【作者】 胡珀；

【Author】 HU Po Department of Computer Science, Huazhong Normal University, Wuhan 430079, China

【机构】 华中师范大学计算机科学系；

【摘要】 文中提出了一种新的图论框架下的多文档自动文摘方法。为了尽可能地降低摘要中的冗余并保持其信息的多样性,采用了一种基于图模型的方法。该方法能自动地确定原始文档集合中包含的不同子主题,通过从各个子主题中提取代表句并重排句子以生成摘要。在中国新闻语料库上进行的实验表明该方法产生的摘要比典型的多文档自动摘要方法具有更好的综合性能。更多还原

【Abstract】 This paper proposes a novel approach for multi-document summarization, which casts the task in a graph-theoretic framework. For reducing redundancy and maintaining diversity in the final summary, we adopt a graph modeling-based method, which can determine the main subtopics in source documents auto-matically, then representative sentences are selected from these subtopics and are reordered to generate summary. Experimental results indicate that the proposed approach achieves better overall performance when compared with the baseline approach running on Chinese News Corpus.更多还原