一种自组织的汉语组合型歧义消歧方法  

Adaptive Chinese combinatorial ambiguities disambiguate method

作　　者：冯素琴[1] 陈惠明[1] 

机构地区：[1]忻州师范学院计算机科学与技术系,山西忻州034000

出　　处：《计算机工程与设计》2007年第3期737-739,742,共4页Computer Engineering and Design

基　　金：山西省忻州师范学院基金项目(200307)

摘　　要：组合型歧义切分字段一直是汉语自动分词的难点。用人工校验后的分词语料提供的搭配实例作为组合歧义字段的初始搭配知识,提出使用搭配统计表的多元最大对数似然比进行消歧;继而根据实验确定了歧义字段的上下文窗口、窗口位置区分、权值估计等要素;在此基础上采用自组织方法自动扩充搭配集,使消歧信息趋于稳定;最后,对提出的方法进行了实验,实验表明,该算法能有效提高消歧准确率。Combinatorial ambiguity is a vital issue in Chinese word segmentation. A multi maximal log disambiguation algorithm by collocations provided by using the checked word segmentation is presented; Then the key factors regarding combinatorial ambiguities is determined by experiments （the size of the context windows, the sensitivity of locations in the windows as well as weighting of feature word）; And then adaptive Chinese combinatorial ambiguities disambiguate method is used to strengthen and stabilize collocations; At lastly, the method is tested and proved that it can well raise accuracy.

关 键 词：自然语言处理 汉语自动分词 组合型切分歧义 自组织方法 消歧 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]