Exploiting Scholar��s Background Knowledge to Improve Recommender System for Digital Libraries

Abstract

Recommender systems for digital libraries have received increasing attention since they assist scholars to find the most appropriate articles for research purposes. Many research studies have recently conducted to model the user interests in order to suggest scientific articles based on the scholar's preferences. However, a major problem of such systems is that they do not subsume user's background knowledge into the recommendation process and scholars typically have to sift manually irrelevant articles retrieved from digital libraries. Therefore, a challenging task is how to collect and exploit sufficient scholar's academic knowledge into the personalization process in order to improve the recommendation accuracy. To address this problem, a recommender framework that consolidates scholar's background knowledge based on the ontological modeling is proposed. The framework exploits Wikipedia as a lexicographic database for concept disambiguation and semantic concept mapping. The practical evaluation by a group of scholars over CiteSeerX digital library indicates an improvement in accuracy of recommendation task.