A novel text - Mining system for generating abstract from extracted summaries using Anaphora resolution

Abstract

The amount of information available varies in length from onedocument to another. It becomes difficult and time-consuming activity to browse the information completely. It is essential to provide the information in a condensed form expressing the central idea of the document. Automatic text summarization is used for generating the summary for the document. This paper presents a novel Abstract Generation System (AGS) to generate an abstract from the extracted summary of an English language text document. The pronominal Anaphora Resolution (AR) Algorithm has been designed and used in AGS for scrutinizing and resolving the anaphors, the referring expressions present in the extract to make the summary more readable. AGS finally generates a fine-tuned summary for the given document. The experiments are conducted using a test set taken from the trained corpus, in AGS and other existing Anaphora Resolution Systems (ARS). The results are compared with the model summary written by human beings. The standard metric of Information Extraction (IE) systems namely the success rate has been used to measure and study the performance of AGS. © 2012 Springer-Verlag.