Please use this identifier to cite or link to this item:
Title: From Extractive to Abstractive Summarization A Journey
Researcher: Mehta, Parth
Guide(s): Majumder, Prasenjit
Keywords: Engineering and Technology,Computer Science,Computer Science Artificial Intelligence
University: Dhirubhai Ambani Institute of Information and Communication Technology (DA-IICT)
Completed Date: 2018
Abstract: Research in the field of text summarisation has primarily been dominated by investigations of various sentence extraction techniques with a significant focus towards news articles. In this thesis, we intend to look beyond generic sentence extraction and instead focus on domain-specific summarisation, methods for creating ensembles of multiple extractive summarisation techniques and using sentence compression as the first step towards abstractive summarisation. Our proposed approach based on attention-based neural network learns to automatically identify these key phrases from pseudo-labelled data, without requiring any annotation or handcrafted rules. The proposed model outperforms existing baselines and state of the art systems by a large margin. newline
Pagination: x, 112p.
Appears in Departments:Department of Information and Communication Technology

Files in This Item:
File Description SizeFormat 
01_title.pdfAttached File78.8 kBAdobe PDFView/Open
02_declaration and certificate.pdf73.56 kBAdobe PDFView/Open
03_acknowledgments.pdf53.08 kBAdobe PDFView/Open
04_table of contents.pdf76.09 kBAdobe PDFView/Open
05_abstract.pdf74.68 kBAdobe PDFView/Open
06_list of tables and figures.pdf63.88 kBAdobe PDFView/Open
07_chapter 1.pdf102.69 kBAdobe PDFView/Open
08_chapter 2.pdf170.97 kBAdobe PDFView/Open
09_chapter 3.pdf275.05 kBAdobe PDFView/Open
10_chapter 4.pdf202.21 kBAdobe PDFView/Open
11_chapter 5.pdf196.44 kBAdobe PDFView/Open
12_chapter 6.pdf207.56 kBAdobe PDFView/Open
13_chapter 7.pdf75.92 kBAdobe PDFView/Open
14_references.pdf91.68 kBAdobe PDFView/Open
15_appendix.pdf98.79 kBAdobe PDFView/Open

Items in Shodhganga are protected by copyright, with all rights reserved, unless otherwise indicated.