摘要
谓语中心词的识别是句法成分分析中的一个非常重要的部分。本文提出了一种规则和特征学习相结合的谓语识别方法 ,将整个谓语识别的过程分为语片捆绑、谓语粗筛选和谓语精筛选三个阶段。在谓语粗筛选中 ,利用规则过滤掉明显不能充当谓语的词 ,得到一个准谓语集 ;在精筛选阶段 ,选择谓语的支持特征 ,根据统计计算得到每个特征对谓语的支持度 ,然后利用准谓语在句子中的上下文出现的特征对准谓语集中的词进行再次筛选 ,从而确定出句子的谓语中心词。经过测试表明 。
Recognizing the predicate head is an important part of the syntactic analysis of Chinese sentences.This paper presents a new approach to recognize the predicate head automatically,which combines a rule based method with a multi feature based method.The process of recognizing is broken into three sub process:preprocess,coarse filter and fine filter.We use a rule based method to filter the quasi predicate that may be the predicate of a sentence.In the fine filter,we select and compute a great diversity of features by statistic,then use these features to recognize the real predicate of this sentence.The result of experiments indicates that this approach is feasible and advanced.
出处
《中文信息学报》
CSCD
北大核心
2003年第2期7-13,共7页
Journal of Chinese Information Processing
