Page 1

CCS369 - TEXT AND SPEECH ANALYSIS

2 MARKS
UNIT I - NATURAL LANGUAGE BASICS

1.What is Natural Language Processing?

Natural Language Processing (NLP) is a fascinating field that sits at the intersection of computer
science, artificial intelligence, and linguistics. The goal of NLP is to enable computers to
understand, interpret, and generate human language in a way that is both meaningful and useful.

2.What are the applications of NLP?

• Parsing: Analyzing sentences to determine their syntactic structure, which can help in
understanding and generating human language.
• Machine Translation: Translating text from one language to another while maintaining
grammatical correctness.
• Information Extraction: Identifying and extracting relevant information from
unstructured text based on its syntactic structure.
• Text Generation: Creating coherent and grammatically correct sentences in various
applications such as chatbots and content creation tools.

3. Define Text preprocessing?

Text preprocessing and wrangling are crucial steps in natural language processing (NLP) and data
analysis. They involve cleaning and transforming raw text data into a format that is suitable for
analysis and model training.

4. What is Text Normalization?

Text Normalization: Converting text to a standard format to reduce variability.
• Lowercasing: Convert all text to lowercase to ensure uniformity.
text = "Hello World"
aa = text.lower()
print(aa)

# Output: "hello world"

5. What is Word Tokenization?

o Word Tokenization: Break a sentence into individual words.

import nltk
nltk.download('punkt')
Output - True
from nltk.tokenize import word_tokenize
text = "Natural Language Processing is fun!"
tokens = word_tokenize(text)
print(tokens)

# Output: ['Natural', 'Language', 'Processing', 'is', 'fun', '!']

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 2

6. Define Text wrangling?

Text wrangling involves further manipulation of the text data to prepare it for analysis or machine
learning tasks.

7.What is Feature Engineering?

Feature engineering for text representation involves converting raw text data into a format suitable
for machine learning models. This process transforms text into numerical features that can be used
to train and evaluate algorithms.

8. What is Term Frequency-Inverse Document Frequency ?

TF-IDF adjusts the word counts by considering the importance of a word in a document relative
to its frequency in the entire corpus. It helps in emphasizing important words and de-emphasizing
common words

9.What is word embeddings?

Word embeddings map words to dense vectors of fixed size, capturing semantic meanings and
relationships. Popular embeddings include Word2Vec, GloVe, and fastText.
10. What are the Application of GloVe Embeddings ?
• Text Classification
• Named Entity Recognition (NER)
• Sentiment Analysis
• Machine Translation
•
11. What is Bag of N-Grams?
The Bag of N-Grams model is an extension of the traditional Bag of Words (BoW) model used
in natural language processing (NLP). While the Bag of Words model considers each word
(unigram) as a separate feature, the Bag of N-Grams model considers sequences of words (n-
grams) as features.

12. What are the applications of Bag of N-Grams

• Sentiment Analysis: Understanding phrases like "not good" versus "good".
• Text Classification: Differentiating between similar words used in different contexts.
• Information Retrieval: Improving the relevance of search results by considering multi-
word expressions.

13. Define stemming

Stemming is the process of reducing a word to its base or root form. It involves trimming or
stripping affixes (such as suffixes or prefixes) from words to simplify them into their stem form.
from nltk.stem import PorterStemmer
stemmer = PorterStemmer()
words = ["running", "eats", "quickly", "happier"]

for word in words:

print(f"{word} -> {stemmer.stem(word)}")

output
running -> run

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 3

eats -> eat

quickly -> quick
happier -> happi

14. What is stopword removal?

Stopword removal is the process of filtering out common words from text that usually don't carry
significant meaning or contribute much to the task at hand (such as search engines, text analysis,
and natural language processing).

from nltk.corpus import stopwords

from nltk.tokenize import word_tokenize
import nltk
nltk.download('punkt')
nltk.download('stopwords')
text = "This is a sample sentence."
words = word_tokenize(text)
filtered_words = [word for word in words if word.lower() not in stopwords.words('english')]
print(filtered_words)

Output
['sample', 'sentence', '.']

15. What is Text Tokenization?

Splitting text into smaller units, such as words or sentences.

16. What is Lemmatization?

Lemmatization is the process of reducing a word to its base or dictionary form (known as the
lemma), which is different from stemming. Unlike stemming, which simply cuts off word endings,
lemmatization considers the context and returns valid words.

For example:

• Stemming: "running" → "run"

• Lemmatization: "better" → "good"

17. Define Bag of Words?

Bag of Words (BoW) is a simple and commonly used model in natural language processing (NLP)
for representing text data. The main idea behind BoW is to treat text (like a document, sentence,
or paragraph) as a collection of individual words, disregarding the order and context in which they
appear.

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 4

UNIT 2 - TEXT CLASSIFICATION

1.What is Vector semantics?
Vector semantics refers to the representation of linguistic items (such as words or sentences) as
vectors. The idea is that similar words or phrases should be close to each other in this vector space,
reflecting their semantic similarity.

2.What is embeddings?
Embeddings are dense, low-dimensional vector representations of words, phrases, or documents,
learned from large text corpora. They are designed to capture the semantic relationships between
words, where words with similar meanings are represented by similar vectors.

3.Define BERT?
BERT (Bidirectional Encoder Representations from Transformers): BERT generates
contextual embeddings, meaning that the representation of a word changes depending on the
words around it. It is based on the Transformer architecture and has achieved state-of-the-art
results in many NLP tasks.

4.What is Word2Vec?
Word2Vec is a popular technique in Natural Language Processing (NLP) for representing words
as vectors. Word2Vec models are used to capture the semantic relationships between words by
representing them in a continuous vector space, where words with similar meanings are placed
closer together.

5. What is Glove?
GloVe, or Global Vectors for Word Representation, is an unsupervised learning algorithm
developed by researchers at Stanford for obtaining vector representations of words. The idea
behind GloVe is to leverage the global statistical information of a corpus to produce dense word
embeddings.

6. What are the applications of Glove Embeddings?

• Text Classification
• Named Entity Recognition (NER)
• Sentiment Analysis
• Machine Translation

7.What is FastText?
FastText is an extension of the Word2Vec model developed by Facebook's AI Research (FAIR)
lab. It addresses some of the limitations of Word2Vec, particularly in handling out-of-vocabulary
(OOV) words and capturing subword information.

8.What are the applications of FastText?

• Handles OOV Words: By utilizing subword information, FastText can create embeddings
for words that were not present in the training data.
• Better for Morphologically Rich Languages: Languages with complex word forms (e.g.,
Finnish, Turkish) benefit significantly from FastText since it captures morphemes.
• Flexibility: FastText can be used in a wide range of applications, from text classification
to machine translation, and even for generating embeddings for unseen words in pre-
trained models.

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 5

9.What is Deep Learning?

Deep learning is a subset of machine learning that focuses on algorithms inspired by the structure
and function of the brain, known as artificial neural networks. It is particularly powerful for tasks
involving large amounts of data, such as image recognition, natural language processing, and
speech recognition.
10. What is Self-supervised learning?
Self-supervised learning is an approach where the model learns from unlabeled data by predicting
parts of the input from other parts. It's used in pre-training large models on vast datasets.
11.What is RNN?
Recurrent Neural Networks (RNNs) are a type of neural network designed for sequential data,
where the order of the data points is crucial. They are particularly effective for tasks where the
context from previous inputs in a sequence influences the current output. RNNs are widely used
in natural language processing (NLP), time series prediction, and other tasks involving sequential
data.
12. What is Transformers?
Transformers are a type of deep learning model architecture that has revolutionized the field of
natural language processing (NLP) and, more recently, has been applied to a variety of other tasks,
including computer vision, time series analysis, and more
13. What is Text summarization?
Text Summarization is a process of distilling the most important information from a source text
and presenting it in a shorter form while maintaining the overall meaning and key points. It's a
crucial tool in managing large amounts of textual data, helping users quickly grasp the main ideas
without having to read the entire content.
14. What is Topic Modeling?
Topic Modeling is an unsupervised machine learning technique used to discover the hidden
thematic structure within a large collection of documents. It identifies patterns of word co-
occurrence in the text and groups words into topics that capture the main themes across the
documents.
15. What are the applications of Topic Modeling?
1. Content Recommendation: By identifying topics within a user's reading history, content
recommendation systems can suggest articles or books that align with their interests.
2. Document Classification: Topic modeling helps in categorizing documents based on
their main themes, such as clustering news articles by topic.
3. Text Mining: Researchers and analysts use topic modeling to explore large text corpora
and identify key themes, trends, or hidden insights.

UNIT III
QUESTION ANSWERING AND DIALOGUE SYSTEMS
1.What is Information Retrieval?
Information Retrieval (IR) is the process of obtaining relevant information from a large
repository, often in response to a user query. The primary goal of IR is to help users find
the information they need quickly and efficiently. This is commonly applied to document
collections, such as web pages, research articles, or databases, where a system retrieves
documents based on their relevance to the user's search terms.

2.What is IR based QA?

IR-based QA is a system that retrieves relevant documents or passages from a large corpus of text
(often unstructured) to answer a user's question. Unlike systems that generate answers directly,

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 6

IR-based QA systems first find relevant information from a database or set of documents and then
extract the most relevant part to form an answer.

3.What is Retrieval Augmented Generation?

Retrieval-Augmented Generation (RAG): Combines retrieval-based methods with generative
models. The system retrieves documents using traditional IR methods but then refines the results
using a generative model.

4.What is Knowledge Based QA?

Knowledge-based question answering (KB-QA) refers to systems that answer questions by

leveraging structured knowledge bases (KBs) like Wikidata, Freebase, or DBpedia. These systems
differ from general-purpose QA systems in that they query explicit facts stored in databases rather
than processing unstructured text.

5.What is Language Model QA?

Language models for question answering (QA) have advanced significantly in recent years,
particularly with the advent of transformer-based architectures like OpenAI’s GPT, Google’s
BERT, and other similar models. These models use vast amounts of text data to learn linguistic
patterns, enabling them to generate or retrieve accurate answers to questions without relying on
explicit knowledge bases.

6. What is Classic QA Model?

Classic QA models, developed before the deep learning and transformer revolutions, relied more
on structured approaches, traditional machine learning, and rule-based systems. These models
typically focused on understanding and retrieving answers from specific types of data, such as
documents, structured databases, or even human-curated knowledge bases.

7.What is Rule based QA?

Rule-based QA systems were among the earliest attempts at automating question answering. These
systems followed manually defined rules or templates to process the question and retrieve the
answer. They were effective for specific, narrow domains but lacked flexibility.

8.What is pipeline based QA systems?

Pipeline-based QA systems break down the QA process into a sequence of independent steps, each
responsible for a specific task such as parsing, entity extraction, relation identification, and answer
generation.

9. What is Statistical Machine Learning?

Statistical machine learning models improved upon rule-based and IR-based systems by learning
patterns from data, often using features derived from questions and text. Classic machine learning
algorithms such as support vector machines (SVMs) and decision trees were applied to QA tasks.

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 7

10. What is Template based QA?

Template-based QA systems relied on predefined templates that represented different types of

questions. A question would be matched to one of the templates, and the corresponding structured
query would be used to retrieve the answer.

11. Define Chatbots?

Chatbots are designed for interactive communication, often implemented in customer service or
personal assistants like Siri and Alexa. While not purely QA systems, they often include QA
components.

12. What are the types of chatbots?

Rule-based chatbots: Simple bots that follow predefined scripts.

ML-based chatbots: More advanced, with natural language understanding (NLU) and dialogue
management systems.

13.What are the applications of Chatbots?

1. Customer Support Chatbots are widely used in customer support for handling FAQs,
troubleshooting issues, and guiding users through product features or services. They reduce
the workload for human agents and provide instant assistance.
2. Virtual Assistants AI chatbots like Siri, Google Assistant, and Alexa act as personal
assistants, helping users perform tasks like setting reminders, controlling smart home
devices, and answering questions.
3. E-commerce In e-commerce, chatbots help customers find products, process orders, and
provide information about discounts or promotions. They can guide users through the
shopping process or offer product recommendations.
4. Healthcare Healthcare chatbots assist patients with appointment scheduling, providing
medical information, reminding users to take medications, and even performing symptom
checking.

14. What is Dialogue systems?

Designing dialogue systems, especially conversational agents like chatbots and virtual assistants,
requires a careful balance of linguistic understanding, interaction flow, and backend integration.
Dialogue systems are typically composed of multiple components that allow them to engage users
in natural, coherent, and goal-oriented conversations.

15.What is Hybrid Systems?

Combining rule-based components with machine learning models creates hybrid systems. For
example, NLU and dialogue management may be rule-based for task-oriented conversations,
while response generation is handled by an AI model for more natural interaction

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 8

16.What is the difference between IR and Knowledge based Question Answering?

UNIT IV
TEXT-TO-SPEECH SYNTHESIS
1.What is Text Normalization?
Text normalization is the process of transforming text into a standard format to facilitate easier
processing and analysis, especially in natural language processing (NLP) tasks. It involves several
steps that help to reduce the variability in text data.

2.What is letter to sound?

Letter-to-sound (L2S), also known as grapheme-to-phoneme (G2P) conversion, is the process
of converting written text (letters or graphemes) into their corresponding sounds (phonemes). This
process is a fundamental part of speech synthesis (text-to-speech systems) and pronunciation
generation in applications like automatic speech recognition (ASR) and text-to-speech (TTS)
systems.
3.What are the applications of Letter to sound

• Text-to-Speech (TTS): Converts written text to speech by first converting letters into
phonemes.
• Automatic Speech Recognition (ASR): Uses phoneme models for recognizing speech
and mapping spoken words to text.
• Language Learning Tools: Helps learners by generating phonetic transcriptions of words.

4.Define Prosody?
Prosody refers to the rhythm, intonation, and stress patterns in speech that convey meaning,
emotion, and structure. It's an essential aspect of natural language and spoken communication,
affecting how messages are perceived beyond the basic phonetic sounds.

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 9

5.What is Signal Processing

Signal processing is the analysis, manipulation, and interpretation of signals to extract useful
information, enhance their quality, or convert them into a desired format. Signals can be anything
that conveys information, such as sound, images, sensor readings, or data streams, and they can
be represented in various forms like analog (continuous) or digital (discrete).

6.What are the types of signal processing

1. Analog Signals: Continuous signals, like sound waves or light, that vary over time and
take any value in a given range.
o Example: Human speech captured by a microphone.
2. Digital Signals: Discrete-time signals, often derived from the sampling of analog signals,
represented as sequences of numbers (binary).
o Example: A digitally recorded audio file.

7.What are the applications of signal processing

Audio and Speech Processing

Image and Video Processing

Biomedical Signal Processing

Communication Systems

8. What is Concatenative Speech Synthesis?

Concatenative TTS systems generate speech by concatenating pre-recorded speech segments.

These systems rely on a large database of recorded speech, typically broken into smaller units like
phonemes, diphones, syllables, or words.

9. What is parametric speech synthesis?

Parametric TTS systems generate speech by modeling the speech production process. Instead of
concatenating pre-recorded speech, parametric approaches synthesize speech by using statistical
models to control parameters like pitch, duration, and formants (vocal tract resonances) to generate
audio waveforms from scratch.

10.What is Deep learning-based text-to-speech (TTS) systems?

Deep learning-based text-to-speech (TTS) systems, particularly those like WaveNet, represent a
major leap in generating natural and high-quality synthetic speech. These systems address many
limitations of traditional methods like concatenative and parametric TTS by using neural
networks to learn the complex patterns of human speech directly from data.

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 10

UNIT V
AUTOMATIC SPEECH RECOGNITION
1.What is Acoustic Modelling?

Acoustic modeling is a crucial component of speech recognition systems, where it deals with the
representation of the relationship between linguistic units of speech (such as phonemes or words)
and the corresponding audio signal. It focuses on how to statistically model the way phonetic units
are produced in various contexts, including differences in speakers, accents, and environmental
noise.

2.What is Phonemes?

Phonemes are the smallest units of sound in a language, and acoustic models attempt to recognize
these by mapping the audio signal to the corresponding phonetic sounds. For example, the words
"cat" and "bat" differ by just one phoneme: /k/ and /b/.

3.What is Gaussian Mixture Models?

GMMs are used to model the distribution of the acoustic features associated with each HMM state.
A GMM is a weighted sum of several Gaussian distributions and helps capture the variability in
speech signals for a particular phoneme.

4. What are the applications of Acoustic Modelling?

• Voice Assistants (e.g., Siri, Alexa, Google Assistant)

• Speech-to-Text systems
• Automatic captioning

5.What is Hidden Markov Model?

A Hidden Markov Model (HMM) is a statistical model that is widely used in speech recognition,
natural language processing, and various other time-series applications. It is particularly well-
suited for modeling sequences where observations are generated by underlying hidden states,
which evolve over time.

6. What are the Advantages of HMM-DNN Systems?

Improved Acoustic Modeling

Ability to Model Large Feature Spaces

Discriminative Training

Better Generalization

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE

 Page 11

7.What are the training process of Acoustic Modelling?

Labeling

Supervised learning

Evaluation

8. What is Acoustic Model Adaptation?

Speech recognition systems often need to adapt to new speakers, environments, or languages.
Techniques like Maximum Likelihood Linear Regression (MLLR) or speaker adaptation
training can be used to fine-tune acoustic models for specific speakers or conditions.

9.What is Feature extraction?

Feature extraction is a fundamental step in speech processing, particularly in speech recognition

systems. The goal of feature extraction is to transform the raw audio signal into a set of concise,
informative representations (features) that retain the essential characteristics of speech and discard
irrelevant information, such as background noise.

10. What are the steps in Feature extraction?

Step 1: Pre-Emphasis
Step 2: Framing
Step 3: Windowing
Step 4: Fourier Transform

CCS369 – TASA EINSTEIN COLLEGE OF ENGINEERING Prepared by K.M.Annammal,AP/CSE