一种结合本体和焦点的问题分类方法

鉴于问题分类是问题分析的主要任务,提出一种结合本体和焦点的问题分类方法.首先依存分析和语义角色标注对问题进行浅层语义分析,再根据预定义的问题焦点结构和焦点抽取规则,获取问题焦点语义表征;然后标示问题的类别为问题焦点中疑问对象在领域本体中的标识;最后,根据焦点不同则问题不同这一事实,将焦点相同的问题归为一类,从而实现问题分类.使用该方法对计算机故障诊断领域1 905个特指问题分类,取得了93.91%的准确率,验证了领域本体和焦点对问题分类方法的有效性.

A Question Classification Method Combining Domain Ontology and Question Focus

Question classification is one of most important tasks of question analysis in question answer system. A novel question classification method combining domain ontology and question focus is proposed in this article. First, with the supports of predefined structure of question focus and extracting rules, the semantic representation of question focus is fetched from the results of shallow semantic analysis including dependency analysis and semantic role labeling. Then, questions are labeled with the IDs of question objects in domain ontology. Thus, according to the fact of "same questions have same focus, and questions with different focuses are of different questions", questions with same focus are classified into the same category. The experimental result found that 93.91% accuracy for 1 905 special questions in restricted domain of computer troubleshooting could be achieved by using the proposed method. It demonstrates the validity of question focus and domain ontology in question classification task.