基于配价结构和语义依存关系的句法分析统计模型

摘要

The mainstream lexicalized syntactic parsing models currently only utilized semantic dependency between words , and did not incorporate other semantic information such as semantic collocation and semantic category .“valency”is a essential se-mantic feature of words ,once the valency of word is determined ,the collocation of the word is clear ,and the sentence structure can be directly derived .Thus a syntactic parsing model combining valence structure with semantic dependency is purposed on the base of head-driven statistical syntactic parsing models .The parsing model incorporates rich semantic information including semantic depen-dency and semantic collocation in the decomposition and probability computation of the rules .Experiments are conducted for the re-fined statistical parser ,it achieves 88.76% precision and 87.43% recall ,F measure is improved 6.65% comparing with the head-driven parsing model introduced by Collins .%目前主流的词汇化句法分析方法仅仅考虑词语之间的语义依存关系，而没有引入语义搭配和语义类等语义信息．“配价”是词语的一个比较本质的特点，一旦一个词语的配价结构确定下来，它应该和怎样的词进行搭配也就比较清楚了，从而也可以比较直接地导出句子的结构．本文结合中心词驱动句法分析模型，提出了基于配价结构和语义依存关系的句法分析模型．模型在规则的分解及概率计算中引入丰富的语义信息，既包括语义依存信息，也包括配价结构等语义搭配信息．用改进的句法分析模型进行句法分析实验，实验结果表明，精确率和召回率分别为88.76％和87.43％，综合指标F值比Collins的中心词驱动句法分析模型提高了6.65个百分点．