Abstract: In light of the inadequacy of popular systems of generating summaries, a context-sensitive program is put forward called SSC (Summarizer Sensitive to Context). The program has one characteristic which can reduce the content redundancy maximally and it is sensitive to the context. First the most important sentence is selected as a sentence in summary, and then the weight of terms in the sentence is set to very small constant so as to prevent the recurrence of these terms. ROUGE is used to evaluate the program. The experimental results show that the method has achieved good results.

Key words: multi-document summarization, sensitive to the context, complexity

摘要： 针对现有的多文档自动摘要生成方法中存在的问题，提出一种多文档自动摘要生成方法，该方法能够最大限度地减小摘要内容的冗余。选取权重最大的句子作为摘要句，把已选句子中包含的词汇的权重设置为接近0的常数，当下次选择摘要句时，可以避免再次选取包含这些词汇的句子。使用自动摘要评测方法ROUGE对该方法进行评测。实验结果表明，根据该方法抽取的机器摘要能够获得较高的 成绩。

关键词: 多文档自动摘要, 上下文敏感, 复杂度

CLC Number: 

• TP391