Deep Web Database Selection with Classification and Rich Features

Abstract

The Web has been rapidly deepened by many searchable databases online, where data are hidden behind query interfaces. There may be hundreds or thousands of Web databases providing data of relevance to a specific domain on the Web. In the face of these large-scale Web databases, the core problem is to select the most appropriate ones to a users query. While this problem has received more attentions recently, current approaches still have the simplified and empirical limitations. In this paper, we propose a Web database selection approach based on classification. We cast Web database selection as a classification problem and combine multiple kinds of features which are about the query and Web databases. We use the classification model to obtain the relevancy of every individual Web database for a user query and select top-K ones to provide the query results. Experiments show that our approach yields high performance.