A Query Expansion and Document Re-Ranking System Based on Single User Profile

• Main Authors: Tsung-han Yang, 楊宗翰
• Other Authors: Yih-chearng Shiue
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/33398161756006023682

碩士 === 國立中央大學 === 資訊管理研究所 === 98 === With the rapid growing information on Internet, the issue of helping user to filter useless information and to reduce the burden of browsing has become important. Therefore, a query expansion and document re-ranking system based on single user profile is proposed to accomplish personalized information retrieval. Collecting the Web pages of the user’s past browsing via web crawler to build a single user profile with multi-topic, and resolve the traditional problem of building multiple profiles for each topic of user’s interest. When user submits a query, the system will recommend personalized expansion words based on the user profile to filter uninterested documents, and just need to compare one single user profile so that expansion words can be produced. Comparing to the traditional process that must first determine which interest user profile is belonged to the query and then expansion words being produced; the proposed system can improve the efficiency. And re-ranking the search results based on user profile to rapidly help user to find personalized search results which are useful and interested and the burden of user browsing can be reduced. The experimental results prove that automatic query expansion based on single user profile can improve the retrieval performance, and after re-rank the search results, the retrieval performance is significantly improved. The proposed query expansion and document re-ranking system based on single user profile provides better efficiency in personalized query expansion, and can recommend expansion words according to user’s interests to filter irrelevant Web documents to acquire the actual needed information, and the burden of user browsing is effectively reduced.