Genetic semantic graph approach for multi-document abstractive summarization

摘要: The aim of automatic multi-document abstractive summarization is to create a compressed version the source text and preserves salient information. Existing graph based methods treat sentence as bag words, rely on content similarity measure did not consider semantic relationships between sentences. These may fail in determining redundant sentences that are semantically equivalent. This paper introduces genetic approach for summarization. Semantic from document set constructed such way nodes represent predicate argument structures (PASs), extracted automatically by employing role labeling (SRL); edges correspond weight determined PAS-to-PAS similarity, PAS-to-document relationship. relationship represented different features, weighted optimized algorithm. (PASs) ranked modified ranking In order reduce redundancy, we utilize maximal marginal relevance (MMR) re-ranks PASs use language generation generate summary top PASs. Experiment this study carried out using DUC-2002, standard corpus Experimental results reveal proposed performs better than other systems.