摘要
问句分析是问答系统的关键,为降低问句完整语法分析的复杂度,该文应用浅层句法分析理论,采用问句语义组块方式来分析问句。以“知网”知识库为基础,提取和定义了表达汉语问句的6种语义块,定义了语义组块最大熵模型的特征表示,通过最大熵原理实现了语义组块特征抽取和特征选取学习算法,并以模型为基础实现了真实问句的语义块的标注,从而为在语义层面上理解汉语问句奠定了基础。实验结果说明最大熵模型应用于汉语问句语义组块分析具有较好的效果。
The key of question-answering system is question parsing. This paper applies shallow sentence parsing theory and adopts the method of question semantic chunk to analyze question in order to decrease the complexity of full syntax parsing. Based on the knowledge database of Hownet, this paper defines six kinds of semantic chunks expressing Chinese question and the feature representation of semantic chunk in maximum entropy model. A learning arithmetic of feature extraction and feature selection of semantic chunks are implemented. The semantic chunks of true question are recognized based on maximum entropy model. The experiment results show that the maximum entropy model has a very good effect on semantic chunk parsing to Chinese question sentence.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2005年第17期3-5,8,共4页
Computer Engineering
基金
云南省信息技术基金资助项目(2002IT03)
关键词
最大熵模型
问句分析
句法分析
组块分析
语义块
Maximum entropy model
Question sentence parsing
Sentence parsing
Chunk parsing
Semantic chunk
