Keyword Query Cleaning with Query Logs

Abstract

Keyword queries over databases are often dirty with some irrelevant or incorrect words, which has a negative impact on the efficiency and accuracy of keyword query processing. In addition, the keywords in a given query often form natural segments. For example, the query “Tom Hanks Green Mile” can be considered as consisting of two segments, “Tom Hanks” and “Green Mile”. The goal of keyword query cleaning is to identify the optimal segmentation of the query, with semantic linkage and spelling corrections also considered. Query cleaning not only helps obtaining queries of higher quality, but also improves the efficiency of query processing by reducing the search space. The seminal work along this direction by Pu and Yu does not consider the role of query logs in performing query cleaning. Query logs contain user-issued queries together with the segmentations chosen by the user, and thus convey important information that reflects user preferences. In this paper, we explore the use of query logs to improve the quality of keyword query cleaning. We propose two methods to adapt the scoring functions of segmentations to account for information gathered from the logs. The effectiveness of our approach are verified with extensive experiments conducted on real data sets.