基于本体的Deep Web语义分类研究

Research on Deep Web semantic categorization based on ontology

针对目前Deep Web分类研究中所采用的Post-query查寻探测方法缺乏语义支持的问题,提出一个基于本体的语义查询探测分类方法。主要思想如下:首先针对一个Deep Web数据库集合,提取查询接口中的属性及其实例,半自动建立领域本体,并且通过领域本体来表示类别特征;然后利用领域本体中的概念以及相应的实例构造语义查询集;最后对待分类的Deep Web数据库利用语义查询集进行查询探测,计算查询探测返回的结果文档在领域本体中的信息覆盖量,并以此对Deep Web进行分类。实验表明:这种语义查询探测分类的方法和以往的方法相比,在准确率、查全率和F1值上有一定的提高。

In view of the problem of lacking semantic support in the Post-query research of Deep Web databases classification,the paper designs a novel semantic query probing classification approach based on ontology.The main idea is as following： firstly,attributes and instances are extracted from query interface,which are used to build domain ontologies semi-automatically,and characteristics of categories are represented by domain ontologies.Then domain query instances are constructed from domain ontologies,which are used as query probing to Deep Web databases.Finally coverage degree between returned result documents of a query and domain ontologies are computed,with which the Deep Web database is classified.The experiments show that semantic query probing classification method we proposed has improved a lot in precision,recall and F1.