摘要： 提出一种基于加权模糊聚类(WFC)和互信息(MI)的主题句提取方法，使主题句尽可能全面覆盖全文主题的同时，缩减自身的冗余，以提高摘要效率，采用加权模糊聚类的方法对文本句子进行分类，对在同一类中的句子使用比较互信息的方法进行排名处理，从而获得高质量的摘要。实验结果表明，与传统聚类方法比较，该方法的正确率提高约15%，可以达到约70%的精确度，并在阅读摘要时能够基本正确地获取文本信息。

关键词: 主题句, 加权模糊聚类, 互信息

Abstract: A topic sentence extraction method based on Weight Fuzzy Clustering(WFC) and Mutual Information(MI) is proposed, which is to cover more topics and lower the redundant information of the text. The abstract efficiency is promoted. Using WFC method, the sentences is classified. Sentences in each cluster is ranked by MI values. High qualified abstact is obtained. Experimental results show that, compared with former clustering method, this method can improve the precision by nearly 15%, and has about 70% accuracy. It can get text information correctly.

Key words: topic sentence, Weight Fuzzy Clustering(WFC), Mutual Information(MI)

中图分类号: 

• TP311.52