摘要
搭配是汉语自动句法分析的重要知识源,而动词是句法分析的核心和前提。通过对已标注真实文本的分析,构造了动词搭配对的上下文变量信息特征模板,给出利用最大熵方法抽取动词—动词搭配,对待测的1000句汉语句子应用最大熵方法自动识别出搭配,其中封闭测试抽取正确率为85.6%,召回率达到70.6%。
Collocation plays an important role in parsing and verb is the kernel and precondition for Chinese parsing.This paper presents a method for verb--verb collocation based on maximum entropy principle,using the constructed characteristic modeling for context variable information via analyzing the real text labeled.By testing of 1 000 sentences with the maximum entropy principle,it has obtained S5.6% accuracy and 70.6% recall ratio.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第3期135-137,共3页
Computer Engineering and Applications
基金
国家自然科学基金(No.60775041)~~
关键词
语料
搭配
特征选择
最大熵
corpus
collocation
feature selection
maximum entropy