Thesis: Text Summarization using Deep Belief Networks

Abstract

The main aim of this thesis work is to build an automated generic extractive summarization system that will summarize legal case reports taken from the Federal Court of Australia (FCA) using deep learning techniques. Deep Belief Networks (DBN) is the architecture that is used to learn the high level abstract representation of the given case reports, and these features could potentially be a part of the summary. Once the features are learnt using the DBN, a summary for each case report (single-document summarization) is generated by scoring the sentences.