질의어 단위 프로파일링을 이용한 북마크 기반 개인화 검색

초록

The approach of a single user profiling for personalized search does not satisfy a person’s diverse information needs. This paper presents a new approach, query-level user profiling, for personalized search in order to improve the situation. Our proposed method extracts query-related words from each user’s bookmarking data and makes use of them in the process of adjusting the user’s profile. We tested our approach using the field data obtained from CiteULike, a popular bookmarking service provider on the Web. Our experiment results show that the proposed approach through query- level user profiling outperforms the existing personalized search method based on single user profiling.

키워드