退役军人领域问答库自动构建方法

Automatic Generation of Veterans Domain"Q&A"Database

问答库的自动生成是智能问答领域的一个基础的研究方向,在问询系统中尤为重要,例如自主研发的面向退役军人咨询服务的智能问答系统。主要是利用陈述句中的句法和语义信息,将其转化为问题并利用同义词词林生成多层次拓展的问题集和答案组成问答对。论文提出一种基于文本简化、问题生成、问题排序和问答对生成的四步骤方法,用来解决自动构建问答库带来的挑战。以《搜狗分类语料库》中的句子为例,对所提出的方法和系统进行了评价,论文方法准确率提升4%,结果表明此方法构成的知识库有更合理、质量更高的问答对。

The automatic generation of question answering database is a basic research direction in the field of intelligent ques tion answering,which is particularly important in the question answering system,such as the self-developed intelligent question an swering system for veterans consulting service.It mainly uses the syntactic and semantic information in declarative sentences to transform them into questions,and uses synonym forest to generate multi-level expanded question sets and answers to form question answer pairs.This paper proposes a four stage method based on text simplification,question generation,question ranking and ques tion and answer pair generation to solve the challenge of automatically constructing question and answer database.Taking the sen tences in Sogou classification corpus as an example,the proposed method and system are evaluated.The accuracy of this method is improved by 4%.The results show that the knowledge base composed of this method has more reasonable and higher quality question and answer pairs.