Skip to Main Content
BM25 is a well known ranking method to measure the relevance of the matching documents to a query.BM25 ranking functions can be used in recommendation when the query is transferred to represent userspsila interest. However, in this scenario,terms in the query need to be handled differently because they may come from sources that users have different levels of interests in. For example, some terms come from areas that user has much interest in; some terms come from areas that user has some interest in.Course recommendation, which is very useful in university portals, can make use of studentspsila course selection history. Course recommendation should consider that students have different levels of interests in different courses they selected. This paper proposes an improvement on BM25 to fulfill this requirement.Other than course recommendation, the improvement can be used in information systems of other areas, for example, financial news recommendation. The experimental result shows the validity and usability of the proposed method.
Date of Conference: June 30 2009-July 2 2009