name suggests, all such kind of information in rule-based POS
tagging is coded in the form of rules. These rules may be either −
• Context-pattern rules
• Or, as Regular expression compiled into finite-state automata,
intersected with lexically ambiguous sentence representation.
We can also understand Rule-based POS tagging by its two-stage
architecture −
• First stage − In the first stage, it uses a dictionary to assign each

word a list of potential parts-of-speech.
• Second stage − In the second stage, it uses large lists of hand-written
disambiguation rules to sort down the list to a single part-of-speech
for each word.
Properties of Rule-Based POS Tagging
Rule-based POS taggers possess the following properties −
• These taggers are knowledge-driven taggers.

• The rules in Rule-based POS tagging are built manually.
• The information is coded in the form of rules.
• We have some limited number of rules approximately around 1000.
• Smoothing and language modeling is defined explicitly in rule-based
taggers.
5.3.1 Stochastic POS Tagging
Another technique of tagging is Stochastic POS Tagging. Now, the
question that arises here is which model can be stochastic. The model that
includes frequency or probability (statistics) can be called stochastic. Any
number of different approaches to the problem of part-of-speech tagging
can be referred to as stochastic tagger.
The simplest stochastic tagger applies the following approaches for POS
tagging −
72
72
Word Frequency Approach Syntax Analysis - II
In this approach, the stochastic taggers disambiguate the words based on

the probability that a word occurs with a particular tag. We can also say
that the tag encountered most frequently with the word in the training set
is the one assigned to an ambiguous instance of that word. The main issue
with this approach is that it may yield inadmissible sequence of tags.
Tag Sequence Probabilities
It is another approach of stochastic tagging, where the tagger calculates
the probability of a given sequence of tags occurring. It is also called n-
gram approach. It is called so because the best tag for a given word is
determined by the probability at which it occurs with the n previous tags.
Properties of Stochastic POST Tagging
Stochastic POS taggers possess the following properties −
• This POS tagging is based on the probability of tag occurring.

• It requires training corpus
• There would be no probability for the words that do not exist in the
corpus.
• It uses different testing corpus (other than training corpus).
• It is the simplest POS tagging because it chooses most frequent tags
associated with a word in training corpus.
Transformation-based Tagging
Transformation based tagging is also called Brill tagging. It is an instance
of the transformation-based learning (TBL), which is a rule-based
algorithm for automatic tagging of POS to the given text. TBL, allows us
to have linguistic knowledge in a readable form, transforms one state to
another state by using transformation rules.
It draws the inspiration from both the previous explained taggers − rule-
based and stochastic. If we see similarity between rule-based and
transformation tagger, then like rule-based, it is also based on the rules
that specify what tags need to be assigned to what words. On the other
hand, if we see similarity between stochastic and transformation tagger
then like stochastic, it is machine learning technique in which rules are
automatically induced from data.
Working of Transformation Based Learning(TBL)
In order to understand the working and concept of transformation-based
taggers, we need to understand the working of transformation-based
learning. Consider the following steps to understand the working of TBL
• Start with the solution − The TBL usually starts with some
solution to the problem and works in cycles.
73
Natural Language Processing • Most beneficial transformation chosen − In each cycle, TBL will
choose the most beneficial transformation.
• Apply to the problem − The transformation chosen in the last step
will be applied to the problem.
The algorithm will stop when the selected transformation in step 2 will not
add either more value or there are no more transformations to be selected.
Such kind of learning is best suited in classification tasks.
Advantages of Transformation-based Learning (TBL)
The advantages of TBL are as follows −
• We learn small set of simple rules and these rules are enough for
tagging.
• Development as well as debugging is very easy in TBL because the
learned rules are easy to understand.
• Complexity in tagging is reduced because in TBL there is interlacing
of machinelearned and human-generated rules.
• Transformation-based tagger is much faster than Markov-model
tagger.
Disadvantages of Transformation-based Learning (TBL)
The disadvantages of TBL are as follows −
• Transformation-based learning (TBL) does not provide tag

probabilities.
• In TBL, the training time is very long especially on large corpora.
The main problem with POS tagging is ambiguity. In English, many

common words have multiple meanings and therefore multiple POS . The
job of a POS tagger is to resolve this ambiguity accurately based on the
context of use. For example, the word "shot" can be a noun or a verb
3.4.1. Introduction to context-free grammar (CFG)
Definition − A context-free grammar (CFG) consisting of a finite set of
grammar rules is a quadruple (N, T, P, S) where
• N is a set of non-terminal symbols.
• T is a set of terminals where N ∩ T = NULL.
• P is a set of rules, P: N → (N ∪ T)*, i.e., the left-hand side of the

production rule P does have any right context or left context.
• S is the start symbol.
74
74
Example Syntax Analysis - II
• The grammar ({A}, {a, b, c}, P, A), P : A → aA, A → abc.

• The grammar ({S, a, b}, {a, b}, P, S), P: S → aSa, S → bSb, S → ε
• The grammar ({S, F}, {0, 1}, P, S), P: S → 00S | 11F, F → 00F | ε
Generation of Derivation Tree
A derivation tree or parse tree is an ordered rooted tree that graphically
represents the semantic information a string derived from a context-free
grammar.
Representation Technique
• Root vertex − Must be labeled by the start symbol.

• Vertex − Labeled by a non-terminal symbol.
• Leaves − Labeled by a terminal symbol or ε.
If S → x1x2 …… xn is a production rule in a CFG, then the parse tree /
derivation tree will be as follows –
There are two different approaches to draw a derivation tree −

Top-down Approach −
• Starts with the starting symbol S

• Goes down to tree leaves using productions
Bottom-up Approach −
• Starts from tree leaves

• Proceeds upward to the root which is the starting symbol S
We have seen that language models estimate the probability of a sequence.

Unlike text classification, we are not predicting any labels so it’s an
75
Natural Language Processing example of unsupervised learning. This week we will look at supervised
learning with sequences. In many NLP tasks, we want to produce a
sequence conditioned on another sequence. Sequence labeling is perhaps
the simplest form of a sequence-to-sequence task. Here our goal is to
predict a label to each token in the input sequence.
Let’s use part-of-speech tagging as a running example. The input a
sequence of tokens (in a sentence), and we want to tag each token by its
grammatical category such as nouns and verbs. For example,
Language/NOUN is/VERB fun/ADJ ./PUNCT
A first thought might be to simply build a word-to-tag dictionary.
However, many words can be assigned multiple tags depending on the
context. In the above example, “fun” can be either a noun (as in “have
fun”) or an adjective (as in “a fun evening”). Therefore, the model must
take context into consideration.
3.5.1 Hidden Markov Model (HMM)
HMMs are the most commonly used generative models for POS tagging
(and other tasks, e.g. in speech recognition)
HMMs make specific independence assumptions in P(t) and P(w| t):
These probabilities don’t depend on the position in the sentence (i), but are
defined over word and tag types.
With subscripts i,j,k, to index word/tag types, they become P(ti | tj), P(ti | tj,
tk), P(wi | tj)
76
76
Syntax Analysis - II
Hidden Markov Model (HMM) POS Tagging

Before digging deep into HMM POS tagging, we must understand the
concept of Hidden Markov Model (HMM).
Hidden Markov Model
An HMM model may be defined as the doubly-embedded stochastic
model, where the underlying stochastic process is hidden. This hidden
stochastic process can only be observed through another set of stochastic
processes that produces the sequence of observations.
Example
For example, a sequence of hidden coin tossing experiments is done and
we see only the observation sequence consisting of heads and tails. The
actual details of the process - how many coins used, the order in which
they are selected - are hidden from us. By observing this sequence of
heads and tails, we can build several HMMs to explain the sequence.
Following is one form of Hidden Markov Model for this problem −
77
Natural Language Processing We assumed that there are two states in the HMM and each of the state
corresponds to the selection of different biased coin. Following matrix
gives the state transition probabilities −
A=[a11a21a12a22]A=[a11a12a21a22]
Here,
• aij = probability of transition from one state to another from i to j.
• a11 + a12 = 1 and a21 + a22 =1
• P1 = probability of heads of the first coin i.e. the bias of the first
coin.
• P2 = probability of heads of the second coin i.e. the bias of the
second coin.
We can also create an HMM model assuming that there are 3 coins or
more.
This way, we can characterize HMM by the following elements −
• N, the number of states in the model (in the above example N =2,
only two states).
• M, the number of distinct observations that can appear with each
state in the above example M = 2, i.e., H or T).
• A, the state transition probability distribution − the matrix A in the
above example.
• P, the probability distribution of the observable symbols in each
state (in our example P1 and P2).
• I, the initial state distribution.
Use of HMM for POS Tagging
The POS tagging process is the process of finding the sequence of tags
which is most likely to have generated a given word sequence. We can
model this POS process by using a Hidden Markov Model (HMM),
where tags are the hidden states that produced the observable
output, i.e., the words.
Mathematically, in POS tagging, we are always interested in finding a tag
sequence (C) which maximizes −
P (C|W)
Where,
C = C1, C2, C3... CT
W = W1, W2, W3, WT
On the other side of coin, the fact is that we need a lot of statistical data to
reasonably estimate such kind of sequences. However, to simplify the
problem, we can apply some mathematical transformations along with
some assumptions.
78
78
The use of HMM to do a POS tagging is a special case of Bayesian Syntax Analysis - II
interference. Hence, we will start by restating the problem using Bayes’
rule, which says that the above-mentioned conditional probability is equal
to −
(PROB (C1, CT) * PROB (W1, WT | C1, CT)) / PROB (W1, WT)
We can eliminate the denominator in all these cases because we are
interested in finding the sequence C which maximizes the above value.
This will not affect our answer. Now, our problem reduces to finding the
sequence C that maximizes −
PROB (C1, CT) * PROB (W1, WT | C1, CT) (1)
Even after reducing the problem in the above expression, it would require
large amount of data. We can make reasonable independence assumptions
about the two probabilities in the above expression to overcome the
problem.
First Assumption
The probability of a tag depends on the previous one (bigram model) or
previous two (trigram model) or previous n tags (n-gram model) which,
mathematically, can be explained as follows −
PROB (C1,..., CT) = Πi=1..T PROB (Ci|Ci-n+1…Ci-1) (n-gram model)
PROB (C1,..., CT) = Πi=1..T PROB (Ci|Ci-1) (bigram model)
The beginning of a sentence can be accounted for by assuming an initial
probability for each tag.
PROB (C1|C0) = PROB initial (C1)
Second Assumption
The second probability in equation (1) above can be approximated by
assuming that a word appears in a category independent of the words in
the preceding or succeeding categories which can be explained
mathematically as follows −
PROB (W1,..., WT | C1,..., CT) = Πi=1..T PROB (Wi|Ci)
Now, on the basis of the above two assumptions, our goal reduces to
finding a sequence C which maximizes
Πi=1...T PROB(Ci|Ci-1) * PROB(Wi|Ci)
Now the question that arises here is has converting the problem to the
above form really helped us. The answer is - yes, it has. If we have a large
tagged corpus, then the two probabilities in the above formula can be
calculated as −
PROB (Ci=VERB|Ci-1=NOUN) = (# of instances where Verb follows Noun) /
(# of instances where Noun appears) (2) 79
Natural Language Processing PROB (Wi|Ci) = (# of instances where Wi appears in Ci) /(# of
instances where Ci appears) (3)
80
80
3.5.2 Maximum Entropy Syntax Analysis - II
What is Max entropy model?
The maximum entropy principle (MaxEnt) states that the most appropriate
distribution to model a given set of data is the one with highest entropy
among all those that satisfy the constrains of our prior knowledge.
Usually, these constrains are given as equations regarding moments of the
desired distribution.
Reference pdf :
• https://web.stanford.edu/group/cslipublications/cslipublications/pdf/
1575863480.pdf
• https://www.researchgate.net/publication/270568943_Syntax_Work
book
• https://linguistics.ucla.edu/people/stabler/isat.pdf

81
6
SEMANTIC ANALYSIS - I
Unit Structure
6.0 Objective
6.1 Introduction
6.1.2 Definition of Semantic Analysis
6.2 Lexical Semantics
6.2.1 Attachment for fragment of English-sentences,
6.2.2 Noun phrases
6.3 Verb phrases
6.3.1 Prepositional phrases
6.4 Relations among lexemes & their senses
6.4.2 -Homonymy, Polysemy, Synonymy
6.5 Homonymy, Robust, Word Senses
6.5.1 Disambiguation (WSD)
6.5.2 Dictionary based approach
6.0 OBJECTIVE
What is the role of semantic analysis in NLP?

The purpose of semantic analysis is to draw exact meaning, or you can
say dictionary meaning from the text. The work of semantic analyzer is
to check the text for meaningfulness.
82
82
What is semantic analysis used for? Semantic Analysis - I
Simply semantic analysis is the process of drawing meaning from text.

It allows computers to understand and interpret sentences, paragraphs, or
whole documents, by analyzing their grammatical structure, and
identifying relationships between individual words in a particular context.
Semantic Analysis
Michael L. Scott, in Programming Language Pragmatics (Third Edition),
2009
One-Pass Compilers
A compiler that interleaves semantic analysis and code generation
with parsing is said to be a one-pass compiler.4 It is unclear whether
interleaving semantic analysis with parsing makes a compiler simpler or
more complex; it's mainly a matter of taste. If intermediate code
generation is interleaved with parsing, one need not build a syntax tree at
all (unless of course the syntax tree is the intermediate code). Moreover, it
is often possible to write the intermediate code to an output file on the fly,
rather than accumulating it in the attributes of the root of the parse tree.
The resulting space savings were important for previous generations of
computers, which had very small main memories. On the other hand,
semantic analysis is easier to perform during a separate traversal of a
syntax tree, because that tree reflects the program's semantic
structure better than the parse tree does, especially with a top-down parser,
and because one has the option of traversing the tree in an order other than
that chosen by the parser.
Decorating a Syntax Tree
In our discussion so far we have used attribute grammars solely to
decorate parse trees. As we mentioned in the chapter introduction, 83
Natural Language Processing attribute grammars can also be used to decorate syntax trees. If our
compiler uses action routines simply to build a syntax tree, then the bulk
of semantic analysis and intermediate code generation will use the syntax
tree as base.
What is semantic analysis in natural language processing?
Semantic analysis describes the process of understanding natural
language–the way that humans communicate–based on meaning and
context. ... It analyzes context in the surrounding text and it analyzes the
text structure to accurately disambiguate the proper meaning of words that
have more than one definition.
What is semantic analysis language?
In linguistics, semantic analysis is the process of relating syntactic
structures, from the levels of phrases, clauses, sentences and paragraphs
to the level of the writing as a whole, to their language-independent
meanings. ... Semantic analysis can begin with the relationship between
individual words.
How important is semantics?
However, in truth, semantics is a very important subject. Semantics is the
study of the meaning of words. Many words have very similar meanings
and it is important to be able to distinguish subtle differences between
them. ... ' It is also important for children to know the opposite meanings
of words (antonyms).
What is meant by semantic features?
Semantic features are theoretical units of meaning-holding components
which are used for representing word meaning. These features play a
vital role in determining the kind of lexical relation which exists between
words in a language.
What are examples of semantics?
semantics Add to list Share. Semantics is the study of meaning in
language. It can be applied to entire texts or to single words. For
example, "destination" and "last stop" technically mean the same thing,
but students of semantics analyze their subtle shades of meaning.
We have learnt how a parser constructs parse trees in the syntax analysis
phase. The plain parse-tree constructed in that phase is generally of no use
for a compiler, as it does not carry any information of how to evaluate the
tree. The productions of context-free grammar, which makes the rules of
the language, do not accommodate how to interpret them.
For example
E→E+T
84
84
The above CFG production has no semantic rule associated with it, and it Semantic Analysis - I
cannot help in making any sense of the production.
Semantics
Semantics of a language provide meaning to its constructs, like tokens and
syntax structure. Semantics help interpret symbols, their types, and their
relations with each other. Semantic analysis judges whether the syntax
structure constructed in the source program derives any meaning or not.
CFG + semantic rules = Syntax Directed Definitions
For example:
int a = “value”;
should not issue an error in lexical and syntax analysis phase, as it is

lexically and structurally correct, but it should generate a semantic error as
the type of the assignment differs. These rules are set by the grammar of
the language and evaluated in semantic analysis. The following tasks
should be performed in semantic analysis:
• Scope resolution
• Type checking
• Array-bound checking
Semantic Errors
We have mentioned some of the semantics errors that the semantic
analyzer is expected to recognize:
• Type mismatch
• Undeclared variable
• Reserved identifier misuse.
• Multiple declaration of variable in a scope.
• Accessing an out of scope variable.
• Actual and formal parameter mismatch.
Attribute Grammar
Attribute grammar is a special form of context-free grammar where some
additional information (attributes) are appended to one or more of its non-
terminals in order to provide context-sensitive information. Each attribute
has well-defined domain of values, such as integer, float, character, string,
and expressions.
Attribute grammar is a medium to provide semantics to the context-free
grammar and it can help specify the syntax and semantics of a
programming language. Attribute grammar (when viewed as a parse-tree)
can pass values or information among the nodes of a tree.
85
Natural Language Processing Example:
E → E + T { E.value = E.value + T.value }
The right part of the CFG contains the semantic rules that specify how the
grammar should be interpreted. Here, the values of non-terminals E and T
are added together and the result is copied to the non-terminal E.
Semantic attributes may be assigned to their values from their domain at
the time of parsing and evaluated at the time of assignment or conditions.
Based on the way the attributes get their values, they can be broadly
divided into two categories : synthesized attributes and inherited attributes.
Synthesized attributes
These attributes get values from the attribute values of their child nodes.
To illustrate, assume the following production:
S → ABC
If S is taking values from its child nodes (A,B,C), then it is said to be a

synthesized attribute, as the values of ABC are synthesized to S.
As in our previous example (E → E + T), the parent node E gets its value
from its child node. Synthesized attributes never take values from their
parent nodes or any sibling nodes.
Inherited attributes
In contrast to synthesized attributes, inherited attributes can take values
from parent and/or siblings. As in the following production,
S → ABC
A can get values from S, B and C. B can take values from S, A, and C.
Likewise, C can take values from S, A, and B.
Expansion : When a non-terminal is expanded to terminals as per a
grammatical rule
86
86
Reduction : When a terminal is reduced to its corresponding non-terminal Semantic Analysis - I
according to grammar rules. Syntax trees are parsed top-down and left to
right. Whenever reduction occurs, we apply its corresponding semantic
rules (actions).
Semantic analysis uses Syntax Directed Translations to perform the above
tasks.
Semantic analyzer receives AST (Abstract Syntax Tree) from its previous
stage (syntax analysis).
Semantic analyzer attaches attribute information with AST, which are
called Attributed AST.
Attributes are two tuple value, <attribute name, attribute value>
For example:
int value = 5;
<type, “integer”>
<presentvalue, “5”>
For every production, we attach a semantic rule.

S-attributed SDT
If an SDT uses only synthesized attributes, it is called as S-attributed SDT.
These attributes are evaluated using S-attributed SDTs that have their
semantic actions written after the production (right hand side).
As depicted above, attributes in S-attributed SDTs are evaluated in

bottom-up parsing, as the values of the parent nodes depend upon the
values of the child nodes.
L-attributed SDT
This form of SDT uses both synthesized and inherited attributes with
restriction of not taking values from right siblings.
In L-attributed SDTs, a non-terminal can get values from its parent, child,
and sibling nodes. As in the following production
87
Natural Language Processing S → ABC
S can take values from A, B, and C (synthesized). A can take values from
S only. B can take values from S and A. C can get values from S, A, and
B. No non-terminal can get values from the sibling to its right.
Attributes in L-attributed SDTs are evaluated by depth-first and left-to-
right parsing manner.
We may conclude that if a definition is S-attributed, then it is also L-

attributed as L-attributed definition encloses S-attributed definitions.
6.1 INTRODUCTION
The purpose of semantic analysis is to draw exact meaning, or you can say
dictionary meaning from the text. The work of semantic analyzer is to
check the text for meaningfulness.
We already know that lexical analysis also deals with the meaning of the
words, then how is semantic analysis different from lexical analysis?
Lexical analysis is based on smaller token but on the other side semantic
analysis focuses on larger chunks. That is why semantic analysis can be
divided into the following two parts −
Studying meaning of individual word
It is the first part of the semantic analysis in which the study of the
meaning of individual words is performed. This part is called lexical
semantics.
Studying the combination of individual words
In the second part, the individual words will be combined to provide
meaning in sentences.
The most important task of semantic analysis is to get the proper meaning
of the sentence. For example, analyze the sentence “Ram is great.” In this
sentence, the speaker is talking either about Lord Ram or about a person
88
88
whose name is Ram. That is why the job, to get the proper meaning of the Semantic Analysis - I
sentence, of semantic analyzer is important.
Meaning Representation
Semantic analysis creates a representation of the meaning of a sentence.
But before getting into the concept and approaches related to meaning
representation, we need to understand the building blocks of semantic
system.
Building Blocks of Semantic System
In word representation or representation of the meaning of the words, the
following building blocks play an important role −
• Entities − It represents the individual such as a particular person,

location etc. For example, Haryana. India, Ram all are entities.
• Concepts − It represents the general category of the individuals such
as a person, city, etc.
• Relations − It represents the relationship between entities and
concept. For example, Ram is a person.
• Predicates − It represents the verb structures. For example,
semantic roles and case grammar are the examples of predicates.
Now, we can understand that meaning representation shows how to put
together the building blocks of semantic systems. In other words, it shows
how to put together entities, concepts, relation and predicates to describe a
situation. It also enables the reasoning about the semantic world.
Approaches to Meaning Representations
Semantic analysis uses the following approaches for the representation of
meaning −
• First order predicate logic (FOPL)

• Semantic Nets
• Frames
• Conceptual dependency (CD)
• Rule-based architecture
• Case Grammar
• Conceptual Graphs
Need of Meaning Representations
A question that arises here is why do we need meaning representation?
Followings are the reasons for the same −
Linking of linguistic elements to non-linguistic elements
The very first reason is that with the help of meaning representation the
linking of linguistic elements to the non-linguistic elements can be done.
89
Natural Language Processing Representing variety at lexical level
With the help of meaning representation, unambiguous, canonical forms
can be represented at the lexical level.
Can be used for reasoning
Meaning representation can be used to reason for verifying what is true in
the world as well as to infer the knowledge from the semantic
representation.
6.1.2 Definition of Semantic Analysis
structures, from the levels of phrases, clauses, sentences and paragraphs to
the level of the writing as a whole, to their language-
independent meanings. It also involves removing features specific to
particular linguistic and cultural contexts, to the extent that such a project
is possible. The elements of idiom and figurative speech, being cultural,
are often also converted into relatively invariant meanings in semantic
analysis. Semantics, although related to pragmatics, is distinct in that the
former deals with word or sentence choice in any given context, while
pragmatics considers the unique or particular meaning derived from
context or tone. To reiterate in different terms, semantics is about
universally coded meaning, and pragmatics, the meaning encoded in
words that is then interpreted by an audience.
Semantic analysis can begin with the relationship between individual
words. This requires an understanding of lexical hierarchy,
including hyponymy and hypernymy, meronomy, polysemy, synonyms,
antonyms, and homonyms. It also relates to concepts like connotation
(semiotics) and collocation, which is the particular combination of words
that can be or frequently are surrounding a single word. This can include
idioms, metaphor, and simile, like, "white as a ghost. In the current
chapter we turn to the topic of semantics. Informally, syntax concerns the
form of a valid program, while semantics concerns its meaning. Meaning
is important for at least two reasons: it allows us to enforce rules (e.g.,
type consistency) that go beyond mere form, and it provides the
information we need in order to generate an equivalent output program. It
is conventional to say that the syntax of a language is precisely that
portion of the language definition that can be described conveniently by a
context-free grammar, while the semantics is that portion of the definition
that cannot. This convention is useful in practice, though it does not
always agree with intuition. When we require, for example, that the
number of arguments contained in a call to a subroutine match the number
of formal parameters in the subroutine definition, it is tempting to say that
this requirement is a matter of syntax. After all, we can count arguments
without knowing what they mean. Unfortunately, we cannot count them
with context-free rules. Similarly, while it is possible to write a context-
free grammar in which every function must contain at least one return
statement, the required complexity makes this strategy very unattractive.
90
90 In general, any rule that requires the compiler to compare things that are
separated by long distances, or to count things that are not properly nested, Semantic Analysis - I
ends up being a matter of semantics. Semantic rules are further divided
into static and dynamic semantics, though again the line between the two
is somewhat fuzzy. The compiler enforces static semantic rules at compile
time. It generates code to enforce dynamic semantic rules at run time (or
to call library routines that do so). Certain errors, such as division by zero,
or attempting to index into an array with an out-of-bounds subscript,
cannot in general be caught at compile time, since they may occur only for
certain input values, or certain behaviors of arbitrarily complex code. In
special cases, a compiler may be able to tell that a certain error will always
or never occur, regardless of run-time input. In these cases, the compiler
can generate an error message at compile time, or refrain from generating
code to perform the check at run time, as appropriate.
Semantic Analysis programs: there will inevitably be cases in which an
error will always occur, but the compiler cannot tell, and must delay the
error message until run time; there will also be cases in which an error can
never occur, but the compiler cannot tell, and must incur the cost of
unnecessary run-time checks. Both semantic analysis and intermediate
code generation can be described in terms of annotation, or decoration of a
parse tree or syntax tree. The annotations themselves are known as
attributes. Numerous examples of static and dynamic semantic rules will
appear in subsequent chapters.
In this current chapter we focus primarily on the mechanisms a compiler
uses to enforce the static rules. We will consider intermediate code
generation (including the generation of code for dynamic semantic checks)
. we consider the role of the semantic analyzer in more detail, considering
both the rules it needs to enforce and its relationship to other phases of
compilation. Most of the rest of the chapter is then devoted to the subject
of attribute grammars. Attribute grammars provide a formal framework for
the decoration of a tree. This framework is a useful conceptual tool even in
compilers that do not build a parse tree or syntax tree as an explicit data
structure. We then consider various ways in which such grammars can be
applied in practice. discusses the issue of attribute flow, which constrains
the order(s) in which nodes of a tree can be decorated. In practice, most
compilers require decoration of the parse tree (or the evaluation of
attributes that would reside in a parse tree if there were one) to occur in
the process of an LL or LR parse. presents action routines as an ad hoc
mechanism for such “on-the-fly” evaluation. (mostly on the PLP CD) we
consider the management of space for parse tree attributes. Because they
have to reflect the structure of the CFG, parse trees tend to be very
complicated. Once parsing is complete, we typically want to replace the
parse tree with a syntax tree that reflects the input program in a more
straightforward way. One particularly common compiler organization uses
action routines during parsing solely for the purpose of constructing the
syntax tree. The syntax tree is then decorated during a separate traversal,
which can be formalized, if desired, with a separate attribute grammar.
91
Natural Language Processing Space Management for Attributes
Any attribute evaluation method requires space to hold the attributes of the
grammar symbols. If we are building an explicit parse tree, then the
obvious approach is to store attributes in the nodes of the tree themselves.
If we are not building a parse tree, then we need to find a way to keep
track of the attributes for the symbols we have seen (or predicted) but not
yet finished parsing. The details differ in bottom-up and top-down parsers.
For a bottom-up parser with an S-attributed grammar, the obvious
approach is to maintain an attribute stack that directly mirrors the parse
stack: next to every state number on the parse stack is an attribute record
for the symbol we shifted when we entered that state. Entries in the
attribute stack are pushed and popped automatically by the parser driver;
space management is not an issue for the writer of action routines.
Complications arise if we try to achieve the effect of inherited attributes,
but these can be accommodated within the basic attribute stack
framework. For a top-down parser with an L-attributed grammar, we have
two principal options. The first option is automatic, but more complex
than for bottom-up grammars. It still uses an attribute stack, but one that
does not mirror the parse stack. The second option has lower space
overhead, and saves time by “shortcutting” copy rules, but requires action
routines to allocate and deallocate space for attributes explicitly. In both
families of parsers, it is common for some of the contextual information
for action routines to be kept in global variables. The symbol table in
particular is usually global. We can be sure that the table will always
represent the current referencing environment, because we control the
order in which action routines (including those that modify the
environment at the beginnings and ends of scopes) are executed. In a pure
attribute grammar, we should need to pass symbol table information into
and out of productions through inherited and synthesized attributes.
6.2 LEXICAL SEMANTICS
One of the approaches or techniques of semantic analysis is the lexicon-

based approach. This technique calculates the sentiment orientations of the
whole document or set of sentence(s) from semantic orientation of
lexicons. Semantic orientation can be positive, negative, or neutral. The
dictionary of lexicons can be created manually as well as automatically
generated. The WorldNet dictionary is used by many researchers. First of
all, lexicons are found from the whole document and then WorldNet or
any other kind of online thesaurus can be used to discover the synonyms
and antonyms to expand that dictionary.
Lexicon-based techniques use adjectives and adverbs to discover the
semantic orientation of the text. For calculating any text orientation,
adjective and adverb combinations are extracted with their sentiment
orientation value. These can then be converted to a single score for the
whole value (Fig. below)
92
92
Semantic Analysis - I
The different approaches to lexicon-based approach are:

1. Dictionary-based approach
In this approach, a dictionary is created by taking a few words initially.
Then an online dictionary, thesaurus or WordNet can be used to expand
that dictionary by incorporating synonyms and antonyms of those words.
The dictionary is expanded till no new words can be added to that
dictionary. The dictionary can be refined by manual inspection.
2. Corpus-based approach
This finds sentiment orientation of context-specific words. The two
methods of this approach are:
Statistical approach: The words which show erratic behavior in positive
behavior are considered to have positive polarity. If they show
negative recurrence in negative text they have negative polarity. If the
frequency is equal in both positive and negative text then the word has
neutral polarity.
Semantic approach: This approach assigns sentiment values to words and
the words which are semantically closer to those words; this can be done
by finding synonyms and antonyms with respect to that word.
Lexical semantics (also known as lexicosemantic), as a subfield
of linguistic semantics, is the study of word meanings. It includes the
study of how words structure their meaning, how they act in grammar
and compositionality, and the relationships between the distinct senses and
uses of a word.
The units of analysis in lexical semantics are lexical units which include
not only words but also sub-words or sub-units such as affixes and
even compound words and phrases. Lexical units include the catalogue of
words in a language, the lexicon. Lexical semantics looks at how the
meaning of the lexical units correlates with the structure of the language
or syntax. This is referred to as syntax-semantics interface.
The study of lexical semantics looks at:
• the classification and decomposition of lexical items

• the differences and similarities in lexical semantic structure cross-
linguistically
• the relationship of lexical meaning to sentence meaning and syntax. 93
Natural Language Processing Lexical units, also referred to as syntactic atoms, can stand alone such as
in the case of root words or parts of compound words or they necessarily
attach to other units such as prefixes and suffixes do. The former are
called free morphemes and the latter bound morphemes. They fall into a
narrow range of meanings (semantic fields) and can combine with each
other to generate new denotations
Lexical Semantics
The first part of semantic analysis, studying the meaning of individual
words is called lexical semantics. It includes words, sub-words, affixes
(sub-units), compound words and phrases also. All the words, sub-words,
etc. are collectively called lexical items. In other words, we can say that
lexical semantics is the relationship between lexical items, meaning of
sentences and syntax of sentence.
Following are the steps involved in lexical semantics −
• Classification of lexical items like words, sub-words, affixes, etc. is
performed in lexical semantics.
• Decomposition of lexical items like words, sub-words, affixes, etc.
is performed in lexical semantics.
• Differences as well as similarities between various lexical semantic
structures is also analyzed.
We understand that words have different meanings based on the context of
its usage in the sentence. If we talk about human languages, then they are
ambiguous too because many words can be interpreted in multiple ways
depending upon the context of their occurrence.
Word sense disambiguation, in natural language processing (NLP), may
be defined as the ability to determine which meaning of word is activated
by the use of word in a particular context. Lexical ambiguity, syntactic or
semantic, is one of the very first problem that any NLP system faces. Part-
of-speech (POS) taggers with high level of accuracy can solve Word’s
syntactic ambiguity. On the other hand, the problem of resolving semantic
ambiguity is called WSD (word sense disambiguation). Resolving
semantic ambiguity is harder than resolving syntactic ambiguity.
For example, consider the two examples of the distinct sense that exist for
the word “bass” −
● I can hear bass sound.
● He likes to eat grilled bass.
The occurrence of the word bass clearly denotes the distinct meaning. In
first sentence, it means frequency and in second, it means fish. Hence, if
it would be disambiguated by WSD then the correct meaning to the above
sentences can be assigned as follows −
● I can hear bass/frequency sound.
• He likes to eat grilled bass/fish.

94
94
7
SEMANTIC ANALYSIS - II
Unit structure
7.0 Attachment for fragment of English-sentences
7.1 Noun phrases
7.2 Verb phrases
7.3 Prepositional phrases
7.4 Relations among lexemes & their senses
7.4.1 Homonymy, Polysemy, Synonymy
7.5 Robust, Word Senses
7.5.1 Disambiguation (WSD)
7.0 ATTACHMENT FOR FRAGMENT OF ENGLISH-

SENTENCES
A sentence fragment is a group of words that cannot stand alone as a

sentence because it does not express a complete thought. Learn about
subordinators, fragment phrases, and fragment temptations to understand
how to identify and correct sentence fragments.
Definition of a Sentence Fragment
Sentence fragments are groups of words that look like sentences, but
aren't. To be a sentence, groups of words need to have at least one
independent clause. An independent clause is any group of words that
contain both a subject and a verb and can stand on its own. For example, 'I
like cheeseburgers' is an independent clause.
Sentence fragments never have independent clauses, but instead are
dependent clauses or phrases. Fragments can masquerade as real sentences
because they begin with a capital letter and end with a period. If you read
them more closely, you'll see that fragments don't form a complete
thought. A sentence fragment is a little like having only half of the pieces
to a puzzle. Without all the pieces, you won't have the whole picture.
Where to Find Sentence Fragments
Sentence fragments usually appear before or after the independent clauses
to which they belong. For example:
When we got in the car. We rolled down the windows.
95
Natural Language Processing 'When we got in the car' is a sentence fragment and a dependent clause. It
clearly belongs to the independent clause that follows it and should be
rewritten like this:
When we got in the car, we rolled down the windows.
Or like this:
We rolled down the windows when we got in the car.
Subordinators
The sentence fragment 'When we got in the car' also has the subordinator
'when'. Some other examples of subordinators are: 'after', 'although',
'before', 'if', 'since', 'until', 'when', 'where', 'while', and 'why'. Clauses with
subordinators can be called either dependent clauses or subordinating
clauses, but when those clauses appear at the beginning of a sentence, they
should be followed by a comma.
Fragment Phrases
Phrases are groups of words that are missing a subject or verb, or both.
Phrases can also masquerade as sentences, like dependent clauses can.
Here are some examples.
Here's an example missing subject and verb:
From morning until night.
This fragment can be made a complete sentence by changing it to:
I worked from morning until night.
Adding 'I' as the subject and 'worked' as the verb corrects this fragment
and makes it an independent clause and a complete thought.
Here's an example of a missing subject:
Start after the weekend.
This fragment can be made a complete sentence by changing it to:
Classes start after the weekend.
Adding the subject 'classes' corrects this fragment and makes it an
independent clause and a complete thought.
Finally, here's an example of a missing verb:
Some girls in the class.
This fragment can be changed to:
Some girls in the class study together.
Adding the verb 'study' corrects this fragment and makes it an independent
clause and a complete thought.
Fragment Temptations
Certain words and expressions make it easy to fall into the sentence
fragment habit. Some of these words include 'also', 'for example', 'and',
'but', 'for instance', 'mainly', 'or', and 'that'. Here's how they appear in a
sentence:
96
96
Harris claims that men and women have different ideas about dating. For Semantic Analysis - II
example, that men should pay for dinner on a date.
7.1 NOUN PHRASES
Noun phrases are groups of words that function like nouns. Typically, they
act as subjects, objects or prepositional objects in a sentence. While that
might seem tricky to grasp, the best way to understand these useful
phrases is to see them in action. Get a clear idea of a noun phrase and how
it is in sentence through example.
What is a noun phrase example?
A noun phrase is either a pronoun or any group of words that can be
replaced by a pronoun. For example, 'they', 'cars', and 'the cars' are
noun phrases, but 'car' is just a noun, as you can see in these sentences (in
which the noun phrases are all in bold) Q: Do you like cars? A: Yes, I like
them.
Noun phrases are simply nouns with modifiers. Just as nouns can act as
subjects, objects and prepositional objects, so can noun phrases. Similarly,
noun phrases can also work in a sentence as adjectives, participles,
infinitives, and prepositional or absolute phrases. Noun phrases are
important for adding more detail to a noun. Examples of simple noun
phrases include:
• the little boy

• the happy puppy
• the building on the corner
• the sharp pencil
• your religion
7.2 VERB PHRASES
Verb phrase (VP): These phrases are lexical units that have a verb
acting as the head word. Usually, there are two forms of verb phrases.
One form has the verb components as well as other entities such as nouns,
adjectives, or adverbs as parts of the object.
The phrase would include the verbal (participle, gerund or infinitive) and
any modifiers, complements or objects. Examples of verb phrases versus
verbal phrases include: The man was texting on his phone. (verb phrase
was texting functions as the action) Texting on his phone, the man
swerved into a ditch.
What is the main verb in a verb phrase?
The main verb is also called the lexical verb or the principal verb. This
term refers to the important verb in the sentence, the one that typically
shows the action or state of being of the subject. Main verbs can stand
alone, or they can be used with a helping verb, also called an auxiliary
verb.
97
Natural Language Processing What is a verb phrase in English language?
In English a verb phrase combines with a noun or noun phrase acting
as subject to form a simple sentence. a phrase consisting of a main verb
and any auxiliaries but not including modifiers, objects, or complements.
What are the types of verb phrases?
Verb phrases generally are divided among two types: finite, of which the
head of the phrase is a finite verb; and nonfinite, where the head is a
nonfinite verb, such as an infinitive, participle or gerund.
How do you find a verb phrase?
A verb phrase consists of a verb plus another word that further
illustrates the verb tense, action, and tone. The other word or words tied
to a verb in a verb phrase are its dependents, which can be adverbs,
prepositional phrases, helping verbs, or other modifiers
Verb Phrase Examples
● She was walking quickly to the mall.
● He should wait before going swimming.
● Those girls are trying very hard.
● Ted might eat the cake.
● You must go right now.
● You can't eat that!
● My mother is fixing us some dinner.
● Words were spoken.
7.3 PREPOSITIONAL PHRASES
Prepositional phrases are groups of words containing prepositions.

Remember that prepositions are words that indicate the relationships
between various elements within a sentence, and you’ll never have
difficulty identifying prepositional phrases.
A prepositional phrase is a group of words that lacks either a verb or a
subject, and that functions as a unified part of speech. It normally consists
of a preposition and a noun or a preposition and a pronoun.
Remember the following rules for prepositional phrases and you will find
that using them becomes much easier.
● Prepositional phrases always consist of two basic parts at minimum:
the object and the preposition.
● In formal English, prepositions are almost always followed by
objects.
● Adjectives can be placed between the prepositions and objects in
prepositional phrases.
98
98
● Prepositional phrases can act as adverbs or adjectives. When they Semantic Analysis - II
are used as adjectives, they modify nouns and pronouns in the same
way single-word adjectives do.
● When prepositional phrases are used as adverbs, they at the same
way single-word adverbs and adverb clauses do, modifying
adjectives, verbs, and other adverbs.
Examples of Prepositional Phrases
The following sentences contain examples of prepositional phrases; the
prepositional phrase in each sentence is italicized for easy identification.
The cupcake with sprinkles is yours.
The cupcake with colorful sprinkles is yours.
We climbed up the hill.
We climbed up the very steep hill.
The rabbits hopped through the garden.
The rabbits hopped through the perfectly manicured garden.
7.4. RELATIONS AMONG LEXEMES & THEIR SENSES
What is the relation among lexemes and their senses?

A lexeme is an individual entry in the lexicon. A lexicon is meaning
structure holding meaning relations of lexemes. A lexeme may have
different meanings. A lexeme's meaning component is known as one of
its senses.
What are the sense relation?
sense relations as “the relations of meaning between words, as
expressed in synonymy, hyponymy, and antonymy.” Thus, sense
relations can be seen from the similarity of meaning as in synonymy, the
inclusion of meaning as in hyponymy, and the oppositeness of meaning as
in antonymy.
What are the types of sense relations?
Two major types of sense relations can be distinguished:
• Sense relations of inclusion, esp. hyponymy and synonymy.

• Sense relations of exclusion, esp. complementarity and antonymy
(both of which are instances of the relationship of incompatibility).
7.4.1 -Homonymy, Polysemy, Synonymy
The word Homonymy (from the Greek—homos: same, onoma: name)
is the relation between words with identical forms but different
meanings—that is, the condition of being homonyms. A stock example is
the word bank as it appears in "river bank" and "savings bank."
99
Natural Language Processing What is the meaning of Homonymy?
Homonymy is the relationship between words that are homonyms—
words that have different meanings but are pronounced the same or
spelled the same or both. It can also refer to the state of being
homonyms.
10 Homonyms with Meanings and Sentences
• Cache – Cash:
• Scents – Sense:
• Chile – Chili:
• Choir – Quire:
• Site – Sight:
• Facts- Fax:
• Finnish – Finish:
What is polysemy and examples?
polysemy Add to list Share. When a symbol, word, or phrase means
many different things, that's called polysemy. The verb "get" is a good
example of polysemy — it can mean "procure," "become," or
"understand."
What are the polysemy words?
A polysemous word is a word that has different meanings that derive
from a common origin; a homograph is a word that has different
meanings with unrelated origins. Polysemous words and homographs
constitute a known problem for language learners.
How many types of polysemy are there?
Types of polysemy
Linear polysemy accounts for a specialization-generalization relation
between senses and, in turn, is divided into four types:
autohyponymy, automeronymy, autosuperordination and autoholonymy.
Synonymy is a relation between individual senses of words, so that a
single word typically has different sets of synonyms for each of its
senses. For example, coat has different synonyms for its senses 'outer
garment' (e.g., jacket) and 'covering layer'(e.g., layer)
What is the concept of synonymy? Meaning of synonymy in English?
the state of being synonymous (= havings the same or almost the same
meaning as another word or phrase in the same language) or the fact
that words or phrases are synonymous: ... Children learn to accept two
labels for the same object (synonymy).
100
100
What are the 5 examples of synonyms? Semantic Analysis - II
II. Examples of Synonyms
• Bad: awful, terrible, horrible.

• Good: fine, excellent, great.
• Hot: burning, fiery, boiling.
• Cold: chilly, freezing, frosty.
• Easy: Simple, effortless, straightforward.
• Hard: difficult, challenging, tough.
• Big: large, huge, giant.
• Small: tiny, little, mini.
7.5 ROBUST, WORD SENSES
What is the meaning of being robust?

Definition of robust
1a : having or exhibiting strength or vigorous health. b : having or
showing vigor, strength, or firmness a robust debate a robust faith.
What is robust example?
The definition of robust is a strong and healthy person or animal,
something rich and full of flavor or an activity that requires someone
strong and healthy to complete. An example of robust is a winning race
horse. An example of robust is strong coffee.
What is word sense in NLP?
In natural language processing, word sense disambiguation (WSD) is the
problem of determining which "sense" (meaning) of a word is
activated by the use of the word in a particular context, a process which
appears to be largely unconscious in people.
What type of word is sense?
noun. any of the faculties, as sight, hearing, smell, taste, or touch, by
which humans and animals perceive stimuli originating from outside or
inside the body: My sense of smell tells me that dinner is ready.
Summary and Concluding Remarks
This chapter has discussed the task of semantic analysis. We reviewed the
sorts of language rules that can be classified as syntax, static semantics,
and dynamic semantics, and discussed the issue of whether to generate
code to perform dynamic semantic checks. We also considered the role
that the semantic analyzer plays in a typical compiler. We noted that both
the enforcement of static semantic rules and the generation of intermediate
code can be cast in terms of annotation, or decoration, of a parse tree or
syntax tree. 101
Natural Language Processing We then presented attribute grammars as a formal framework for this
decoration process. An attribute grammar associates attributes with each
symbol in a context-free grammar or tree grammar, and attribute rules
with each production.
Synthesized attributes are calculated only in productions in which their
symbol appears on the left-hand side. The synthesized attributes of tokens
are initialized by the scanner. Inherited attributes are calculated in
productions in which their symbol appears within the right-hand side; they
allow calculations internal to a symbol to depend on the context in which
the symbol appears. Inherited attributes of the start symbol (goal) can
represent the external environment of the compiler. Strictly speaking,
attribute grammars allow only copy rules (assignments of one attribute to
another) and simple calls to semantic functions, but we usually relax this
restriction to allow more-or-less arbitrary code fragments in some existing
programming language. Just as context-free grammars can be categorized
according to the parsing algorithm(s) that can use them, attribute
grammars can be categorized according to the complexity of their pattern
of attribute flow. S-attributed grammars, in which all attributes are
synthesized, can naturally be evaluated in a single bottom up pass over a
parse tree, in precisely the order the tree is discovered by an LR family
parser. L-attributed grammars, in which all attribute flow is depth-first
left-to-right, can be evaluated in precisely the order that the parse tree is
predicted and matched by an LL-family parser.
Attribute grammars with more complex patterns of attribute flow are
not commonly used in production compilers, but for syntax-based editors,
incremental compilers, and various other tools. While it is possible to
construct automatic tools to analyze attribute flow and decorate parse
trees, most compilers rely on action routines, which the compiler writer
embeds in the right-hand sides of productions to evaluate attribute rules at
specific points in a parse. In an LL-family parser, action routines can be
embedded at arbitrary points in a production’s right-hand side. In an LR-
family parser, action routines must follow the production’s left corner.
Space for attributes in a bottom-up compiler is naturally allocated in
parallel with the parse stack, but this complicates the management of
inherited attributes. Space for attributes in a top-down compiler can be
allocated automatically, or managed explicitly by the writer of action
routines. The automatic approach has the advantage of regularity, and is
easier to maintain; the ad hoc approach is slightly faster and more flexible.
In a one-pass compiler, which interleaves scanning, parsing, semantic
analysis, and code generation in a single traversal of its input, semantic
functions or action routines are responsible for all of semantic analysis and
code generation. More commonly, action routines simply build a syntax
tree, which is then decorated during separate traversal(s) in subsequent
pass(es). we will consider a wide variety of programming language
constructs. Rather than present the actual attribute grammars required to
implement these constructs, we will describe their semantics informally,
and give examples of the target code. We will return to attribute grammars
in Chapter 14, when we consider the generation of intermediate code in
102
102 more detail.
7.5.1 Disambiguation (WSD) Semantic Analysis - II
What do you mean by word sense disambiguation WSD?

In natural language processing, word sense disambiguation (WSD) is the
problem of determining which "sense" (meaning) of a word is activated
by the use of the word in a particular context, a process which appears to
be largely unconscious in people
What is word sense disambiguation write down the application of
WSD?
Word Sense Disambiguation (WSD), has been a trending area of research
in Natural Language Processing and Machine Learning. WSD is basically
solution to the ambiguity which arises due to different meaning of
words in different context. ... The ambiguity that arises due to this, is
tough for a machine to detect and resolve
What is disambiguation example?
The definition of a disambiguation is a removal of uncertainty or
confusion. An example of disambiguation is when a study explains the
discrepancy between two different scientific studies that point to different
results that create uncertainty. noun. 2
What are the different approaches to WSD?
WSD approaches are categorized mainly into three types,
Knowledge-based,
Supervised and
Unsupervised methods,
Word Sense Disambiguation is an important method of NLP by which the
meaning of a word is determined, which is used in a particular context.
NLP systems often face the challenge of properly identifying words, and
determining the specific usage of a word in a particular sentence has many
applications.
Word Sense Disambiguation basically solves the ambiguity that arises in
determining the meaning of the same word used in different situations.
Word Sense Disambiguation Applications
WSD has many applications in various text processing and NLP fields.
• WSD can be used alongside Lexicography. Much of the

modern Lexicography is corpus-based. WSD, used in Lexicography
can provide significant textual indicators.
• WSD can also be used in Text Mining and Information Extraction
tasks. As the major purpose of WSD is to accurately understand the
meaning of a word in particular usage or sentence, it can be used for
the correct labeling of words.
103
Natural Language Processing • For example, from a security point of view, a text system should be
able to understand the difference between a coal “mine” and a land
“mine”.
• While the former serves industrial purposes, the latter is a security
threat. So a text mining application must be able to determine the
difference between the two.
• Similarly, WSD can be used for Information Retrieval purposes.
Information Retrieval systems work through text data primarily
based on textual information. Knowing the relevance of using a
word in any sentence will surely help.
• Challenges in Word Sense Disambiguation
• WSD faces a lot of challenges and problems.
• The most common problem is the difference between various
dictionaries or text corpus. Different dictionaries have different
meanings for words, which makes the sense of the words to be
perceived as different. A lot of text information is out there and
often it is not possible to process everything properly.
• Different applications need different algorithms and that is often a
challenge for WSD.
• A problem also arises is that words cannot be divided into discrete
meanings. Words often have related meanings and this causes a lot
of problems.
How to implement WSD?
There are four main ways to implement WSD.
These are:
Dictionary- and knowledge-based methods:
These methods rely on text data like dictionaries, thesaurus, etc. It is based
on the fact that words that are related to each other can be found in the
definitions. The popularly used Leask method, which we shall discuss
more later is a seminal dictionary-based method.
Supervised methods:
In this type, sense-annotated corpora are used to train machine learning
models. But, a problem that may arise is that such corpora are very tough
and time-consuming to create.
Semi-supervised Methods:
Due to the lack of such corpus, most word sense disambiguation
algorithms use semi-supervised methods. The process starts with a small
amount of data, which is often manually created.
This is used to train an initial classifier. This classifier is used on an
untagged part of the corpus, to create a larger training set. Basically, this
104
104
method involves bootstrapping from the initial data, which is referred to as Semantic Analysis - II
the seed data.
Semi-supervised methods thus, use both labeled and unlabeled data.
Unsupervised Methods:
Unsupervised Methods pose the greatest challenge to researchers and NLP
professionals. A key assumption of these models is that similar meanings
and senses occur in a similar context. They are not dependent on manual
efforts, hence can overcome the knowledge acquisition deadlock.
Lesk Algorithm
Lesk Algorithm is a classical Word Sense Disambiguation algorithm
introduced by Michael E. Lesk in 1986.
The Lesk algorithm is based on the idea that words in a given region of the
text will have a similar meaning. In the Simplified Lesk Algorithm, the
correct meaning of each word context is found by getting the sense which
overlaps the most among the given context and its dictionary meaning.
Read More about the Lesk Algorithm here.
Let us start by importing the libraries.
So, the NLTK library can implement the Lesk method properly.
Do check out the code here.
Why is WSD any relevant?
Word Sense Disambiguation is closely related to Parts of speech tagging
and is an important part of the whole Natural Language Processing
process.
WSD if implemented properly, can lead to breakthroughs in NLP. A
problem that often arises is the whole meaning of word sense. Word sense
is not a numeric quantity that can be measured or a true or false value that
can be denoted as 1 or 0.
The whole idea of word sense is controversial. The meaning of a word is
highly contextual and depends on its usage. It is not something that can be
easily measured as a discrete quantity.
Lexicography deals with generalizing the corpus and explaining the full
and extended meaning of a word. But, sometimes these meanings might
not apply to the algorithms or data.
My personal opinion and experience, say that working with text data can
be tricky. So, implementation of WSD can be often very difficult and
problem-creating.
But, WSD has immense applications and uses.
If a computer algorithm can just read a text and identify different uses of a
text, it would mean vast improvements in the field of text analytics. 105
Natural Language Processing Comparing and evaluating various WSD methods can be difficult. But,
with time more research is going on regarding WSD and it can be
improved.
Dictionary Based Algorithm. A simple approach to segment text is to scan
each character one at a time from left to right and look up those
characters in a dictionary. If the series of characters found in the
dictionary, then we have a matched word and segment that sequence as a
word.
Functional Modelling and Mathematical Models: A Semantic Analysis
How the model M is connected to the system S.
Feynman's weight on a spring is a convenient place to start. In all three
examples below, S is a weight on a spring, either a real one or one that we
propose to construct. But the connection takes a different form in each
example.
(A) The model is a computer mockup in virtual reality. It looks like the
real thing, and maybe sounds like the real thing if the programmer was
having fun. In this case the connection with the system
is resemblance or similarity; we say that the model is pictorial.
In functional modelling the modeler will sometimes turn an early stage of
the specification into a toy working system, called a prototype. The
prototype is a pictorial model of the final system. It shows how the final
system will operate, by working more or less like the final system but
maybe with some features missing.
(B) The model is a graph, for example
This graph describes the system less directly. Left to right in the graph
represents time, up and down represents the vertical distance of the center
of mass of the weight from its resting position. In both dimensions a
distance in the graph is proportional to a distance in space or time. A
model that can be read in this way, by taking some dimensions in the
model as corresponding to some dimensions in the system, is called
an analogue model.
106
106
In hydraulic and aeronautical engineering one often meets scale models. Semantic Analysis - II
These are analogue models where the dimensions of the final system are
accurately scaled up or down (usually down) so that the model is a more
convenient size than the final system. But if all the dimensions are scaled
down in a ratio r, then the areas are scaled down in ratio r2 and the
volumes (and hence the weights) in ratio r3. So given the laws of physics,
how should we scale the time if we want the behavior of the model to
predict the behavior of the system?
Dimensional analysis answers this question.
A model can be both pictorial and analogue. For example, the architect's
model is both. But the model is clearly not a pictorial model; it doesn't
look anything like a weight on a spring.
(C) Feynman himself models his system with an equation, (6). His
equation is a piece of text which makes a statement about the system. We
call it a textual model.
Some fields have developed specialist notations for their subject matter.
Generally, these notations are textual, in the sense that they build up
expressions from a finite alphabet, though there may be pictorial reasons
why one symbol was chosen rather than another. Often, they are meant to
be written and read rather than spoken. Musical scores are an obvious
example. There is no inherent problem about translating the score of a
Beethoven piano sonata symbol by symbol into English (‘The time
signature is common time, the first bar begins with a minim at g in the
right hand … ‘) in such a way that the original score could be recovered
from the English-though I can't think why anybody would want to do it.
The analogue model doesn't translate into English in any similar way.
Another example of a textual notation is Universal Modelling Language
(UML), which is often used in early stages of software modelling; it's less
specialist than musical scores but still very limited in what it can express.
The diagram notation of Barwise and Etchemin’s logic software Hyper
proof [Barwise and Etchemin, 1994] looks pictorial but can be read as
textual. The upper part of the screen carries pictures of a chessboard with
various objects arranged on it, while the lower part carries conventional
logical formulas. Built-in rules (Observe and Apply) carry information
from pictures to formulas and from formulas to pictures. Although we can
say in formulas anything we can say in pictures, and to some extent vice
versa, the two kinds of representation encourage different habits of
reasoning.
Keith Stenning [Stenning, 2002] reports some valuable experiments on
how students with different cognitive styles use Hyperproof. His surveys
different forms of representation from a cognitive point of view.
Textual models are particularly important for us because in principle
Tarski's semantic analysis applies to them. Tarski himself said
107
Natural Language Processing Whoever wishes … to pursue the semantics of colloquial language with
the help of exact methods will be driven first to undertake the thankless
task of a reform of this language…. It may, however, be doubted whether
the language of everyday life, after being ‘rationalized’, in this way, would
still preserve its naturalness and whether it would not rather take on the
characteristic features of the formalized languages.
Tarski may have intended these remarks to discourage people from
extending his semantic theory beyond the case of formalized languages.
But today his theory is applied very generally, and the ‘rationalization’,
that he refers to is taken as part of the job of a semanticist. For example
the diagrams of Barwise and Etchemendy (above) are studied in this spirit.
Very many models are text from the outset, or can be read as text. One
important case that we need to consider is computer models. For example
models for wind turbines are usually presented as computer programs
together with some accompanying theory to justify the programs. For
semantic analysis we need to be more precise about exactly what feature
of a computer model is the actual model. Let me give my own answer;
other analysts may see things differently.
The information about the proposed wind turbine is got by running the
program. So we should count the model as being the output of the
program. The output may include text printed on the screen or saved in a
file; in this respect the model is textual. The output may also consist of
pictures on the screen, or graphs; in this respect the model is pictorial, and
possibly also analogue. Dynamic real-time simulations are certainly
analogue; they may include sound as well as graphics.
Often the same program can be run with different parameters. (For
example in wind turbine modelling one uses programs that simulate wind
conditions and are seeded with a random number.
For practical purposes these programs can be thought of as allowing

infinitely many parameter values.) Only a few of these runs will ever take
place; the rest are virtual outputs. So we have to allow that a textual model
can consist of virtual text-or perhaps better, it can consist of a family of
different virtual texts.
On this reading, the computer program itself is not the model; it is a

device for generating the model. Also the background theory supporting
the program is not the model-though often it contains explanatory models
of other things.
Feynman's example throws up a further object that we need to notice. This

is the function f which, for any real numbers A, γ, t, ωγ, takes the value
(13)f(A,γ,t.ωγ)=Ae-γt/2cosωγt.
108
108
This function is an abstract object. For definiteness some people give it a Semantic Analysis - II
set-theoretic form by identifying it with a set of ordered 5-tuples of real
numbers. Although the function clearly bears some close relationship to
the equation , it's a wholly different kind of object. We can't put it on a
page or a screen, or make it out of wood or plaster of paris. In short it is
not ‘accessible’, to us in any direct way.
We can only have any cognitive relationship to it through some

description of it-for example the equation (6). For this reason I think we
should hesitate to call the function a ‘model’, of the spring-weight system.
(Later we will see that it's closer to a semantic model, though it isn't quite
that either.) Nor should we confuse functions in this sense with the
‘function’, of an artefact as in functional modelling (on which see the
chapter by Vermaas and Garbacz in this Volume).
Reference book pdf link
https://www.researchgate.net/publication/289220287_Semantic_Analysis
A Coursebook SECOND EDITION JAMES R. HURFORD Professor of
General Linguistics, University of Edinburgh BRENDAN HEASLEY
Consultant (Postgraduate Training), Sharjah Women’s College, United
Arab Emirates MICHAEL B. SMITH Associate Professor of Linguistics,
Oakland University

109
8
TEXT SUMMARIZATION, TEXT
CLASSIFICATION
Unit Structure
8.0 Objectives
8.1 Introduction
8.2 Overview
8.3 Text summarization- LEXRANK
8.4 Optimization based approaches for summarization
8.5 Summarization evaluation
8.6 Text classification
8.7 Let us Sum Up
8.9 Bibliography
8.10 Unit End Exercises
8.0 OBJECTIVES
An essential NLP task is text summarization. Extraction and abstraction

are the two basic kinds of approaches to text summarization. To create a
summary, extractive methods choose a subset of existing words, phrases,
or sentences from the original text. Abstractive methods, on the other
hand, construct an internal semantic representation first, then generate a
summary using natural language generation techniques. Such a summary
may include words that aren't explicitly mentioned in the original paper.
The majority of text summarizing systems uses extractive summarization.
The main issues in text summarization are topic identification,
interpretation, summary generation, and evaluation of the created
summary. Identifying significant phrases in the document and applying
them to pick sentences for inclusion in the summary are critical jobs in
extraction-based summarization. Abstraction-based approaches, on the
other hand, paraphrase chunks of the source document.
8.1 INTRODUCTION
The text summary is a technique for condensing vast passages of text into
manageable chunks. The goal is to develop a logical and fluent summary
that only includes the document's major ideas. In machine learning and
natural language processing, automatic text summarization is a prevalent
challenge (NLP).
110
110
The technique has proven to be crucial in quickly and accurately Text Summarization, Text
Classification
summarizing large texts, which would be costly and time-consuming if
done manually.
Before producing the requisite summary texts, machine learning models
are frequently trained to comprehend documents and condense the useful
information.
The purpose of text summarization Data is to this century what oil was to
the previous one, propelled by modern technical breakthroughs. The
collection and transmission of massive volumes of data have parachuted
into our world today.
With so much data moving in the digital world, machine learning
algorithms that can automatically condense lengthy texts and offer
accurate summaries that carry the intended contents fluently are needed.
Furthermore, using text summarization reduces reading time, speeds up
the research process, and expands the quantity of information that may fit
in a given space.
8.2 OVERVIEW
Text summarization is an important activity in Natural Language

Processing (NLP) that will undoubtedly have a significant impact on our
lives. With the rise of the digital world and its impact on the publishing
industry, devoting time to thoughtfully read an article, document, or book
to determine its relevance is no longer an option, especially given time
constraints.
Furthermore, with the growing number of articles being published and the
digitization of traditional print magazines, keeping track of the growing
number of publications available on the web has become practically
impossible. This is where text summary can aid in the reduction of lengthy
texts.
THE DEFINITION OF TEXT SUMMARIZATION
Automatic Text Summarization is a technique in which computer software
shortens lengthy texts and provides summaries to convey the desired
content. It is a prevalent challenge in machine learning and natural
language processing (NLP).
The process of constructing a concise, cohesive, and fluent summary of a
lengthier text document, which includes highlighting the text's important
points, is known as text summarization.
• Extractive Summarization
• Abstractive Summarization
111
Extractive approaches aim to summarize articles by finding key sentences

or phrases from the original text and piecing together sections of it to
create a shortened version. The summary is then created using the
retrieved sentences.
Summarization of the Abstract unlike extraction, this method depends on
advanced natural language algorithms to paraphrase and reduce portions
of a document. Abstractive machine learning algorithms can construct new
phrases and sentences to capture the meaning of the source content. When
done effectively in deep learning issues, such abstraction can help
overcome grammatical mistakes.
8.3 TEXT SUMMARIZATION- LEXRANK
LexRank is a graph-based unsupervised technique for text summarization.

Sentence grading is done using the graph approach. LexRank is a tool for
calculating sentence relevance in a graph representation of sentences
based on the concept of eigenvector centrality.
In this paradigm, the adjacency matrix of the graph representation of
sentences is a connectivity matrix based on intra-sentence cosine
similarity. This sentence extraction is primarily concerned with a group of
sentences with the same goal in mind, i.e. a centroid sentence is chosen as
the mean for all other sentences in the document. The sentences are then
graded based on their similarity.
Any multi-set documents can be used as the target document.
Based on Eigen Vector Centrality, a graphical approach was developed.
The vertices of the graphs are where sentences are inserted. The cosine
similarity measure is used to calculate the weight of the Edges.
112
112
Text Summarization, Text
Classification
Where Si denotes the sentences at the vertices, and Wij denotes the
weights on the edges.
Computation of cosine similarity
The bag of words model is used to describe N-dimensional vectors to
define similarity, where N is the number of all possible worlds in a given
language. The value of the appropriate dimension in the vector
representation of the sentence for each word that appears in the sentence is
the number of occurrences of the word in the sentence times the idf of the
word.
Where tfw,s is the number of occurrences of the word w in the sentence s.

Eigen Vector Centrality and LexRank
To establish the total centrality, each node's contribution is calculated.
However, all Nodes in a graphical network do not have to be deemed
equally essential. If there are unrelated texts with major important
sentences, the centrality of these sentences will be automatically
increased. The origin nodes are taken into account to avoid this issue. As a
result, here's a quick way to figure out centrality:
113
Natural Language Processing Matrix B is the adjacency matrix formed by dividing each element by the
associated row sum in the similarity graph. The left eigenvector of matrix
B has an eigenvalue of 1 and is denoted by pT.
ALGORITHM
Algortihm to calculate LexRank Scores is as follows:
The required arrays for sentences, threshold, LexRank Scores, Cosine

matrix, Degree, and Eigen Values are all initialized in the first four phases.
the-idf modified values for the given sentences are used to initialize our
matrix in steps 5 to 9. The element is replaced with value 1 if the values
are larger than a threshold idf value. If the value is less than the specified
threshold, it is replaced with 0. In these stages, we establish a conventional
tf-idf table or matrix.
Each value of the cosine matrix is divided by the degree of each node in
steps 17 to 21. The matching degree of each node is called degree
centrality in this case.
Finally, we calculate the final score using the Power iteration approach.
114
114
8.4 OPTIMIZATION BASED APPROACHES FOR Text Summarization, Text
Classification
SUMMARIZATION
With the growth of the World Wide Web and electronic government
services over the last decade, automatic text summarization has gotten a
lot of attention (e.g., electronic document management systems). Users
can access a vast amount of information thanks to the continued growth of
the WWW and online text collections from e-government services.
Material overload either wastes a large amount of time viewing all of the
information or causes useful information to be overlooked. As a result,
new technologies that can successfully process documents are in high
demand. In technological surroundings, automatic text summarization is a
critical technology for overcoming this barrier. Automatic text
summarization technology is improving, and it may give a solution to the
information overload problem
The practice of automatically constructing a compressed version of a
given text that delivers helpful information to readers is known as
document summarising. Summarization has been defined as a reductive
transformation of a given text, usually stated as a three-step process:
selection of salient bits of text, aggregation and abstraction of information
for various selected chunks, and lastly display of the final summary text.
This method can be applied to a variety of tasks, including information
retrieval, intelligence gathering, data extraction, text mining, and indexing.
Extractive and abstractive approaches to document summarization can be
broadly distinguished. In general, an abstract is a summary of
concepts/ideas taken from a source that is then "reinterpreted" and given
differently, whereas an extract is a summary of units of text taken from a
source and presented verbatim. In reality, the vast majority of studies have
focused on summary extraction, which takes bits from the source such as
keywords, sentences, or even paragraphs to create a summary. Abstractive
approaches, in contrast to extractive approaches, are more difficult to
adopt since they need more work.
To read source texts and generate new texts, you'll need a lot of domain
knowledge. "Reading and understanding the text to recognize its content,
which is subsequently compiled in a succinct text" is how abstraction is
defined.
The summary might be a single document or a multi-document, depending
on the number of papers to be summarised. Only one document can be
condensed into a shorter form using single-document summarising. Multi-
document summarising reduces a large number of documents into a single
summary. It gives users a domain overview of a topic, indicating what is
similar and different across multiple papers, as well as relationships
between pieces of information in different documents, and lets them zoom
in for more information on specific elements of interest. Multi-document
summary seeks to extract relevant information about a subject or topic
from various texts written about that subject or topic. It accomplishes
115
Natural Language Processing knowledge synthesis and knowledge discovery by combining and
integrating information from several publications, and it can also be used
for knowledge acquisition.
There are two types of extraction-based summarization methods:
supervised and unsupervised. Supervised approaches are based on
algorithms that use a large number of human-made summaries, and are
thus best suited to texts that are relevant to the summarizer model. As a
result, they don't always give a sufficient summary for papers that aren't
exactly like the model. In addition, when users modify the aim of
summarising or the properties of documents, the training data must be
rebuilt or the model must be retrained. To train the summarizer,
unsupervised methods do not require training data such as human-made
summaries. To score and rank sentences using unsupervised approaches,
early researchers frequently choose various statistic and linguistic criteria.
On the internet, you can now obtain a wealth of information on people's
perspectives on practically any topic. In complicated, high-stakes
decision-making processes, the capacity to interpret such data is crucial. A
person interested in purchasing a laptop can browse consumer reviews
from others who have purchased and utilized the device. Customer
feedback on a freshly introduced product at the corporate level might help
discover flaws and features that need to be improved.
To express people's opinions to users, effective summarising techniques
are required. The creation of a content selection strategy that defines what
vital information should be given is a difficult problem when adopting this
approach in a specific domain. In general, content selection is a crucial
task at the heart of both summarization and NLG, and it's an area where
cross-fertilization could be fruitful.
Existing NLG systems typically handle content selection by creating a
heuristic based on a number of relevant parameters and optimising it.
ILEX (Intelligent Labelling Explorer) is a database-based system for
creating labels for groups of objects, such as museum artefacts (O'Donnell
et al., 2001). Its content selection approach comprises assigning
knowledge elements a heuristic relevance score and returning the things
with the highest scores.
Evaluative arguments are generated in GEA (Generator of Evaluative
Arguments) to describe an entity as good or negative. An entity is divided
into a hierarchy of features, with a relevance score generated individually
for each feature depending on the user's preferences and the product's
value of that feature. The most relevant characteristics for the current user
are chosen during content selection.
8.5 SUMMARIZATION EVALUATION
The purpose of automatic summarization1 is to take an information

source, extract content from it, and deliver the most significant content to
the user in a condensed form that is sensitive to the demands of the user or
116
116
application. Summaries can be user-focused (or topic-focused, or query- Text Summarization, Text
Classification
focused), that is, tailored to the needs of a specific user or set of users, or
they can be 'generic,' that is, directed at a certain—usually broad—
readership community. They can be in the form of an extract, which is a
summary made entirely of content copied from the input, or an abstract,
which is a summary made up of at least some material not found in the
input.
Automatic summarization has long been interested in a review, with
thorough evaluations dating back to the 1960s, for example. While there
hasn't been a lot of agreement on evaluation issues, a reasonably clear
picture emerges from looking at the evaluation approaches that have been
studied thus far. These strategies will be briefly described and evaluated in
this short work.
• Communicated in plain language. There may be a proper answer in

circumstances when the output is an answer to a question, but it is
difficult to determine what the correct result is in other cases.
There's always the chance that a system will produce an excellent
summary that differs significantly from any human summary used as
a rough approximation to the correct result.
• Review could skyrocket. Because it is easy to repeat, an evaluation
that uses a score program rather than human judgments is ideal.
Because summarization entails compression, it's critical to be able to
assess summaries at various compression rates. This expands the scale and
the evaluation's complexity.
These considerations must be considered since summarising entails
presenting information in a way that is sensitive to the demands of the user
or application. As a result, the design of an evaluation becomes more
difficult.
There are two basic groups of methods for evaluating text summarization
(and, indeed, natural language processing systems). The first, a self-
evaluation, puts the summarization method to the test. The second, called
an extrinsic evaluation, looks at how the summary affects the completion
of another job. The coherence and informativeness of summaries have
been the focus of intrinsic evaluations. Extrinsic evaluations, on the other
hand, have looked at how summarization affects tasks such as relevance
assessment, reading comprehension, and so on.
8.6 TEXT CLASSIFICATION
Over the last few decades, text categorization problems have been
extensively researched and handled in a variety of real-world applications.
Many researchers are currently interested in building applications that use
text categorization algorithms, especially in light of recent achievements
in Natural Language Processing (NLP) and text mining. Feature
extraction, dimension reductions, classifier selection, and assessments are
the four processes that most text classification and document
117
Natural Language Processing categorization systems go through. The organization and technical
implementations of text classification systems are discussed in this study
in terms of the pipeline shown in Figure 1.
The raw text data collection serves as the pipeline's initial input. Text data
sets, in general, contain text sequences in documents like D = X1, X2,...,
XN, where Xi refers to a data point (i.e., document, text segment) with s
number of sentences, each sentence containing ws words with lw letters. A
class value is assigned to each point from a collection of k discrete value
indices.
Then, for our training purposes, we should develop a structured collection
called Feature Extraction. The dimensionality reduction step is an optional
part of the pipeline that could be part of the classification system (for
example, if we use Term Frequency-Inverse Document Frequency (TF-
IDF) as our feature extraction and the train set has 200k unique words,
computational time will be very expensive, so we could reduce this option
by bringing feature space into another dimensional space). Choosing the
correct classification algorithm is the most important stage in document
categorization.
The evaluation step, which is separated into two parts, is the other half of
the pipeline (prediction the test set and evaluating the model). In general,
there are four main levels of scope that can be applied to the text
classification system:
1. Document-level: The method collects the necessary categories of a
full document at the document level.
2. Paragraph level: The algorithm obtains the appropriate categories of
a single paragraph at the paragraph level (a portion of a document).
3. Sentence level: Gets the relevant categories of a single sentence (a
section of a paragraph) at the sentence level.
4. Sub-sentence level: The algorithm obtains the relevant categories of
sub-expressions within a sentence (a piece of a sentence) at the sub-
sentence level.
Figure 1 shows a flowchart of the text categorization pipeline.

(I) Text and Document Feature Extraction: Texts and documents are
unstructured data sets in general. When employing mathematical
modeling as part of a classifier, however, these unstructured text
sequences must be translated into a structured feature space. To
118
118
begin, the data must be cleaned to remove any extraneous characters Text Summarization, Text
Classification
or words. Formal feature extraction approaches can be used once the
data has been cleansed. Term Frequency-Inverse Document
Frequency (TF-IDF), Term Frequency (TF), Word2Vec, and Global
Vectors for Word Representation (GloVe) are some of the most
prevalent features extraction techniques.
(II) Dimensionality Reduction: Because text or document data sets
frequently contain a large number of unique words, data pre-
processing operations might take a long time and require a lot of
memory. Using low-cost algorithms is a typical approach to this
problem. This kind of low-cost algorithm, however, does not
perform as well as expected in particular data sets. Many researchers
prefer to employ dimensionality reduction to lower the time and
memory complexity of their applications to avoid performance
degradation. Pre-processing with dimensionality reduction could be
more efficient than constructing low-cost classifiers.
Principal Component Analysis (PCA), Linear Discriminant Analysis
(LDA), and non-negative matrix factorization are some of the most
prominent dimensionality reduction approaches (NMF). We also go
through new strategies for dimensionality reduction in unsupervised
feature extraction, such as random projection, autoencoders, and t-
distributed stochastic neighbor embedding (t-SNE).
(III) Text Classification Techniques: Choosing the optimal classifier is
the most critical phase in the text classification pipeline. We can't
properly choose the most efficient model for a text categorization
application without a comprehensive conceptual knowledge of each
approach. Following that, we discuss ensemble-based learning
approaches like boosting and bagging, which have primarily been
applied to query learning strategies and text analysis. Logistic
regression (LR) is a simple classification approach that has been
addressed in most data mining applications. The history of
information retrieval can be traced back to the dawn of time.
The Naive Bayes Classifier (NBC) was a very popular application.
We'll take a quick look at the Naive Bayes Classifier, which is
computationally light and requires very little memory. Non-
parametric approaches, such as a k-nearest neighbor, have been
investigated and employed as classification tasks. Another
prominent technique is the Support Vector Machine (SVM).
For document categorization, it uses a discriminative classifier. This
method can be applied to any situation. Bioinformatics, image,
video, human activity classification, safety, and other data mining
disciplines security, and so on Many researchers utilize this model as
a benchmark against which to assess their findings.
Attempts to bring attention to new ideas and contributions
119
Natural Language Processing Non-parametric approaches, such as a k-nearest neighbor, have been
investigated and employed as classification tasks. Another
prominent technique is the Support Vector Machine (SVM). For
document categorization, it uses a discriminative classifier. This
method can be applied to any situation.
Bioinformatics, image, video, human activity classification, safety,
and other data mining disciplines security, and so on Many
researchers utilize this model as a benchmark against which to
assess their findings. Attempts to bring attention to new ideas and
contributions Deep learning approaches have recently outperformed
prior machine learning algorithms on problems including image
classification, natural language processing, and face recognition,
among others. The ability of these deep learning algorithms to
represent complicated and non-linear relationships inside data is
critical to their success.
(IV) Evaluation: The text categorization pipeline's final step is
evaluation. The use and development of text categorization methods
require an understanding of how a model works. There are numerous
methods for assessing supervised techniques. The simplest way of
evaluation is accuracy calculation, but it does not work for
unbalanced data sets. F Score, Matthews Correlation Coefficient
(MCC), receiver operating characteristics (ROC), and area under the
ROC curve are all ways for evaluating text categorization systems
(AUC).
feature extractions, dimensionality reduction, and other pipeline steps
Techniques of classification and methods of evaluation In this part, state-
of-the-art techniques are compared based on a variety of criteria, including
model architecture, the novelty of the work, feature extraction technique,
corpus (the data set/s used), validation measure, and limitation of each
work. Choosing a feature extraction approach is necessary for determining
the optimum system for a certain application. Because some feature
extraction strategies are inefficient for a specific application, this decision
is entirely dependent on the application's purpose and data source. For a
given use, it is effective. for example, is trained on Wikipedia and does not
perform as well as TF-IDF when used for short text messages such as
short message service (SMS). Furthermore, due to the tiny amount of data,
this model cannot be trained as well as other strategies. The categorization
technique is the next step in this pipeline, where we briefly discuss the
limitations and drawbacks of each technique.
Text classification is a serious difficulty for scholars in many domains and
fields. Text categorization methods are extensively used in information
retrieval systems and search engine applications. Text classification could
be utilized for information filtering (e.g., email and text message spam
filtering) as an extension of existing applications. Following that, we
discuss document categorization's acceptance in public health and human
behavior. Document organization and knowledge management are two
120
120
further areas where text classification has aided. Finally, we'll talk about Text Summarization, Text
Classification
recommender systems, which are common in marketing and advertising.
TEXT PREPROCESSING
For text classification applications, feature extraction and pre-processing
are critical tasks. This section introduces strategies for cleaning text data
sets, reducing implicit noise and allowing for more accurate analysis. For
enlightenment. We also go over two popular strategies for extracting text
features: weighted words and word embedding.
CLEANING AND PRE-PROCESSING OF TEXT
Many redundant words, such as stopwords, misspellings, and slang, can be
found in most text and document data sets. Noise and superfluous
characteristics can degrade system performance in many algorithms,
particularly statistical and probabilistic learning algorithms. In this section,
we'll go through several text cleaning and pre-processing approaches and
methodologies.
TOKENIZATION
Tokenization is a pre-processing technique that divides a stream of text
into tokens, which are words, phrases, symbols, or other meaningful
pieces. The investigation of the words in a sentence is the primary purpose
of this step. Both text classification and text mining necessitate the use of
a parser to handle the tokenization of documents, such as the following
sentence:
He opted to sleep for another four hours after four hours of sleep.
The tokens in this scenario are as follows:
"After" "four" "hours of" "sleeping," "he" "decided" "to" "sleep" "for"
"another" "four."
STOP WORDS
Many terms in text and document classification are not significant enough
to be employed in classification algorithms, such as "a", "about", "above",
"across", "after", "afterwards", "again", and so on. To deal with these
words, the most typical method is to eliminate them from texts and
documents.
CAPITALIZATION
To compose a sentence, text and document data points have a variety of
caps. Since
When a document has a lot of sentences, different capitalization might be
a real pain.
the classification of huge documents Dealing with uneven capitalization is
the most usual solution.
121
Natural Language Processing is to change all of the letters to lower case. This approach converts all text
and document words into
the same feature area, yet it complicates the understanding of some words
significantly
(For instance, "US" (United States of America) to "us" (pronoun))
Converters for slang and abbreviations are available.
assist in accounting for these occurrences
ABBREVIATIONS AND SLANG
Other types of text abnormalities that are handled in the pre-processing
step include slang and abbreviation. SVM stands for Support Vector
Machine, and an abbreviation is a shorter form of a word or phrase that
contains mostly first letters from the terms.
Slang is a subset of spoken or written language that has several meanings,
such as "lost the plot," which essentially means "gone insane". Converting
these terms into formal language is a typical approach to dealing with
them
NOISE ABATEMENT
Many unneeded characters, such as punctuation and special characters,
can be found in most text and document data sets. Although critical
punctuation and special characters are necessary for human
comprehension of documents, they can harm categorization systems.
CORRECTIONS TO SPELLING
Spelling correction is a pre-processing step that can be skipped. Typos
(short for typographical errors) are widespread in texts and documents,
particularly in text data sets from social media (e.g., Twitter). This
challenge has been addressed by many algorithms, strategies, and methods
in NLP . For researchers, there are a variety of strategies and methods
accessible, including hashing-based and context-sensitive spelling
correction algorithms, as well as spelling correction utilizing the Trie and
Damerau–Levenshtein distance bigrams.
STEMMING
In NLP, a single word might exist in multiple forms (for example, singular
and plural noun forms), all of which have the same semantic meaning.
Stemming is one way for combining different variants of a word into the
same feature space. Text stemming uses linguistic processes like affixation
(adding affixes) to modify words to produce variant word forms. The stem
of the word "studying," for example, is "study."
LEMMATIZATION
Lemmatization is a natural language processing technique that replaces a
word's suffix with a different one or eliminates the suffix to reveal the
fundamental word form (lemma)
122
122
8.7 LET US SUM UP Text Summarization, Text
Classification
The text summary is a technique for condensing vast passages of text into
manageable chunks. In machine learning and natural language processing,
text summarization is a prevalent challenge. The main issues are topic
identification, interpretation, summary generation, and evaluation of the
created summary. Text summarization is the process of constructing a
concise, cohesive, and fluent summary of a lengthier text document. This
involves finding key sentences or phrases from the original text and
piecing together sections to create a shortened version.
It is a prevalent challenge in machine learning and natural language
processing (NLP). The bag of words model is used to describe N-
dimensional vectors to define similarity. Matrix B is the adjacency matrix
formed by dividing each element by the associated row sum in the
similarity graph. The degree of each node is called degree centrality in this
case
8.8 REFFERENCES
• https://www.sciencedirect.com/topics/computer-science/text-
summarization
• https://towardsdatascience.com/a-quick-introduction-to-text-
summarization-inmachine-learning-3d27ccf18a9f
• https://iq.opengenus.org/lexrank-text-summarization/
• Text Classification Algorithms: A Survey Kamran Kowsari 1,2,* ID
, Kiana Jafari Meimandi 1 , Mojtaba Heidarysafa 1 , Sanjana Mendu
1 ID , Laura Barnes 1,2,3 ID and Donald Brown 1,3 I
• Summarization Evaluation: An Overview Inderjeet MANI The
MITRE Corporation, W640 11493 Sunset Hills Road Reston, VA
20190-5214, USA
• AN OPTIMIZATION APPROACH TO AUTOMATIC GENERIC
DOCUMENT SUMMARIZATION RASIM M. ALGULIEV,
RAMIZ M. ALIGULIYEV, AND CHINGIZ A. MEHDIYEV
Institute of Information Technology of Azerbaijan National
Academy of Sciences, Baku, Azerbaijan
• “Optimization-based Content Selection for Opinion Summarization
Jackie Chi Kit Cheung Department of Computer Science University
of Toronto Toronto”, ON, M5S 3G4, Canada
jcheung@cs.toronto.edu Giuseppe Carenini and Raymond T. N
8.9 BIBLIOGRAPHY
• https://www.sciencedirect.com/topics/computer-science/text-
summarization
• https://towardsdatascience.com/a-quick-introduction-to-text-
summarization-inmachine-learning-3d27ccf18a9f
• https://iq.opengenus.org/lexrank-text-summarization/ 123
Natural Language Processing • “Text Classification Algorithms: A Survey Kamran Kowsari” 1,2,*
ID , Kiana Jafari Meimandi 1 , Mojtaba Heidarysafa 1 , Sanjana
Mendu 1 ID , Laura Barnes 1,2,3 ID and Donald Brown 1,3 I
• “Summarization Evaluation: An Overview Inderjeet MANI The
MITRE Corporation”, W640 11493 Sunset Hills Road Reston, VA
20190-5214, USA
• “AN OPTIMIZATION APPROACH TO AUTOMATIC GENERIC
DOCUMENT SUMMARIZATION RASIM M. ALGULIEV,
RAMIZ M. ALIGULIYEV, AND CHINGIZ A. MEHDIYEV”
Institute of Information Technology of Azerbaijan National
Academy of Sciences, Baku, Azerbaijan
• “Optimization-based Content Selection for Opinion Summarization
Jackie Chi Kit Cheung Department of Computer Science University
of Toronto Toronto”, ON, M5S 3G4, Canada
jcheung@cs.toronto.edu Giuseppe Carenini and Raymond T. N
8.10 UNIT EXCERISE
1. What should the input dimension for text summarization be?

2. Is it feasible to categorize a conversation's topic?
3. What are the most effective methods for text simplification?
4. How would you address a problem of text summarization?
5. How do you go about putting text summarising into practice?
6. What does a categorization text serve?
7. What are the various approaches to text classification?
8. What exactly is a text classification issue?

124
124
9
SENTIMENT ANALYSIS AND
OPINION MINING - I
Unit Structure
9.0 Objective
9.1 Introduction
9.1.1 Definition of Sentiment Analysis
9.2 Sentiment analysis types
9.2.1 Sentiment Analysis - Affective lexicons
9.2.2 Learning affective lexicons
9.3.1 Computing with affective lexicons
9.3.2 Aspect based sentiment analysis
9. OBJECTIVE
• Opinion mining and sentiment analysis done from big data

• Opinion mining in data mining
• Sentiment analysis in text mining
• Explain sentiment analysis
9.1 INTRODUCTION
Sentiment analysis systems help organizations gather insights from

unorganized and unstructured text that comes from online sources such as
emails, blog posts, support tickets, web chats, social media channels,
forums and comments. Algorithms replace manual data processing by
implementing rule-based, automatic or hybrid methods. Rule-based
systems perform sentiment analysis based on predefined, lexicon-based
rules while automatic systems learn from data with machine learning
techniques. A hybrid sentiment analysis combines both approaches.
In addition to identifying sentiment, opinion mining can extract the
polarity (or the amount of positivity and negativity), subject and opinion
holder within the text. Furthermore, sentiment analysis can be applied to
varying scopes such as document, paragraph, sentence and sub-sentence
levels.
Vendors that offer sentiment analysis platforms or SaaS products include
Brandwatch, Hootsuite, Lexalytics, NetBase, Sprout Social, Sysomos and
Zoho. Businesses that use these tools can review customer feedback more
regularly and proactively respond to changes of opinion within the market.
125
Natural Language Processing Sentiment analysis, also called opinion mining, is the field of study that
analyzes people’s opinions, sentiments, evaluations, appraisals, attitudes,
and emotions towards entities such as products, services, organizations,
individuals, issues, events, topics, and their attributes. It represents a large
problem space. There are also many names and slightly different tasks,
e.g., sentiment analysis, opinion mining, opinion extraction, sentiment
mining, subjectivity analysis, affect analysis, emotion analysis, review
mining, etc. However, they are now all under the umbrella of sentiment
analysis or opinion mining. While in industry, the term sentiment analysis
is more commonly used, but in academia both sentiment analysis and
opinion mining are frequently employed. They basically represent the
same field of study., we use the terms sentiment analysis and opinion
mining interchangeably.
To simplify the presentation, throughout this book we will use the term
opinion to denote opinion, sentiment, evaluation, appraisal, attitude, and
emotion. However, these concepts are not equivalent. We will distinguish
them when needed. The meaning of opinion itself is still very broad.
Sentiment analysis and opinion mining mainly focuses on opinions which
express or imply positive or negative sentiments. Although linguistics and
natural language processing (NLP) have a long history, little research had
been done about people’s opinions and sentiments before the year 2000.
Since then, the field has become a very active research area. There are
several reasons for this. First, it has a wide arrange of applications, almost
in every domain.
The industry surrounding sentiment analysis has also flourished due to the
proliferation of commercial applications. This provides a strong
motivation for research. Second, it offers many challenging research
problems, which had never been studied before. Sentiment Analysis and
Opinion Mining 8 the first time in human history, we now have a huge
volume of opinionated data in the social media on the Web. Without this
data, a lot of research would not have been possible. Not surprisingly, the
inception and the rapid growth of sentiment analysis coincide with those
of the social media. In fact, sentiment analysis is now right at the center of
the social media research. Hence, research in sentiment analysis not only
has an important impact on NLP, but may also have a profound impact on
management sciences, political science, economics, and social sciences as
they are all affected by people’s opinions. Although the sentiment analysis
research mainly started from early 2000, there were some earlier work on
interpretation of metaphors, sentiment adjectives, subjectivity, view
points, and affects (Hatzivassiloglou and McKeown, 1997; Hearst, 1992;
Wiebe, 1990; Wiebe, 1994; Wiebe, Bruce and O'Hara, 1999).
9.1.1 Definition of Sentiment Analysis
Sentiment analysis is contextual mining of text which identifies and
extracts subjective information in source material, and helping a
business to understand the social sentiment of their brand, product or
service while monitoring online conversations.
126
126
Sentiment Analysis is the most common text classification tool that Sentiment Analysis and
Opinion Mining - I
analyses an incoming message and tells whether the underlying sentiment
is positive, negative our neutral.
structures, from the levels of phrases, clauses, sentences and paragraphs to
the level of the writing as a whole, to their language-
independent meanings. It also involves removing features specific to
particular linguistic and cultural contexts, to the extent that such a project
is possible. The elements of idiom and figurative speech, being cultural,
are often also converted into relatively invariant meanings in semantic
analysis. Semantics, although related to pragmatics, is distinct in that the
former deals with word or sentence choice in any given context, while
pragmatics considers the unique or particular meaning derived from
context or tone. To reiterate in different terms, semantics is about
universally coded meaning, and pragmatics, the meaning encoded in
words that is then interpreted by an audience.[1]
Semantic analysis can begin with the relationship between individual
words. This requires an understanding of lexical hierarchy,
including hyponymy and hypernymy, meronomy, polysemy, synonyms,
antonyms, and homonyms.[2] It also relates to concepts like connotation
(semiotics) and collocation, which is the particular combination of words
that can be or frequently are surrounding a single word. This can include
idioms, metaphor, and simile, like, "white as a ghost."
Some important remarks about this definition are in order:
1. In this definition, we purposely use subscripts to emphasize that the
five pieces of information in the quintuple must correspond to one
another. That is, the opinion must be given by opinion holder about
aspect of entity at time. Any mismatch is an error.
2. The five components are essential. Missing any of them is
problematic in general. For example, if we do not have the time
component, we will not be able to analyze opinions on an entity
according to time, which is often very important in practice because
an opinion two years ago and an opinion yesterday is not the same.
Without opinion holder is also problematic. For example, in the
sentence “the mayor is loved by the people in the city, but he has
been criticized by the state government,” the two opinion holders,
“people in the city” and “state government,” are clearly important
for applications.
3. The definition covers most but not all possible facets of the semantic
meaning of an opinion, which can be arbitrarily complex. For
example, it does not cover the situation in “The view finder and the
lens are too close,” which expresses an opinion on the distance of
two parts. It also does not cover the context of the opinion, e g.,
“This car is too small for a tall person,” which does not say the car is
too small for everyone. “Tall person” is the context here. Note also
that in the original definition of entity, it is a hierarchy of parts, sub- 127
Natural Language Processing parts, and so on. Every part can have its set of attributes. Due to the
simplification, the quintuple representation can result in information
loss. For example, “ink” is a part/component of a printer. In a printer
review, one wrote “The ink of this printer is expensive.” This does
not say that the printer is expensive (which indicates the aspect
price). If one does not care about any attribute of the ink, this
sentence just gives a negative opinion to the ink, which is an aspect
of the printer entity. However, if one also wants to study opinions
about different aspects of the ink, e.g., price and quality, the ink
needs to be treated as a separate entity. Then, the quintuple
representation still applies, but the part-of relationship needs to be
saved. Of course, conceptually we can also expand the
representation of opinion target using a nested relation. Despite the
limitations, the definition does cover the essential information of an
opinion which is sufficient for most applications.
4. This definition provides a framework to transform unstructured text
to structured data. The quintuple above is basically a database
schema, based on which the extracted opinions can be put into a
database table. Then a rich set of qualitative, quantitative, and trend
analyses of opinions can be performed using the whole suite of
database management systems (DBMS) and OLAP tools.
5. The opinion defined here is just one type of opinion, called regular
opinion. Another type is comparative opinion (Jindal and Liu,
2006b; Liu, 2006 and 2011), which needs a different definition.
Section 2.3 will discuss different types of opinions. Chapter 8
defines and analyzes comparative opinions. For the rest of this
section, we only focus on regular opinions. For simplicity, we just
called them opinions.
9.2 SENTIMENT ANALYSIS TYPES
Modern-day sentiment analysis approaches are classified into three

categories: knowledge-based, statistical, and hybrid. Here's how to
perform sentiment analysis. Knowledge-Based: This approach included
the classification of text based on words that emanate emotion.
128
128
One of the things that define us as human Is our ability to showcase a wide Sentiment Analysis and
Opinion Mining - I
range of emotions. We can be sad, excited, worried, and angry within
seconds. We can learn and analyze situations like no other animal can.
These are just some of the things that makes us unique and special.
This is evident during shopping, too. Almost all purchases are emotion-
driven. It could be out of fear, jealousy, or happiness. With emotions
playing a critical role in customer behavior, it has become essential for
brands to analyze the sentiments of their consumers.
Here’s where the concept of sentiment analysis comes into play. In this
post, we’ll discuss the idea and different types of sentiment analysis.
Sentiment Analysis Research
As discussed above, pervasive real-life applications are only part of the
reason why sentiment analysis is a popular research problem. It is also
highly challenging as a NLP research topic, and covers many novel
subproblems as we will see later. Additionally, there was little research
before the year 2000 in either NLP or in linguistics. Part of the reason is
that before then there was little opinion text available in digital forms.
Since the year 2000, the field has grown rapidly to become one of the
most active research areas in NLP. It is also widely researched in data
mining, Web mining, and information retrieval. In fact, it has spread from
computer science to management sciences (Archak, Ghose and Ipeirotis,
2007; Chen and Xie, 2008; Das and Chen, 2007; Dellarocas, Zhang and
Awad, 2007; Ghose, Ipeirotis and Sundararajan, 2007; Hu, Pavlou and
Zhang, 2006; Park, Lee and Han, 2007).
Different Levels of Analysis
Introduction to the main research problems based on the level of
granularities of the existing research. In general, sentiment analysis has
been investigated mainly at three levels:
Document level: The task at this level is to classify whether a whole
opinion document expresses a positive or negative sentiment. For
example, given a product review, the system determines whether the
review expresses an overall positive or negative opinion about the product.
This task is commonly Sentiment Analysis and Opinion Mining 11 known
as document-level sentiment classification. This level of analysis assumes
that each document expresses opinions on a single entity (e.g., a single
product). Thus, it is not applicable to documents which evaluate or
compare multiple entities.
Sentence level: The task at this level goes to the sentences and determines
whether each sentence expressed a positive, negative, or neutral opinion.
Neutral usually means no opinion. This level of analysis is closely related
to subjectivity classification (Wiebe, Bruce and O'Hara, 1999), which
distinguishes sentences (called objective sentences) that express factual
information from sentences (called subjective sentences) that express
subjective views and opinions. However, we should note that subjectivity
129
Natural Language Processing is not equivalent to sentiment as many objective sentences can imply
opinions, e.g., “We bought the car last month and the windshield wiper
has fallen off.” Researchers have also analyzed clauses (Wilson, Wiebe
and Hwa, 2004), but the clause level is still not enough, e.g., “Apple is
doing very well in this lousy economy.” Entity and Aspect level: Both the
document level and the sentence level analyses do not discover what
exactly people liked and did not like.
Aspect level performs finer-grained analysis. Aspect level was earlier
called feature level (feature-based opinion mining and summarization) .
Instead of looking at language constructs (documents, paragraphs,
sentences, clauses or phrases), aspect level directly looks at the opinion
itself. It is based on the idea that an opinion consists of a sentiment
(positive or negative) and a target (of opinion). An opinion without its
target being identified is of limited use. Realizing the importance of
opinion targets also helps us understand the sentiment analysis problem
better. For example, although the sentence “although the service is not that
great, I still love this restaurant” clearly has a positive tone, we cannot say
that this sentence is entirely positive. In fact, the sentence is positive about
the restaurant (emphasized), but negative about its service (not
emphasized). In many applications, opinion targets are described by
entities and/or their different aspects.
Thus, the goal of this level of analysis is to discover sentiments on entities
and/or their aspects. For example, the sentence “The iPhone’s call quality
is good, but its battery life is short” evaluates two aspects, call quality and
battery life, of iPhone (entity). The sentiment on iPhone’s call quality is
positive, but the sentiment on its battery life is negative. The call quality
and battery life of iPhone are the opinion targets. Based on this level of
analysis, a structured summary of opinions about entities and their aspects
can be produced, which turns unstructured text to structured data and can
be used for all kinds of qualitative and quantitative analyses. Both the
document level and sentence level classifications are already Sentiment
Analysis and Opinion Mining 12 highly challenging. The aspect-level is
even more difficult. It consists of several sub-problems, which we will
discuss.
To make things even more interesting and challenging, there are two types
of opinions, i.e., regular opinions and comparative opinions .
A regular opinion expresses a sentiment only on a particular entity or an
aspect of the entity, e.g., “Coke tastes very good,” which expresses a
positive sentiment on the aspect taste of Coke.
A comparative opinion compares multiple entities based on some of their
shared aspects, e.g., “Coke tastes better than Pepsi,” which compares Coke
and Pepsi based on their tastes (an aspect) and expresses a preference for
Coke ,
130
130
What Is Sentiment Analysis? Sentiment Analysis and
Opinion Mining - I
First things first, what is sentiment analysis? Sentiment analysis is a type
of market analysis that includes the use of text analysis, biometrics,
natural language processing (NLP), and computational linguistics to
recognize the state of the said information.
In simple terms, it’s the process of determining whether a piece of content
– email, social media post, or article – is negative, positive, or neutral.
Sentiment analysis enables you to ascertain public opinion and understand
consumer experiences.
But why should you even bother about sentiment analysis? For starters,
it’s extremely helpful in social media monitoring. It helps you gauge
public opinion on certain topics on an enormous scale.
Besides, it can play a pivotal role in market research and customer service.
With sentiment analysis, you can see what people think about your
products or your competitors’ products. This helps you understand
customer attitudes and preferences, enabling you to make informed
decisions.
Types of Sentiment Analysis
People have a wide range of emotions – sad or happy, interested or
uninterested, and positive or negative. Different sentiment analysis models
are available to capture this variety of emotions.
Let’s look at the most important types of sentiment analysis.
1. Fine-Grained
This sentiment analysis model helps you derive polarity precision. You
can conduct a sentiment analysis across the following polarity categories:
very positive, positive, neutral, negative, or very negative. Fine-grained
sentiment analysis is helpful for the study of reviews and ratings.
For a rating scale from 1 to 5, you can consider 1 as very negative and five
as very positive. For a scale from 1 to 10, you can consider 1-2 as very
negative and 9-10 as very positive.
2. Aspect-Based
While fine-grained analysis helps you determine the overall polarity of
your customer reviews, aspect-based analysis delves deeper. It helps you
determine the particular aspects people are talking about.
Let’s say; you’re a mobile phone manufacturer, and you get a customer
review stating, “the camera struggles in artificial lighting conditions.”
With aspect-based analysis, you can determine that the reviewer has
commented on something “negative” about the “camera.”
131
Natural Language Processing 3. Emotion Detection
As the name suggests, emotion detection helps you detect emotions. This
can include anger, sadness, happiness, frustration, fear, worry, panic, etc.
Emotion detection systems typically use lexicons – a collection of words
that convey certain emotions. Some advanced classifiers also utilize robust
machine learning (ML) algorithms.
It’s recommended to use ML over lexicons because people express
emotions in a myriad of ways. Take this line, for example: “This product
is about to kill me.” This line may express feelings of fear and panic.
A similar line – this product is killing it for me – has an entirely different
and positive meaning. But the word “kill” might be associated with fear or
panic in the lexicon. This may lead to inaccurate emotion detection.
4. Intent Analysis
Accurately determining consumer intent can save companies time, money,
and effort. So many times, businesses end up chasing consumers that don’t
plan to buy anytime soon. Accurate intent analysis can resolve this hurdle.
The intent analysis helps you identify the intent of the consumer – whether
the customer intends to purchase or is just browsing around.
If the customer is willing to purchase, you can track them and target them
with advertisements. If a consumer isn’t ready to buy, you can save your
time and resources by not advertising to them.
9.2.1 Sentiment Analysis - Affective lexicons
Definition:
● It is a Natural Language Processing task; Sentiment Analysis refers
to finding patterns in data and inferring the emotion of the given
piece of information which could be classified into one of these
categories:
• Negative
• Neutral
• Positive
Utility:
• We are truly living in the information age where the data is

generated by both humans and machines at an unprecedented rate,
therefore it’s nearly impossible to gain insights into such data for
making intelligent decisions, manually. One such insight is
assessing/calculating the sentiment of a big (rather huge) dataset
• Software to the rescue yet again. One way of assessing, rather
calculating the emotion on top of such a huge dataset is by way of
employing sentiment analysis techniques (more on those later)
132
132
Applications and examples: Sentiment Analysis and
Opinion Mining - I
● Market sentiment – Huge amount of data is being generated in the
global financial world, every second. It is imperative for investors to
remain on top of such information so that they can manage their
portfolios more effectively. One way of gauging the market
sentiment is by way of implementing sentiment analyzing tools
● Gauging the popularity of the products, books, movies, songs,
services by analyzing the data generated from different sources such
as tweets, reviews, feedback forums, gestures (swipes, likes) etc.
Now that we know what sentiment analysis is on a broader level, let’s dig
a little deeper. Analyzing data can be quite tricky because the data, in
almost all the cases, is unstructured, opinionated and in some cases
incomplete- meaning it is addressing only a fraction of the subject matter
and therefore unsuitable for calculating the emotion. This data must be
cleaned for it to be turned into something actionable. It is intended, in
most cases, to focus on deducing the polarity (good or bad, spam or ham,
black or white etc.) from the given piece of information as the opinions
can be nuanced and can lead to ambiguity. Putting it another way, the
“problem” should be framed in a certain way so that a binary classification
(polarity) can be deduced from it. Now I know most of you will say, what
about that “neutral” classification between “negative” and “positive”
classes? Fair question. Well the answer is, if the deduction/calculation
comes out to be somewhere in the middle of 0 (negative) and 1 (positive),
let’s say 0.5 then it is obvious that the given piece of information also
known as problem instance can neither be classified
as negative nor positive and therefore the classification of neutral seems
more appropriate.
Analysis Mechanisms:
On a higher level, there are two techniques that can be used for performing
sentiment analysis in an automated manner, these are: Rule-
based and Machine Learning based. I will explore the former in this blog
and take up the latter in part 2 of the series.
• Rule based
Rule based sentiment analysis refers to the study conducted by the
language experts. The outcome of this study is a set of rules (also known
as lexicon or sentiment lexicon) according to which the words classified
are either positive or negative along with their corresponding intensity
measure.
Generally, the following steps are needed to be performed while applying
the rule-based approach:
• Extract the data

• Tokenize text. The task of splitting the text into individual words
133
Natural Language Processing • Stop words removal. Those words which do not carry any significant
meaning and should not be used for the analysis activity. Examples
of stop words are: a, an, the, they, while etc.
• Punctuation removal (in some cases)
• Running the preprocessed text against the sentiment lexicon which
should provide the number/measurement corresponding to the
inferred emotion
• Example:
• Consider this problem instance: “Sam is a great guy.”
• Remove stop words and punctuation
• Running the lexicon on the preprocessed data, returns a positive
sentiment score/measurement because of the presence of a positive
word “great” in the input data.
• This is a very simple example. Real world data is much more
complex and nuanced e.g. your problem sentiment may contain
sarcasm (where seemingly positive words carry negative meaning or
vice versa), shorthand, abbreviations, different spellings (e.g. flavor
vs flavour), misspelled words, punctuation (especially question
marks), slang and of course, emojis.
• In order to tackle the complex data for analysis is to make use of
sophisticated lexicons that can take into consideration the intensity
of the words (e.g. if a word is positive then how positive it is,
basically there is a difference between good, great and amazing and
that is represented by the intensity assigned to a given word), the
subjectivity or objectivity of the word and the context also. There
are several such lexicons available. Following are a couple of
popular ones:
• VADER (Valence Aware Dictionary and Sentiment
Reasoner): Widely used in analyzing sentiment on social
media text because it has been specifically attuned to analyze
sentiments expressed in social media (as per the linked docs).
It now comes out of the box in Natural language toolkit,
NLTK. VADER is sensitive to both polarity and the intensity.
Here is how to read the measurements:
• -4: extremely negative
• 4: extremely positive
• 0: Neutral or N/A
Examples (taken from the docs):
TextBlob: Very useful NLP library that comes prepackaged with its own
sentiment analysis functionality. It is also based on NLTK. The sentiment
property of the api/library returns polarity and subjectivity.
• Polarity range: -1.0 to 1.0

• Subjectivity range: 0.0 – 1.0 (0.0 is very objective and 1.0 is very
subjective)
• Example (from the docs):
134
134
• Problem instance: “Textblob is amazingly simple to use. What great Sentiment Analysis and
Opinion Mining - I
fun!”
• Polarity: 39166666666666666
• Subjectivity: 0.39166666666666666
• Polarity: as mentioned earlier, it is a measurement of how positive
or negative the given problem instance is. In other words, it is
related to the emotion of the given text
• Subjectivity: It refers to opinions or views (can be allegations,
expressions or speculations) that need to be analyzed in the given
context of the problem statement. The more subjective the instance
is, the less objective it becomes and vice versa. A subjective instance
(e.g. a sentence) may or may not carry any emotion. Examples:
• “Sam likes watching football”
• “Sam is driving to work”
• Sentiwordnet: This is also built into NLTK. It is used for opinion
mining. This helps in deducing the polarity information from the
given problem instance. SWN extends wordnet which is a lexical
database of words (the relationship between words, hence the
term net), developed at Princeton and is a part of NLTK corpus.
Here I’d focus primarily on synsets, which are the logical groupings
of nouns, verbs, adjectives and adverbs, into sets or collections of
cognitive synonyms, hence the term synset (you can read more on
wordnet online, I found the links this, this and this quite helpful).
Coming back to Sentiwordnet and its relationship with wordnet;
Sentiwordnet assigns polarity to each synset.
Example:
• Let’s take a word hunger

• The wordnet part:
• Get a list of all the synsets against hunger
• The Sentiwordnet (swn) part:
• Get the polarity for all those synsets
• Use one or maybe a combination of all of them to determine the
polarity score for that word (hunger in ours). The following
diagrams
9.2.2 Learning affective lexicons
More generally, we use the phrase affective lexicon to refer to that subset
of words in a language that are about affect or affective conditions.
Many, but by no means all, of the words in the affective lexicon refer to
emotions. Emotion lexicons describe the affective meaning of words and
thus constitute a centerpiece for advanced sentiment and emotion analysis.
Yet, manually curated lexicons are only available for a handful of
languages, leaving most languages of the world without such a precious
resource for downstream applications. Even worse, their coverage is often
limited both in terms of the lexical units they contain and the emotional
135
Natural Language Processing variables they feature. In order to break this bottleneck, we here introduce
a methodology for creating almost arbitrarily large emotion lexicons for
any target language. Our approach requires nothing but a source language
emotion lexicon, a bilingual word translation model, and a target language
embedding model. Fulfilling these requirements for 91 languages, we are
able to generate representationally rich high-coverage lexicons comprising
eight emotional variables with more than 100k lexical entries each. We
evaluated the automatically generated lexicons against human judgment
from 26 datasets, spanning 12 typologically diverse languages, and found
that our approach produces results in line with state-of-the-art
monolingual approaches to lexicon creation and even surpasses human
reliability for some languages and variables
Sentiment Lexicon and Its Issues
Not surprisingly, the most important indicators of sentiments are sentiment
words, also called opinion words. These are words that are commonly
used to express positive or negative sentiments. For example, good,
wonderful, and amazing are positive sentiment words, and bad, poor, and
terrible are negative sentiment words. Apart from individual words, there
are also phrases and idioms, e.g., cost someone an arm and a leg.
Sentiment words and phrases are instrumental to sentiment analysis for
obvious reasons. A list of such words and phrases is called a sentiment
lexicon (or opinion lexicon). Over the years, researchers have designed
numerous algorithms to compile such lexicons. Although sentiment words
and phrases are important for sentiment analysis, only using them is far
from sufficient. The problem is much more complex. In other words, we
can say that sentiment lexicon is necessary but not sufficient for sentiment
analysis.
Below, we highlight several issues:
1. A positive or negative sentiment word may have opposite
orientations in different application domains. For example, “suck”
usually indicates negative sentiment, e.g., “This camera sucks,” but
it can also imply positive sentiment, e.g., “This vacuum cleaner
really sucks.”
2. A sentence containing sentiment words may not express any
sentiment. This phenomenon happens frequently in several types of
sentences. Question (interrogative) sentences and conditional
sentences are two important types, e.g., “Can you tell me which
Sony camera is good?” and “If I can find a good camera in the shop,
I will buy it.” Both these sentences contain the sentiment word
“good”, but neither expresses a positive or negative opinion on any
specific camera. However, not all conditional sentences or
interrogative sentences express no sentiments, e.g., “Does anyone
know how to repair this terrible printer” and “If you Sentiment
Analysis and Opinion Mining 13 are looking for a good car, get
Toyota Camry.”
136
136
Sarcastic sentences with or without sentiment words are hard to deal with, Sentiment Analysis and
e.g., “What a great car! It stopped working in two days.” Sarcasms are not Opinion Mining - I
so common in consumer reviews about products and services, but are very
common in political discussions, which make political opinions hard to
deal with. Many sentences without sentiment words can also imply
opinions. Many of these sentences are actually objective sentences that are
used to express some factual information. Again, there are many types of
such sentences. Here we just give two examples. The sentence “This
washer uses a lot of water” implies a negative sentiment about the washer
since it uses a lot of resource (water). The sentence “After sleeping on the
mattress for two days, a valley has formed in the middle” expresses a
negative opinion about the mattress. This sentence is objective as it states
a fact. All these sentences have no sentiment words. These issues all
present major challenges. In fact, these are just some of the difficult
problems.
Natural Language Processing Issues Finally, we must not forget sentiment
analysis is a NLP problem. It touches every aspect of NLP, e.g.,
coreference resolution, negation handling, and word sense disambiguation,
which add more difficulties since these are not solved problems in NLP.
However, it is also useful to realize that sentiment analysis is a highly
restricted NLP problem because the system does not need to fully
understand the semantics of each sentence or document but only needs to
understand some aspects of it, i.e., positive or negative sentiments and
their target entities or topics. In this sense, sentiment analysis offers a
great platform for NLP researchers to make tangible progresses on all
fronts of NLP with the potential of making a huge practical impact
content.
Researchers now also have a much better understanding of the whole
spectrum of the problem, its structure, and core issues. Numerous new
(formal) models and methods have been proposed. The research has not
only deepened but also broadened significantly. Earlier research in the
field mainly focused on classifying the sentiment or subjectivity expressed
in documents or sentences, which is insufficient for most real-life
applications. Practical applications often demand more in-depth and fine-
grained analysis. Due to the maturity of the field, the book is also written
in a structured form in the sense that the problem is now better defined and
different research directions are unified around the definition.
Opinion Spam Detection
A key feature of social media is that it enables anyone from anywhere in
the world to freely express his/her views and opinions without disclosing
his/her true identify and without the fear of undesirable consequences.
These opinions are thus highly valuable. However, this anonymity also
comes with a price. It allows people with hidden agendas or malicious
intentions to easily game the system to give people the impression that
they are independent members of the public and post fake opinions to
promote or to discredit target products, services, organizations, or
individuals without disclosing their true intentions, or the person or
137
Natural Language Processing organization that they are secretly working for. Such individuals are called
opinion spammers and their activities are called opinion spamming.
Opinion spamming has become a major issue. Apart from individuals who
give fake opinions in reviews and forum discussions, there are also
commercial companies that are in the business of writing fake reviews and
bogus blogs for their clients. Several high-profile cases of fake reviews
have been reported in the news. It is important to detect such spamming
activities to ensure that the opinions on the Web are a trusted source of
valuable information. Unlike extraction of positive and negative opinions,
opinion spam detection is not just a NLP problem as it involves the
analysis of people’s posting behaviors. It is thus also a data mining
problem. Chapter 10 will discuss the current state-of-the-art detection
techniques.

138
138
10
SENTIMENT ANALYSIS AND
OPINION MINING - II
Unit Structure
10.0 Computing with affective lexicons
10.1 Aspect based sentiment analysis
10.0 COMPUTING WITH AFFECTIVE LEXICONS
Why compute affective meaning?

Affective computing is the study and development of systems and devices
that can recognize, interpret, process, and simulate human affects.
In general, affective computing refers to a computer's capabilities to
recognize a user's emotional states, to express its own emotions, and to
respond to the user's emotions (Picard, 1997).
Affective computing has potential applications in human–computer
interaction, such as affective mirrors allowing the user to see how he or
she performs; emotion monitoring agents sending a warning before
one sends an angry email; or even music players selecting tracks
based on mood.
Human-centered computation through affective computing can help in
recognizing emotions, and generate proper actions to have richer
communication. Applications of affective computing range from
education , autonomous driving , entertainment , to health-care.
Affective Computing. The artificial intelligence (AI) field picked up the
idea that human rational thinking depends on emotional
processing. Rosalind Picard's “Affective Computing” had a major effect
on both the AI and HCI fields (Picard, 1997)
139
Natural Language Processing What is an affective experience? As we use the term, an affective
experience is an experience that presents its objects in a positively or
negatively valenced way. To explain what we mean, we'll begin with
some examples. Our first example is the feeling of desire.
What is affective AI?
Emotion AI, also known as affective computing, is essentially all

about detecting emotions using artificial intelligence. Machines with
this kind of emotional intelligence are able to understand not only the
cognitive but also the emotive channels of human communication
What is meant by cognitive computing?
Cognitive computing is the use of computerized models to simulate the
human thought process in complex situations where the answers may
be ambiguous and uncertain. The phrase is closely associated with
IBM's cognitive computer system, Watson.
What principles guide our social behavior?
Affect refers to the feelings that we experience as part of life and includes
both moods and emotions. Social behavior is influenced by principles
of reciprocal altruism and social exchange.
Computer vision algorithms identify key landmarks on the face – for
example the corners of your eyebrows, the tip of your nose, the corners of
your mouth. Machine learning algorithms then analyse pixels in those
regions to classify facial expressions.
Narrow AI is goal-oriented, designed to perform singular tasks - i.e. facial
recognition, speech recognition/voice assistants, driving a car, or
searching the internet - and is very intelligent at completing the specific
task it is programmed to do.
140
140
Major sub-fields of AI now include: Machine Learning, Neural Sentiment Analysis and
Opinion Mining - II
Networks, Evolutionary Computation, Vision, Robotics, Expert
Systems, Speech Processing, Natural Language Processing, and
Planning.
Affective computing is becoming more and more important as it enables
to extend the possibilities of computing technologies by incorporating
emotions. In fact, the detection of users' emotions has become one of the
most important aspects regarding Affective Computing.
Affective computing is becoming more and more important as it enables
to extend the possibilities of computing technologies by incorporating
emotions. In fact, the detection of users' emotions has become one of the
most important aspects regarding Affective Computing.
Social Media are the platforms created using these Web 2.0 technologies. -
Social Computing is simply the branch of computer science that
focuses on analyzing the relationship between human behavior and
computer systems.
How do you identify primary emotions?
Thomas says that primary emotions are simply our initial reactions to
external events or stimuli. Secondary emotions are the reactions we then
have to our reaction
Knowing the Difference Between Primary and Secondary Emotions Could
Be the Key to Fighting Fairly with Your Partner
“It’s just emotion that’s taking me over,” they weren’t exaggerating—

emotions affect us in powerful ways, impacting our behavior and our
relationships. So, it’s important we understand them as best we can.
Which, if we’re being honest, is easier said than done. In fact, civilizations
have been trying to categorize human emotions for centuries. In today’s 141
Natural Language Processing modern era, one of those categorizations focuses on primary and
secondary emotions. Read on to learn more about these types
of emotions—and how to best identify them.
What are emotions in the first place?
Happy, sad, mad, scared, friendly, thoughtful, lonely, cranky, grateful,
delighted—the list goes on and on. How do you actually describe these
invisible forces? Merriam-Webster defines emotions as “a conscious
mental reaction (such as anger or fear) subjectively experienced as strong
feeling usually directed toward a specific object and typically
accompanied by physiological and behavioral changes in the body,”
meaning that these feelings aren’t just fleeting, but they have real-life
impacts on our actions a psychologist and emotional scientist who works
with emotionally sensitive people, likes to think of emotions as energy in
motion (e-motion—get it?).
So, what’s the difference between primary and secondary emotions?
Think of it most this way: Primary comes first and secondary comes, well,
second. From there, the psych field tends to differ on what exactly primary
vs. secondary emotions are.. Psychology Today, “In the 20th century, Paul
Ekman identified six basic emotions (anger, disgust, fear, happiness,
sadness, and surprise) and Robert Plutchik [identified] eight, which he
grouped into four pairs of polar opposites (joy-sadness, anger-fear, trust-
distrust, surprise-anticipation).” Some experts, explains Burton, believe
these basic emotions are hard-wired, “innate and universal, automatic, and
fast, and trigger behavior with a high survival value.” Secondary
emotions—like apprehension, interest, trust, for example—are
hypothesized to come from a combination of the primary emotions
The key word there is hypothesize. Human emotions are so complicated
that experts are continually studying, testing, disagreeing on and updating
these theories.
Dr. Thomas, on another hand, sees emotions on a more individual level—
instead of looking at a color-coded wheel of definite primary and
secondary emotions, Dr. Thomas says that primary emotions are simply
our initial reactions to external events or stimuli. Secondary emotions are
the reactions we then have to our reactions. For example, If you’re a
sensitive person, this might ring especially true for you—most emotionally
sensitive people tend to not only react to stimuli in their environments but
also to that aforementioned reaction. This leads to a chain of reactions that
can get more intense and last a longer time.
Are happiness, anger and sadness primary or secondary emotions?
That answer depends on which psychologist you talk to. If a professional
follows the Plutchik Model, the basic, primary emotions are joy, trust,
fear, surprise, sadness, anticipation, anger and disgust. Emotional scientist
Dr. Thomas, however, doesn’t believe emotions are experienced
exclusively as primary or secondary. Says Thomas: “One can actually
142
142
experience a full range of emotions as a primary response to our Sentiment Analysis and
environment.” Secondary emotions, Dr. Thomas argues, are responses to Opinion Mining - II
our primary emotions. Think: Your submission to a literary journal is

rejected. Your primary response is discouragement. Your
secondary response might be anxiety. Why? You've become anxious in
response to your own emotions.
Emotions are also arguably highly impacted by external factors, like
culture. In clinical psychologist Harriet Lerner’s The Dance of Anger: A
Woman's Guide To Changing The Patterns Of Intimate Relationships, the
author argues that the oh-so common anger women feel is not a primary
emotion but an alarm bell signaling a deeper issue. Kathy Caprino
interviewed Lerner in 2014 for Forbes asking, “Do you consider anger a
negative emotion, especially in our culture of positivity?” Lerner
responded: “Anger is neither positive nor negative. Anger simply is. It’s
an important emotion that deserves our attention and respect. But most of
us have little experience using our anger as a vehicle for positive change.
Instead we silence our anger or vent it in a way that leaves us feeling
helpless and powerless.”
How do primary vs. secondary emotions impact my relationships?
While we’re not going to tell you the frustration you’re feeling towards
your partner who left the bathroom floor sopping wet is a primary or
secondary emotion, it is clear that our emotions are complicated, and we
should honor them by taking a step back to really consider what it is we’re
feeling. If we can get to the crux of what we’re experiencing—and why
we’re experiencing it—we can make better decisions. It’s important to be
able to understand the reason why any feeling, from joy to deep sadness,
arises.
Here’s an example: Your wife broke your favorite dish and you’re
angry.
Are you angry that your wife would be so careless with something you
love or are you embarrassed at how you yelled at her when she dropped
the platter? This would feed into Dr. Thomas’s ideas of emotions.
Remember, Dr. Thomas says that when it comes to our relationships
(including the one with yourself!), the most important thing is to be able to
recognize whether you’re reacting to your internal stimuli or the
environment around you.
In the Plutchik Model, your anger surrounding the broken platter might be
considered a secondary reaction to the initial, primal terror you felt when
you heard it crash to the ground and instinctively thought, “danger!”
But Dance of Anger author, Harriet Lerner, has another postulation: Your
anger surrounding the broken dish might be trying to alert you to a more
important story about your relationship.
If you’re yelling at your wife about how you’ll never be able to find a
replacement, but really, you’re hurt that your wife is often careless with
things that are important to you, you won’t get the resolution you
143
Natural Language Processing need. Think of the big waves of emotions you experience as opportunity
for positive change—ride the waves safely to the shore instead of battling
against the current. Take a breath (try some square breathing) and ask
yourself, is this really a time to lambast your loved one or is this an
occasion to access what’s really going on with your relationship and it
address it head on? If it's the latter, perhaps you calmly suggest, "I'm hurt
you broke something important to me, but I need some time to process my
thoughts. Let's enjoy dinner and make time to talk about this tomorrow."
OK, so how do I interpret my emotions?
There are lots of ways to gain clarity on your feelings. Here are some of
our suggestions.
1. Seek professional help
Whether it’s a psychiatrist, psychologist, therapist, social worker or
other mental health professional, often times gaining insight into
your own emotions means bringing in some outside help. Seeing the
big picture can be impossible when you’re stuck inside the whirling
chaos of it all. But as soon as you bring in a third, neutral party
whose job is to help you understand yourself the truth can unravel
fairly quickly—but only if you’re open to it.
2. Journaling
You’d be surprised at how effective putting pen to paper can be. Try
spending ten minutes every day free writing. Jot down any thought
that comes to you for that set period of time without worrying about
making sense, and you’ll be amazed at what comes to the surface.
Who knew you were so moved by your roommate’s cooking? Or so
offended by the criticism your friend texted you? Not sure you can
commit to journaling just yet? Try a one-word journal (consider it
self-care for when you’re feeling a little lazy.)
3. Mindful Meditation
You’ve heard all the fuss about mindful meditation, but the hype is
legit. The 2,000-year-old Buddhist practice has been a proven way
to improve awareness around the present moment and self. While
meditation can seem lofty and unfamiliar, there are so many ways to
bring the ancient practice into your life. Start with a quick ten-
minute session with an app like Headspace or Calm. Not into sitting
still? Try mindful running!
4. Read
If one of the emotions you’re feeling often is anger, get your hands
on a copy of Harriet Lerner’s Dance of Anger ASAP. If talking or
thinking about your feelings is brand-new to you, try Marc
Brackett’s Permission to Feel: Unlocking the Power of Emotions to
Help Our Kids, Ourselves, and Our Society Thrive, which provides a
blueprint for acknowledging and skillfully understanding our
emotions in a society that shames emotions.
144
144
Effective Computing: Sentiment Analysis and
Opinion Mining - II
Scherer’s typology of affective states
145
10.1 Aspect based sentiment analysis
Aspect-Based Sentiment Analysis (ABSA) is a type of text analysis that

categorizes opinions by aspect and identifies the sentiment related to
each aspect. By aspects, we consider attributes or components of an entity
(a product or a service, in our case).
Traditional document-level sentiment classification tries to identify the
general sentiment polarity of a given text as positive, negative, or neutral.
Unlike document-level sentiment, aspect-level sentiment
classification identifies the sentiment of one specific aspect in its
context sentence.
What is aspect in NLP?
We determining the orientation of sentiment expressed on each aspect in a
sentence. There are two main approaches, i.e., the supervised learning
approach and the lexicon-based approach. For the supervised learning
approach, the learning-based methods used for sentence-level and clause-
level sentiment classification. In a hierarchical classification model was
also proposed. However, the key issue is how to determine the scope of
each sentiment expression, i.e., whether it covers the aspect of interest in
the sentence. The current main approach is to use parsing to determine the
dependency and the other relevant information. For example, a
dependency parser was used to generate a set of aspect dependent features
for classification. A related approach was also used which weights each
feature based on the position of the feature relative to the target aspect in
the parse tree. For comparative sentences, “than” or other related words
can be used to segment a sentence. Supervised learning is dependent on
the training data. As we discussed a model or classifier trained from
labeled data in one domain often performs poorly in another domain.
146
146
Although domain adaptation (or transfer learning) has been studied by Sentiment Analysis and
Opinion Mining - II
researchers, the technology is still far from mature, and the current
methods are also mainly used for document level sentiment classification
as documents are long and contain more features for classification than
individual sentences or clauses. Thus, supervised learning has difficulty to
scale up to a large number of application domains. Sentiment Analysis and
Opinion Mining 60 The lexicon-based approach can avoid some of the
issues , and has been shown to perform quite well in a large number of
domains. Such methods are typically unsupervised. They use a sentiment
lexicon (which contains a list of sentiment words, phrases, and idioms),
composite expressions, rules of opinions , and (possibly) the sentence
parse tree to determine the sentiment orientation on each aspect in a
sentence. They also consider sentiment shifters, but-clauses and many
other constructs which may affect sentiments.
Of course, the lexicon-based approach also has its own shortcomings,
which we will discuss later. An extension of this method to handling
comparative sentences . Below, we introduce one simple lexicon-based
method to give a flavor of this approach. The method and it has four
steps. Here, we assume that entities and aspects are known. Mark
sentiment words and phrases: For each sentence that contains one or more
aspects, this step marks all sentiment words and phrases in the sentence.
Each positive word is assigned the sentiment score of +1 and each
negative word is assigned the sentiment score of −1. For example, we have
the sentence, “The voice quality of this phone is not good, but the battery
life is long.” After this step, the sentence becomes “The voice quality of
this phone is not good [+1], but the battery life is long” because “good” is
a positive sentiment word (the aspects in the sentence are italicized). Note
that “long” here is not a sentiment word as it does not indicate a positive
or negative sentiment by itself in general, but we can infer its sentiment in
this context shortly. In fact, “long” can be regarded as a context-dependent
sentiment word, which we will discuss in Chapter 10. In the next section,
we will see some other expressions that can give or imply positive or
negative sentiments.
2. Apply sentiment shifters: Sentiment shifters (also called valence
shifters in (Polanyi and Zaenen, 2004)) are words and phrases that
can change sentiment orientations. There are several types of such
shifters. Negation words like not, never, none, nobody, nowhere,
neither, and cannot are the most common type. This step turns our
sentence into “The voice quality of this phone is not good[-1], but
the battery life is long” due to the negation word “not.” We will
discuss several other types of sentiment shifters in the next section.
Note that not every appearance of a sentiment shifter changes the
sentiment orientation, e.g., “not only … but also.” Such cases need
to be dealt with care. That is, such special uses and patterns need to
be identified beforehand.
3. Handle but-clauses: Words or phrases that indicate contrary need
special handling because they often change sentiment orientations
too. Sentiment Analysis and Opinion Mining 61 The most 147
Natural Language Processing commonly used contrary word in English is “but”. A sentence
containing a contrary word or phrase is handled by applying the
following rule: the sentiment orientations before the contrary word
(e.g., but) and after the contrary word are opposite to each other if
the opinion on one side cannot be determined. The if-condition in
the rule is used because contrary words and phrases do not always
indicate an opinion change, e.g., “Car-x is great, but Car-y is better.”
After this step, the above sentence is turned into “The voice quality
of this phone is not good[-1], but the battery life is long[+1]” due to
“but” ([+1] is added at the end of the but-clause). Notice here, we
can infer that “long” is positive for “battery life”. Apart from but,
phrases such as “with the exception of,” “except that,” and “except
for” also have the meaning of contrary and are handled in the same
way. As in the case of negation, not every but means contrary, e.g.,
“not only … but also.” Such non-but phrases containing “but” also
need to be identified beforehand.
4. Aggregate opinions: This step applies an opinion aggregation
function to the resulting sentiment scores to determine the final
orientation of the sentiment on each aspect in the sentence. Let the
sentence be s, which contains a set of aspects {a1, …, am} and a set
of sentiment words or phrases {sw1, …, swn} with their sentiment
scores obtained from steps 1- 3. The sentiment orientation for each
aspect ai in s is determined by the following aggregation function: , (
, ) . ( , ) ∑∈ = ow s j i j i j dist sw a sw so score a s (5) where swj is
an sentiment word/phrase in s, dist(swj, ai) is the distance between
aspect ai and sentiment word swj in s. swj.so is the sentiment score
of swi. The multiplicative inverse is used to give lower weights to
sentiment words that are far away from aspect ai. If the final score is
positive, then the opinion on aspect ai in s is positive. If the final
score is negative, then the sentiment on the aspect is negative. It is
neutral otherwise. This simple algorithm performs quite well in
many cases. It is able to handle the sentence “Apple is doing very
well in this bad economy” with no problem. Note that there are
many other opinion aggregation methods. For example, (Hu and Liu,
2004) simply summed up the sentiment scores of all sentiment
words in a sentence or sentence segment. Kim, and Hovy (2004)
used multiplication of sentiment scores of words. Similar methods
were also employed by other researchers (Wan, 2008; Zhu et al.,
2009). To make this method even more effective, we can determine
the scope of each individual sentiment word instead of using words
distance as above. In Sentiment Analysis and Opinion Mining 62
this case, parsing is needed to find the dependency as in the
supervised method discussed above. We can also automatically
discover the sentiment orientation of context dependent words such
as “long” above. In fact, the above simple approach can be enhanced
in many directions. For example, Blair-Goldensohn et al. (2008)
integrated the lexicon-based method with supervised learning.
Kessler and Nicolov (2009) experimented with four different
strategies of determining the sentiment on each aspect/target
148
148
(including a ranking method). They also showed several interesting Sentiment Analysis and
Opinion Mining - II
statistics on why it is so hard to link sentiment words to their targets
based on a large amount of manually annotated data. Along with
aspect sentiment classification research, researchers also studied the
aspect sentiment rating prediction problem which has mostly been
done together with aspect extraction in the context of topic
modeling, As indicated above, apart from sentiment words and
phrases, there are many other types of expressions that can convey
or imply sentiments. Most of them are also harder to handle. Below,
we list some of them, which are called the basic rules of opinions .
Aspect-based sentiment analysis goes one step further than sentiment
analysis by automatically assigning sentiments to specific features or
topics. It involves breaking down text data into smaller fragments,
allowing you to obtain more granular and accurate insights from your data.
Aspect-Based Opinion Mining ( ABOM ) involves extracting aspects or
features of an entity and figuring out opinions about those aspects. It's
a method of text classification that has evolved from sentiment analysis
and named entity extraction ( NER ). ABOM is thus a combination of
aspect extraction and opinion mining.
Which type of analytics is used in sentiment analysis?
Sentiment analysis is used to determine whether a given text contains
negative, positive, or neutral emotions. It's a form of text analytics that
uses natural language processing (NLP) and machine learning.
Sentiment analysis is also known as “opinion mining” or “emotion
artificial intelligence.
Issues and Challenges of Aspect-based Sentiment Analysis: A
Comprehensive Survey Ambreen Nazir, Yuan Rao, Lianwei Wu, Ling Sun
Abstract—The domain of Aspect-based Sentiment Analysis, in which
aspects are extracted, their sentiments are analyzed and sentiments are
evolved over time, is getting much attention with increasing feedback of
public and customers on social media. The immense advancements in the
field urged researchers to devise new techniques and approaches, each
sermonizing a different research analysis/question, that cope with
upcoming issues and complex scenarios of Aspect-based Sentiment
Analysis. Therefore, this survey emphasized on the issues and challenges
that are related to extraction of different aspects and their relevant
sentiments, relational mapping between aspects, interactions,
dependencies and contextual-semantic relationships between different data
objects for improved sentiment accuracy, and prediction of sentiment
evolution dynamicity. A rigorous overview of the recent progress is
summarized based on whether they contributed towards highlighting and
mitigating the issue of Aspect Extraction, Aspect Sentiment Analysis or
Sentiment Evolution.
The reported performance for each scrutinized study of Aspect Extraction
and Aspect Sentiment Analysis is also given, showing the quantitative
evaluation of the proposed approach. Future research directions are 149
Natural Language Processing proposed and discussed, by critically analysing the presented recent
solutions, that will be helpful for researchers and beneficial for improving
sentiment classification at aspect-level.
Index Terms— Aspect, Computational linguistic, Deep learning,
Sentiment analysis, Sentiment evolution, Social media —
INTRODUCTION HE field of Natural Language Processing (NLP)
dealing Sentiment Analysis (SA), also named as opinion mining, is an
active research area to display emotions and to automatically discover the
sentiments expressed within the text.
The object of SA is usually a product or service that is of keen interest
among people, that they care to put a sentiment towards it. Traditionally,
SA has been considered as opinion polarity that whether someone has
expressed positive, negative or neutral sentiment about an event .
Since last decade, researchers are putting efforts to capture, quantify and
measure dynamic public sentiments through different methods, tools and
techniques, and thus allowing SA as one of the rapidly growing research
areas.
SA applications have been widely spread to nearly every domain, like
social events, consumer products and services, healthcare, political
elections and financial services.
Influential groups and business organizations, like, Google, Microsoft,
SAP and SAS, have designed their own in-house capabilities that support
them in decision making and assist them in developing better business
applications to track and predict evolving market trends. From studies, SA
has been generally categorized at three levels.
Document-level
sentence-level
and aspect level
To classify whether a whole document, sentence (subjective or objective)
and an aspect expresses a sentiment, i.e., positive, negative or neutral. The
Aspect based Sentiment Analysis (AbSA) helps to understand the problem
of SA better comparatively, because it directly focuses on sentiments
rather than language structure. Where, an aspect is related to an entity, and
the basic concept of an aspect is not just limited to judgement but also
extends towards thoughts, point of views, ways of thinking, perspectives,
an underlying theme or social influence towards an occurrence.
Hence, AbSA provides a great opportunity to analyse sentiments (public)
over time across different contents present on media .
AbSA can be categorized by three main processing phases, i.e., Aspect
Extraction (AE), Aspect Sentiment Analysis (ASA) and Sentiment
Evolution (SE). The first phase deals with the extraction of aspects, which
can be
150
150
explicit aspects , Sentiment Analysis and
Opinion Mining - II
implicit aspects ,
aspect terms ,
entities and Opinion Target Expressions (OTE) .
The second phase classifies sentiment polarity for a predefined aspect,
target or entity .
This phase also formulates interactions, dependencies and contextual
semantic relationships between different data objects, e.g., aspect, entity,
target, multi-word target, sentiment word, for achieving improved
sentiment classification accuracy .
The expressed sentiment can be classified into ternary, or fine-gained
sentiment values .
The third phase concerns with the dynamicity of peoples’ sentiment
towards aspects (events) over a period of time. Social characteristics and
self-experience are considered as the leading causes for SE .
Focus of this Survey The field of AbSA is not a straight road, it has
suffered many diverse changes and touched many new eras to ponder
over. Researchers have been working hard to resolve multi-faceted
challenges containing many issues. They have come up with thorough
solutions of many complicated challenges through different machine
learning techniques, mostly deep-learning techniques, that represent their
critical idea in the field. They have provided pictorial representations and
numerical modeling for handling complex scenarios.
Organization of this Survey :The survey has been started with the brief
introduction and paramount significance of AbSA. The rest of the survey
is organized as follows: Section II provides the definitions of sentiment
with respect to aspect, and lists down the major issues and challenges
related to AE, ASA and SE. Section III and IV discusses the major issues
of AE and ASA, and concisely describes the recent solutions for these
issues. Section V discusses the dynamics of SE. Section VI highlights the
future research directions. Section VII concludes this survey.
KEY SOLUTIONS : This section presents AbSA definitions, and
outlines the major issues and sub-issues of AE, ASA and SE. This section
also illustrates three main processing steps of AbSA, i.e., AE, ASA and
SE, through framework for the ease and understandability of the reader.
Definitions (AbSA) In AbSA, sentiment is a subjective consciousness of
human beings towards an aspect (objective existence). People get affected
by the ups and downs of life, at any time and place, which results in the
change of their sentiment towards a specific aspect. This change depicts
human behavior flexibility, decision autonomy and creative thinking. Pang
and Lee defined SA as: “A sentiment is basically an opinion that a person
expresses towards an aspect, entity, person, event, feature, object, or a
certain target.” Now Title Critical Idea Challenges Pang . Opinion Mining
and SA ∙ presented applications that determine sentiment of the document,
and organized approaches related to opinion-oriented classification 151
Natural Language Processing problems How to perform SA at document-level using machine-learning
approaches?
A Survey on Sentiment Detection of Reviews ∙ discussed approaches
related to subjectivity classification, word classification and opinion
discovery in customer review domain How to perform SA at document-
level using machine-learning approaches?
A Survey on the Role of Negation in SA ∙ presented computational
methods for handling negation in SA How to cope with the scope and
challenges of negation modeling?
Survey on Mining Subjective Data on The Web ∙ differentiated between
four different approaches that classify word-sentiment value, i.e.,
machine-learning, semantic, statistical and dictionary-based approaches
How to perform subjective SA at documentlevel?
SA and Opinion Mining ∙ covered the field of SA at document, sentence
and aspect-level ∙ discussed various issues related to AE, sentiment
classification, sentiment lexicons, NLP and opinion-spam detection ∙
surveyed the till date practical solutions along with the future directions
How to cope with review ranking, redundancy issues, viewpoints quality,
genuine aspects, spammer detection etc… ?
A Survey on Opinion Mining and SA: Tasks, Approaches and
Applications ∙ organized subtasks of machine learning, NLP and SA
techniques, such as, subjectivity classification, sentiment classification,
lexicon relation, opinion-word extraction, and various applications of SA ∙
discussed open issues and future directions in SA How to focus on
sentence-level and document-level SA and their subtasks?
Like It or Not: A Survey of Twitter SA Methods ∙ discussed the deep-
learning algorithms related to twitter SA ∙ elaborated tasks specific to
emotion detection, change of sentiment over time, sarcasm detection, and
sentiment classification How to tackle the challenges, tasks and feature
selection methods limited to twitter SA?
Survey on Aspect-Level SA ∙ performed approach-based categorization
of different solutions those were related to AE, aspect classification and a
combination of both ∙ proposed future research direction for semantically-
rich-concept-centric AbSA How to cope with the challenges of
comparative opinions, conditional sentences, negation modifiers and
presentation?
Deep Learning for SA: A Survey ∙ presented applications and deep-
learning approaches for the SA related tasks such as sentiment
intersubjectivity, lexicon expansion, stance detection How to achieve
advances in SA using deep learning approaches?
The State-of-the-Art in Twitter SA: A Review and Benchmark
Evaluation ∙ focused on challenges and key trends related to classification
errors, twitter monitoring and event detection to perform twitter SA
152
152
effectively How to reveal the root causes of commonly occurring Sentiment Analysis and
Opinion Mining - II
classification errors?
A Survey of SA in Social Media ∙ categorized the latest technologies and
techniques in SA ∙ introduced latest tools for research approaches related
to subjectivity classification in customer-review domain How to focus on
different types of data and advance tools, to overcome the limitations of
social media SA?
A Survey on Sentiment and Emotion Analysis for Computational
Literary Studies ∙ presented approaches of SA and emotion analysis ∙
discussed the computational methods for sentiment and emotion
classification How to classify and interpret the emotional text through
sentiment and emotional analysis in digital human community? Our Issues
and Challenges of Aspect-based Sentiment Analysis: A Comprehensive
Survey ∙ discusses the issues and challenges of AE, ASA and SE ∙ presents
the progress of AbSA by concisely describing the recent solutions ∙
highlight factors responsible for SE dynamicity ∙ proposes future research
directions by critically analyzing the present solutions How to improve the
mechanism of AE? What measures should be taken to achieve good
classification accuracy at aspect-level? How to predict SE dynamicity?
Reference pdf:
• https://www.cs.uic.edu/~liub/FBS/SentimentAnalysis-and-
OpinionMining.pdf
• https://www.pdfdrive.com/sentiment-analysis-and-opinion-mining-
e1590362.html
• https://www.researchgate.net/publication/265163299_Sentiment_An
alysis_and_Opinion_Mining_A_Survey
• https://link.springer.com/referenceworkentry/10.1007/978-1-4899-
7687-1_907
• https://www.morganclaypool.com/doi/abs/10.2200/S00416ED1V01
Y201204HLT016

153

Natural Language Processing Inside Pages 2

Uploaded by

Copyright:

Available Formats

Natural Language Processing Inside Pages 2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Natural Language Processing Inside Pages 2

Uploaded by

Copyright:

Available Formats

F.Y.

Dr. Suhas Pednekar

Programme Co-ordinator : Shri Mandar Bhanushe

Course Co-ordinator : Shyam Mohan T.

Course Writers : Mrs. Vidya Bharde

: Mr. Krishna Kumar Sharma

: Mrs. Hema Darne

May 2022, Print - 1

DTP composed and Printed by:

Chapter No. Title Page No.

2. Word level analysis - I.............................................................................................22

3. Word level analysis - II...........................................................................................34

4. Syntax Analysis _ I..................................................................................................56

5. Syntax Analysis - II.................................................................................................69

6. Semantic Analysis - I..............................................................................................82

7. Semantic Analysis - II.............................................................................................95

8. Text Summarization, Text Classification...............................................................110

9. Sentiment Analysis and Opinion Mining - I.........................................................125

10. Sentiment Analysis and Opinion Mining - II........................................................139

NATURAL LANGUAGE PROCESSING

Module No Detailed Contents Hrs

The objective of this module is to learn about history of Natural Language

1.1 HISTORY OF NLP

As we all know Natural Language refers to the language spoken by people

1.2 GENERIC NLP SYSTEM

Figure 1.2.1 Generic NLP system.

Figure 1.2.2 Pipeline view of the components of a generic NLP system.

There are two components of Natural Language Processing

It is the process of reading and It is the process of writing or generating

NLU understands the human NLG uses the structured data

Figure 1.3.1 Levels of NLP

A natural language understanding system must have detailed information

Pragmatics is the extension of the meaning or semantics. Pragmatics

1.5 AMBIGUITY IN NATURAL LANGUAGE

Ambiguity can occur at all NLP levels. It is a property of linguistic

• She made a silver speech.

• The car, while moving, hit the pole

• The car the pole while the pole was moving.

• Anaphoric Ambiguity: Anaphora's are the entities that have been

• I love you (as well as liking you)

1.6 STAGES IN NLP

There are general five steps in natural language processing

Figure 1.6.1 General five steps of NLP

Figure 1.6.2 Stages of NLP

Syntactic analysis must exploit the results of morphological analysis to

1.7 CHALLENGES IN NLP

There are a number of challenges of natural language processing because

1.8 APPLICATIOS OF NLP

In this chapter we studied what is natural language processing (NLP),

1.10 MULTIPLE CHOICE QUSETION ANSWERS

i. What is the field of Natural Language Processing (NLP)?

i. NLP is concerned with the interactions between computers and

1.12 SAMPLE QUESTIONS

1. What is NLP? Discuss the different stages in NLP

The objective of this module is to learn about morphology, morphology

Affixes are further divided into prefixes, suffixes, infixes, and

• prefixes precede the stem,

Regular Nouns Irregular Nouns

Plural cats thrushes mice oxen

The possessive suffix is realized by apostrophe + -s for regular singular