摘要： 研究了特定领域的文本的信息抽取，主要考虑了文本分布的观点。首先从未标注的语料中学习主题和主题间的关系，然后把它应用在同领域的文本信息抽取。经测试，其信息抽取的效果有所提高。

关键词: 主题；信息抽取；聚类；k 近邻；

Abstract: This paper studies information extraction of special domain and mainly considers the view of text distribution. First, it studies topic and relation of topic from un-annotated corpus, then applies it in text information extraction of the same domain. The experiment indicates that the result of the method improves a lot than previous ones to some extent.

Key words: Topic; Information extraction; Clustering; K-means