Improving graph based multidocument text summarization using an enhanced sentence similarity measure

Abstract

Multi document summarization is a process to produce a single summary from a set of related documents collected from heterogeneous sources. Since the documents may contain redundant information, the performance of a multi document summarization system heavily depends on the sentence similarity measure used for removing redundant sentences from the summary. For graph based multi document summarization where existence of an edge between a pair of sentences is determined based on how much two sentences are similar to each other, the sentence similarity measure also plays an important role. This paper presents an enhanced method for computing sentence similarity aiming for improving multidocument summarization performance. Experiments using two different datasets show the effectiveness of the proposed sentence similarity measure in improving the performance of a graph based multidocument summarization system.