Suggestions for Fresh Search Queries by Mining Mircoblog Topics

Query suggestion of Web search has been an effective approach to help users quickly express their information need and more accurately get the information they need. All major web-search engines and most proposed methods that suggest queries rely on query logs of search engine to determine possible query suggestions. However, for search systems, it is much more difficult to effectively suggest relevant queries to a fresh search query which has no or few historical evidences in query logs. In this paper, we propose a suggestion approach for fresh queries by mining the new social network media, i.e, mircoblog topics. We leverage the comment information in the microblog topics to mine potential suggestions. We utilize word frequency statistics to extract a set of ordered candidate words. As soon as a user starts typing a query word, words that match with the partial user query word are selected as completions of the partial query word and are offered as query suggestions. We collect a dataset from Sina microblog topics and compare the final results by selecting different suggestion context source. The experimental results clearly demonstrate the effectiveness of our approach in suggesting queries with high quality. Our conclusion is that the suggestion context source of a topic consists of the tweets from authenticated Sina users is more effective than the tweets from all Sina users.