基于词序列频率有向网的中文组合词提取算法  

Chinese combined-word detection based on directed net of word-sequence frequency

作　　者：陈建超[1] 郑启伦[1] 李庆阳[1] 严桂夺[1] 

机构地区：[1]华南理工大学计算机科学与工程学院,广州510640

出　　处：《计算机应用研究》2009年第10期3746-3749,共4页Application Research of Computers

基　　金：广东省自然科学基金资助项目(07006474);广东省科技攻关资助项目(2007B010200044)

摘　　要：随着人类知识体系的不断拓展和深化,很多组合词(多个词或语素组成的词)被创造出来用于表达新的概念。由于无法及时把组合词收录进词库,分词系统无法识别它们。为此,从文本中提取组合词成为智能计算领域的一个热门的研究方向。借鉴人类的认知心理模式,提出一种基于词序列频率有向网的组合词抽取算法,以识别自由文本中的组合词。算法首先建立描述文本中的词序列出现频率的有向网,然后通过独特的矩阵运算,逐步把组合词提取出来。算法的优点是无须借助专业的语言知识,在实验分析中,算法显示了较好的效果。Inspired by one of human being＇ s cognition patterns, this paper proposed a new detection algorithm based on directed net of word-sequence frequency to discover combined-words from free texts. The algorithm created a directed net which implicated frequency of any word-sequence in a certain text first, then extracted combined-words from the directed net along with our special matrix operations. The algorithm is free from any linguistics knowledge. In the experiment analysis, the algorithm achieved very good result.

关 键 词：有向图 组合词 词序列 认知心理模式 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]