Vol 11, Issue 4 , April/ 2020

ISSN NO: 0377-9254

Text Summarization using Natural Language Processing

Syed Muqtadir Uddin Hussaini1, Faraaz Mohd Khan2, Faisal Khan3, Dr. Abdul Subhan4

UG Scholar1, 2, 3, Asst. Professor4,

Department of Computer Science & Engineering1, 2, 3, 4,
Isl Engineering College, Hyderabad, India.

Abstract summarization is the process of automatically

generating natural language summaries from an input
In this project, Automatic text summarization is document while retaining the important points. It
basically summarizing of the given paragraph using would help in easy and fast retrieval of information.
natural language processing and machine learning. There are two prominent types of summarization
There has been an explosion in the amount of text algorithms.
data from a variety of sources. This volume of text is
an invaluable source of information and knowledge • Extractive summarization systems form summaries
which needs to be effectively summarized to be by copying parts of the source text through some
useful. In this review, the main approaches to measure of importance and then combine those
automatic text summarization are described. We part/sentences together to render a summary.
review the different processes for summarization and Importance of sentence is based on linguistic and
describe the effectiveness and shortcomings of the statistical features.
different methods. The system works by assigning
scores to sentences in the document to be • Abstractive summarization systems generate new
summarized, and using the highest scoring sentences phrases, possibly rephrasing or using words that were
in the summary. Score values are based on features not in the original text. Naturally abstractive
extracted from the sentence. A linear combination of approaches are harder. For perfect abstractive
feature scores is used. Almost all of the mappings summary, the model has to first truly understand the
from feature to score and the coefficient values in the document and then try to express that understanding
linear combination are derived from a training in short possibly using new words and phrases. Much
corpus. Some anaphor resolution is performed. The harder than extractive. Has complex capabilities like
system was submitted to the Document generalization, paraphrasing and incorporating real-
Understanding Conference for evaluation. In addition world knowledge. Majority of the work has
to basic summarization, some attempt is made to traditionally focused on extractive approaches due to
address the issue of targeting the text at the user. The the easy of defining hard-coded rules to select
intended user is considered to have little background important sentences than generate new ones. Also, it
knowledge or reading ability. The system helps by promises grammatically correct and coherent
simplifying the individual words used in the summary. But they often don’t summarize long and
summary and by drawing the pre-requisite complex texts well as they are very restrictive.
background information from the web.
Potential applications
Introduction
Possible current uses of summarization:
In the modern Internet age, textual data is ever
1. People need to learn much from texts. But they
increasing. Need some way to condense this data
tend to want to spend less time while doing
while preserving the information and meaning. We
this.
need to summarize textual data for that. Text

www.jespublication.com Page No:10

 Vol 11, Issue 4 , April/ 2020
ISSN NO: 0377-9254

2. It aims to solve this problem by supplying or not. For example, Search engines typically use
them the summaries of the text from which Extractive summary generation methods to generate
they want to gain information. summaries from web page. Many types of logical and
3. Goals of this project are that these summaries mathematical formulations have been used to create
will be as important as possible in the aspect summary. The regions are scored and the words
of the texts’ intention. containing highest score are taken into the
4. The user will be eligible to select the summary consideration. In extraction only important sentences
length. are selected. This approach is easier to implement.
5. Supplying the user, a smooth and clear There are three main obstacles for extractive
interface. approach. The first thing is ranking problem which
6. Configuring a fast replying server system. includes ranking of the word. The second one
selection problem that includes the selection of subset
Objectives of particular units of ranks and the third one is
coherence that is to know to select various units from
The objective of the project is to understand the understandable summary. There are many algorithms
concepts of natural language processing and creating which are used to solve ranking problem. The two
a tool for text summarization. The concern in obstacles i.e. - selection and coherence are further
automatic summarization is increasing broadly so the solved to improve diversity and helps in minimizing
manual work is removed. The project concentrates the redundancy and pickup the lines which are
creating a tool which automatically summarizes the important. Each sentence is scored and arranged in
document. decreasing order according to the score. It is not
trivial problem which helps in selecting the subsets of
Scope
sentences for coherent summary. It helps in reduction
The project is wide in scope | all of the limitations of redundancy. When the list is put in ordered
stated below may seem to contradict that, but they are manner than the first sentence is the most important
the only restrictions applied. This project looks at sentence which helps in forming the summary. The
single document summarization - the area of multi sentence having the highest similarity is selected in
document summarization is not covered. Also, the next step is picked from the top half of the list. The
summaries produced are largely extracts of the process has to be repeated until the limit is reached
document being summarized, rather than newly and relevant summary is generated.
generated abstracts. The parameters used are optimal
People by and large utilize abstractive outlines. In the
for news articles, although that can be changed
wake of perusing content, Individuals comprehend
easily.
the point and compose a short outline in their own
Methodologies particular manner creating their very own sentences
without losing any essential data. In any case, it is
For obtaining automatic text summarization, there are troublesome for machine to make abstractive
basically two major techniques i.e.-Abstraction based synopses. Along these lines, it very well may be said
Text Summarization and Extraction based Text that the objective of reflection based outline is to
Summarization. make a synopsis utilizing regular dialect preparing
procedure which is utilized to make new sentences
Extraction Based Extraction that are syntactically right. Abstractive rundown age
is difficult than extractive technique as it needs a
The Extractive summaries are used to
semantic comprehension of the content to be
highlight the words which are relevant, from input
encouraged into the Common Dialect framework.
source document. Summaries help in generating
Sentence Combination being the significant issue
concatenated sentences taken as per the appearance.
here offers ascend to irregularity in the produced
Decision is made based on every sentence if that
outline, as it's anything but an all around created field
particular sentence will be included in the summary
yet. Abstractive arrangement to grouping models is

www.jespublication.com Page No:11

 Vol 11, Issue 4 , April/ 2020
ISSN NO: 0377-9254

by and large prepared on titles and captions. The Results:

comparative methodology is embraced with archive
setting which helps in scaling. Further every one of
the sentences is revamped in the request amid the
inference. Document synopsis can be changed over to
regulated or semi-administered learning issue. In
directed learning methodologies, indications or signs,
for example, key-phrases, point words, boycott
words, are utilized to recognize the sentences as
positive or negative classes or the sentences are
physically labelled. At that point the parallel more
tasteful can be prepared for getting the scores or
synopsis of each sentence. Anyway they are not
effective in removing archive explicit summaries. If
the report level data isn't given then these
methodologies give same expectation independent of
the record. Giving archive setting in the models
diminishes this issue.

System Architecture :

www.jespublication.com Page No:12

 Vol 11, Issue 4 , April/ 2020
ISSN NO: 0377-9254

Summarized Text text summarizers with great capabilities and giving

good results. We have learned all the basics of
Extractive and Abstractive Method of automatic text
summarization and tried to implement extractive one.
We have made a basic automatic text summarizer
using nltk library using python and it is working on
small documents. We have used extractive approach
to do text summarization.

References

[1] Ahmad T. Al-Taani. “,Automatic Text

Summarization Approaches” International
Conference on Infocom Technologies and Unmanned
Systems (ICTUS'2017)

[2] Neelima Bhatia, ArunimaJaiswal, “Automatic

Text Summarization: Single and Multiple
Collected Information as paragraph so, total 68 Summarizations ”, International Journal of Computer
paragraphs have been collected from URL web page Applications
and extracted the main content as summary.
[3] Mehdi Allahyari, SeyedaminPouriyeh, Mehdi
Assefi, SaeidSafaei, Elizabeth D. Trippe, Juan B.
Gutierrez, KrysKochut, “ Text Summarization
Techniques: A Brief Survey”, (IJACSA)
International Journal of Advanced Computer Science
and Applications

Conclusion [4]Pankaj Gupta, Ritu Tiwari and

NirmalRobert,”Sentiment Analysis and Text
Text summarization is one of the major problems in Summarization of Online Reviews: A
the field of Natural Language Processing. Methods Survey”InternationalConzatiference on
such as Deep Understanding, Sentence Extraction, Communication and Signal Processing,August 2013
Paragraph Extraction, Machine Learning, and even
some which employ all these methods along with [5]Vishalgupta,Gurpreet Singh Lehal,”A Survey of
Traditional NLP Techniques(Semantic Analysis, Text Summarization Extractive
etc.). As such, keeping these accomplishments in Techniques.”JOURNAL OF EMERGING
mind, there is still ample amount of research left in TECHNOLOGIES IN WEB INTELLIGENCE,
the domain of Text Summarization, as a meaningful VOL. 2, NO. 3, AUGUST 2010
summary is still difficult to attain in all domains and
languages. [6]Jiwei Tan, XiaojunWan,Jianguo Xiao Institute of
Computer Science and Technology,Peking University
As with time internet is growing at a very fast rate “Abstractive document summarization with a
and with it data and information is also increasing. it GraphBased attentional neural model. ”
will going to be difficult for human to summarize
large amount of data. Thus there is a need of [7]SeonggiRyang, Graduate school of Information
automatic text summarization because of this huge science and technology, University of Tokyo Takeshi
amount of data. Until now, we have read multiple Abekawa, National institute of informatics
papers regarding text summarization, natural “Framework of automatic text summarization using
language processing. There are multiple automatic Reinforcement learning” 48

www.jespublication.com Page No:13

 Vol 11, Issue 4 , April/ 2020
ISSN NO: 0377-9254

[8]Tianshi, YaserKeneshloo, Narenramakrishnan,

Chandan K. Reddy, Senior member, IEEE “ Neural
Abstractive text summarization with sequence-to -
sequence models”

[9] Josef Steinberger, KarelJežek, “Using latent

Semantic analysis In Text Summarization and
Summary Evaluation”, Department of Computer
Science and Engineering, UniverzitníCZ-306 14
Plzeň.

[10] Sumitha C., Dr. A. Jaya, Amal Ganesh, “A study

on Abstract Summarization Techniques in Indian
Languages”, Elsevier Proceeding of Computer
Science, No. 87, pp.25-31, 2016.

[11] Dipanjan Das, Andre F.T. Martins, “A Survey

on Automatic Text Summarization”,Language
Technologies Institute, Carnegie Mellon University,
November 2007.

www.jespublication.com Page No:14