Query Recommendation from Search Engine Query Logs Using Query Enrichment

This article presents a technique to find semantically related queries from the search engine query log. The proposed technique is applied to the query recommendation task. The experiment is carried out on the log of user queries written in Thai. We first discuss the nature of Thai queries that makes the problem of finding related queries much more challenging. Not only are the queries short and ambiguous, the ambiguity of word boundary and ill-written queries also make the fundamental process of Thai text processing like word segmentation problematical. To cope with these problems, we propose the use of results from search engines to expand and enrich the meanings of queries. The experimental results show that the proposed technique is quite robust although the errors of word segmentation inevitably occur.