基于语义句模的语义理解方法研究

Research on Method of Semantic Comprehension Based on Semantic Sentence Template

针对目前情境信息丰富问题语义理解的困难,提出了一种基于语义句模的语义理解方法。该方法借鉴汉语言学中句模的研究成果,以及问题的分类,构建语义句模,将语义信息蕴含在句模中,利用文本中能够穷举的信息,为千变万化的情境信息赋予相应的语义。实验构建了基于自然语言处理工具Gate的语义句模jape规则库,以收集的102道四则运算应用题为例进行语义理解,实验结果为完全理解的题目为82.4%,部分理解为17.6%,完全不理解的0%。得出的实验结论是,该方法能够较好地实现情境信息丰富问题的语义理解。

According to the difficulty of semantic comprehension for the problems with rich contextual infomlation, present a method of semantic comprehension based on semantic sentence template. This method is to construct the semantic senteace templates based on the research of sentence template in the field of Chinese linguistic and the classifications of problems. The semantic is contained in the seman- tic sentence templates. The semantics of the contextual information can be recognized by the exhaustive information. Experiment con- structs jape rule library of semantic sentence template based on natural language processing tool Gate, 102 word problems are collected as examples for semantic comprehension, and the results demonstrate that the full comprehension topic accounts for 82.4% ,part comprehen- sion 17.6% ,without not comprehension. It is concluded that this method is effective for semantic comprehension of the problems with rich contextual information.