An Optimization Algorithm for Extractive Multi-document Summarization Based on Association of Sentences

Designing of automatic summary extraction technology becomes more and more important as the number of documents increasing rapidly. At present, the indicators that are often used for summary evaluation including the coverage, redundancy, and relevance. However, the summary cannot be completely extracted using only these three evaluation indicators. Therefore, with the concept of centrality, we design an association-based centrality criterion firstly, which can be used to evaluate associations of sentences to reinterpret the centrality of summary. Based on the three commonly used and the designed centrality factors, an optimization algorithm is then proposed for obtaining the summary from the given multiple documents. Experiments were also made on the dataset from the Document Understanding Conference 2002 (DUC2002) to show the effectiveness of the proposed algorithm in terms of ROUGE scores.