Effective XML Keyword Search through Valid Lowest Information Unit

Abstract

Keyword search for XML documents has attracted much attention recently. Existing approaches mainly retrieve search results through identifying the Lowest Common Ancestor (LCA) of keyword match nodes. There are also techniques for removing redundant and irrelevant search results. However, the problem of whether a LCA fragment contains complete information has not been adequately investigated. To address this challenge, we first introduce the notion of Information Unit (IU). Its structural property ensures that it can provide complete information on topic to user. We then reason whether a search result is valid by analyzing the context of keyword match nodes. We propose that a meaningful returned result should be a Valid Lowest Information unit (VLIU), which is the lowest IU that contains all the keywords and is semantically valid. Finally, we conduct extensive experiments to demonstrate the effectiveness and efficiency of our approach.