Improving the Performance of Text Summarization

This book is concerned with the area of automatic summarization of multiple documents. We propose several methods that improve the accuracy of the summaries. We have explored document density in graph-based multi-document summarization. We have tested the assumption that the higher the density of a document, the higher the salience (score) of its sentences leading to better summaries.We have also done work on user-based models of multi-document summarization. We note that different users can generate rather different summaries on the basis of the same source data and query. We have exploited machine learning techniques to generate models oriented to specific users, but for the same query and source data. Also, we have explored the actor-object relationship (AOR) between sentences. This work lead again to marked improvements in the overall result, particularly when combined with the rather successful approach that involves ensemble summarizing system.