摘要： 在论述了面向自然语言文本的知识供应方法已有研究之后，提出了面向推理设计语义关系标记，将自然语言理解与推理技术相结合，使自然语言理解结果可转化为便于计算机推理的语句；立足科技文献开发语义关系树库，提出面向语义关系分析的语句相似度计算方法，实现基于统计与规则相结合的自然语言语义关系分析；使自然语言科技文献可成为专家系统知识库，解决专家系统知识库表现形式不统一及知识获取的瓶颈。

关键词: 自然语言理解, 知识获取, 推理技术, Prolog

Abstract: After discussing current research on knowledge acquisition method for document in natural language,the paper designs a set of semantic relation markups, which make for reasoning, integrates natural language understanding and reasoning technology, so that the result of natural language understanding can be turned into sentences which can used to reason by computer; rooted in literature of science and technology, developed semantic relation trees storeroom and put forward a method of calculating similarity degree between two sentences for semantic relation analysis, and realize analysis of semantic relation of natural language based on the combine of statistics and rules, so that document in natural language can be used as knowledge base of ES, so that the backward of uniform representation manner and bottleneck of knowledge acquisition were settled.

中图分类号: 

• TP182