Research on Sentence Optimum Selection Algorithm for Multi-Document Summarization

Analyzing sentences selection in summarization, an approach based on deleting sentences in a sentences set to obtain summary is proposed, which differs from the traditional method of adding sentences to get the summary. It has two stages, one is the process of obtaining the candidate summary sentences set with direct obtaining algorithm and redundancy-based obtaining algorithm, the other is the process of deleting sentences with sentences optimum algorithm. With DUC 2004 as the test corpus, the ROUGE value of summaries gotten by sentences selection proves the necessity of sentences optimum selection for multi-document summarization. Compared with the redundancy-based sentences selection method, the validity of the approach proposed is proved.