2. Introduction
What is Text Summarization?
Making short, accurate, fluent summary of major point of given text or text documents.
Documents Summary
3. • Why text summarization?
Reduce reading time
Find key and main point
Effectiveness of indexing
Automatic summarization are less biased with human summarization
• Examples of Text Summarization
Headlines
Outlines
Previews
Reviews
History
Biography
4. Types Of Text Summarization
Extractive Abstractive
Summarization
System
একজন ছাত্ররেে পডার ানা কো অরনক
প্রর াজন। পডারেখাে পা াপাশ পাঠক্রম
বশির্ভূ ত কার্ূকো এ অং গ্রিণ কো ও
দেকাে।তারদে মানশিক শবকা এে জনয
শন শমত খখোধুো কোে ও প্রর াজন।
ছাত্ররদে পডারেখা, পাঠয বশির্ভূ ত কাজ ও খখোধুো
কো প্রর াজন।
একজন ছারত্রে পডার ানাে পা াপাশ পাঠয বশির্ভূ ত
কার্ূকো এ অং গ্রিণ ও মানশিক শবকার ে জনয
খখোধুো কোে ও প্রর াজন।
Abstractive
Extractive
5. • Extractive
Drag main sentence from source document and combine them to make summary
Easier
Too Prevailing
Most Past work
• Abstractive
Overcome the grammar inconsistency of Extractive
More difficult
More flexible
Necessary for future work
6. Deep Learning
• Artificial Neuron similar to the human brain
• That’s teach computer to do what comes naturally to human
• Sub set of machine learning
• Representing meaning of multiple level of representation
7. Objective
The main goal of our research is develop an abstractive Bengali text summarizer for generate
Bengali text summary. There are two ways to summarize text, one is extractive and other is
abstractive. In extractive method drag main sentence from main document and combine them to
make summary but in abstractive method its overcome the grammatically inconsistency of
extractive method. Abstractive method is more efficient then extractive method. Grammar
inconsistency is major obstacle to make a summary of text document. If we use abstractive method
in deep learning algorithm we can reduce computational time and reduce inconsistency. So we
need to provide a inconsistent free and computationally fast text summarizer which give accurate
and fluent text summary.
8. Motivation
• Hence there is less available text summarizer for Bengali language rather than other language, so
our main goal to develop an abstractive text summarizer which is computationally efficient and
create summaries automatically.
• Find relevant information quickly
• Clustering similar document and present summary
• Reduce the size of document while presenting it’s content
• Understand the main concept of information in a short time
9. Future outcome
• A better and efficient Bengali text summarizer
• Enrich lexical data base for Bengali language
• Create large resource of Bengali data
• Automatic Bengali text summarization