基于本体的领域自动问答系统研究

由于传统信息检索返回的结果难以充分理解用户的问题语义,因此以医学领域本体为例,提出一种具有语义推理的自动问答系统。系统在领域知识本体上,通过链式索引结构抽取问题对应在领域知识本体中的命名实体。为理解问题的语义层次,通过改进CFN汉语框架网结构,给出从领域知识本体直接生成能理解问题语义的QFN问题框架本体的映射算法。运用QFN将自然语言问题转化成RDF三元组结构,自动生成问题对应的SPARQL查询语句,同时调用Jena推理机完成语义推理查询在知识本体中查找并给出问题的相关回答。实验结果表明,该方法相比传统的信息检索,可以理解问题表达语义并给出与问题语义相关度高的答案。

DOMAIN AUTOMATIC QUESTION ANSWERING SYSTEM BASED ON ONTOLOGY

It is difficult to fully understand the user s problem semantics by traditional retrieval.Taking the medical domain ontology as an example,we proposed an automatic question answering system with semantic reasoning.On the domain knowledge ontology,the system extracted the named entity which was corresponding to the problem in the domain ontology through the chain index structure.In order to understand the semantic level of the problem,we presented a mapping algorithm of QFN problem framework ontology which could directly understand the problem semantics from domain knowledge ontology by improving the structure of CFN Chinese framework network.QFN was used to transform the natural language problem into RDF triples,and the corresponding SPARQL query statement was automatically generated.Jena inference machine was called to complete the semantic inference query to find relevant answer in the knowledge ontology.Experimental results show that compared with the traditional information retrieval method,the proposed method can understand the problem expression semantics and provide the answer with a high degree of relevance to the semantic expression of the question.