Abstract: Because of more and more data nestled in Deep Web, how to find the most relevant Web databases for user’s query requirements has become a problem demanding prompt solution. An approach based on topic distribution of Web database is proposed for Web database selection of Deep Web data integration. It acquires the content summary of Web database in the form of topic coverage, and then gets the topics of user query by using the appointed Web database. The database selection is made under query topics and topic coverage distribution matrix. The experiments on the real Web database have proved that this approach can not only achieve high recall, but also reduce price of building database content summary.

Key words: Deep Web, Web database selection, topic distribution, topic coverage

摘要： 随着越来越多的信息隐藏在Deep Web中，针对用户查询找出最相关的Web数据库成为亟待解决的问题。提出了一种基于Web数据库主题分布的方法用于Deep Web数据集成中的Web数据库选择。获取主题覆盖度形式的Web数据库内容描述，而后利用选定的Web数据库获取查询主题，最终由查询主题和主题分布矩阵来选择Web数据库。在真实Web数据库上的实验结果表明，该方法既取得了较高的查询召回率，也可有效降低数据库内容描述建立的代价。

关键词: Deep Web, Web数据库选择, 主题分布, 主题覆盖度