Stepwise Refinement for Semantic Analysis of Natural Language Sentences

自然语言句子的语义分析过程可以看作是连续接收新词，对单词和部分句子的解释进行整合和提炼的过程。本研究提出了构建基于逐步细化解释的自然语言理解系统的新技术。我们的方法使用带有信息的变量，这些变量在我们的逻辑编程语言PAL中是非常强大的对象，由带有信息的变量表示的对象可以有效地处理对单词和句子的不同解释。我们还提出了句子语义分析的知识编译方法、单词和句子意义的表示方法以及利用带有信息的变量寻找句子解释的新框架。

The process of semantic analysis of natural language sentences can be regarded as the process of successively receiving new words and integrating and refining interpretations of words and part of sentences. This research has proposed new techniques for constructing natural language understanding systems based on stepwise refinement of interpretations. Our method uses variables with information which are very powerful objects in our logic programming language PAL. Objects represented by variables with information can effectively deal with varying interpretations of words and sentences. We also proposed methods of knowledge compiling for semantic analysis of sentences, method of representation of meanings of words and sentences and a new framework for finding interpretation of sentences using variables with information in PAL.