Mining query subtopics from social tags

Introduction. One of the principal difficulties confronted by modern search engines is the overwhelming amount of, and lack of focus in, retrieved results. The goal of this study is identification and extraction of an assortment of diverse subtopics for the query and reduction and re-ranking of search results, according to the most prominent and discriminative subtopics. Method. Social tagging as a 'wisdom of crowds' source for characteristic query subtopics was employed. A new method for extraction of subtopics from social tags was devised. Three different ranking measures for query result are presented and comparatively evaluated. Analysis. We propose a new automatic methodology for evaluation of the quality of the extracted subtopics. For this purpose, we assess the minimal potential contribution of our approach to enhancement of query result ranking and focusing. The proposed methodology is based on standard information retrieval measures, such as precision, recall and the mean average precision. Results. The obtained results with the Google search engine and the Delicious social bookmarking site as a source of social tags indicate that the mean average precision of search results was improved after their filtering and re-ranking according to the subtopics extracted from social tagging. Conclusion. Social tags can be effectively utilized for query subtopics extraction. The quality of the subtopics can be automatically evaluated by their contribution to focusing and ranking query results. Query result representation can be improved by their selection and re-ranking according to the prominent social subtopics.