基于语义要素组合的知识库问答方法  

Knowledge Base Question Answering Method Based on Semantic Element Combination

作　　者：刘飞龙 郝文宁 余晓晗 陈刚 刘冲 LIU Feilong;HAO Wenning;YU Xiaohan;CHEN Gang;LIU Chong(Institute of Command Information System,Army Engineering University of PLA,Nanjing 210000,China)

机构地区：[1]陆军工程大学指挥信息系统学院,南京210000

出　　处：《计算机工程》2018年第11期46-55,共10页Computer Engineering

基　　金：国家自然科学基金青年基金(71501186)

摘　　要：为进行知识库问答系统中的问题语义分析,提出基于语义要素组合的问答方法。采取词库识别和规则识别提取问题语义要素并依据预定义模式识别函数,结合问题依存分析树结构和函数类型组合语义要素,生成问题语义表达式后利用映射和联合消歧组成知识库语义表达式,然后将知识库语义表达式转换为SPARQL语句后执行问答。实验结果表明,该方法的F1平均值为0. 841,能够有效理解并分析问题语义。In order to analyze the question semantics in the knowledge base question answering system,a question answering method based on semantic element combination is proposed.Firstly,the semantic elements of the problem are extracted by lexicon recognition and rule recognition,and recognize functions based on predefined patterns.Then the semantic elements of the problem are combined with the structure of the problem dependency analysis tree and the type of the function to generate the semantic expression of the problem,then the semantic expression of the knowledge base is formed by mapping and joint disambiguation.Finally,the semantic expression of the knowledge base is converted to SPARQL statement and execute question and answer.Experimental results show that,the average F1 value of the method is 0.841,which can effectively understand and analyze the problem semantics.

关 键 词：问答系统 问题理解 语义要素组合 联合消歧 语义表示 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]