Data extraction for search engine using safe matching

Abstract

Our study shows that algorithms used to check the similarity of data records affect the efficiency of a wrapper. A closer examination indicates that the accuracy of a wrapper can be improved if the DOM Tree and visual properties of data records can be fully utilized. In this paper, we develop algorithms to check the similarity of data records based on the distinct tags and visual cue of the tree structure of data records and the voting algorithm which can detect the similarity of data records of a relevant data region which may contain irrelevant information such as search identifiers to distinguish the potential data regions more correctly and eliminate data region only when necessary. Experimental results show that our wrapper performs better than state of the art wrapper WISH and it is highly effective in data extraction. This wrapper will be useful for meta search engine application, which needs an accurate tool to locate its source of information. © 2011 Springer-Verlag.