A Flexible Re-ranking System Based on Sub-keyword Extraction and Importance Adjustment

Most of current search engines invoke a function of query expansion for a re-retrieval or a re-ranking, which provides sub-keywords related to query keywords entered by a user. However, users cannot make full use of sub-keywords, except for selecting or abandoning them. This paper proposes a user-tunable system for re-ranking search results. The system automatically extracts sub-keywords complementing the meaning of query keywords, and constructs a radar chart interface by using the extracted sub-keywords as its chart items. The radar chart interface enables users to freely adjust the importance between multiple sub-keywords or replace inappropriate sub-keywords, so that user-desired pages can be re-ranked to the top place. Empirical evaluation, conducted by not only the author but also a total of 100 individuals, shows that the appropriate sub-keywords can be extracted and the re-ranking performance is good.