Improving Retrieval Performance with Wikipedia's Category Knowledge

For text search systems, the ambiguity of short queries often leads to poor performance. To solve this problem, relevance feedback via query-expansion is considered as one effective technique. However, many methods of relevance feedback barely use the knowledge of search results and the improvement of effectiveness is limited because the knowledge used is limited. In this paper we try to include Wikipedia's category knowledge to improve the poor retrieval performance. A method of category feedback is proposed, which is based on the information of Wikipedia categories. Categories instead of terms and documents are provided to users for feedback. Finally, an experimental search system is developed which demonstrates the effectiveness of our method.