Interest-based Recommendation in Digital Library

Abstract

With the huge amount and large variety of information available in a digital library, its becoming harder and harder for users to identify and get hold of their interested documents. To alleviate the difficulty, personalized recommendation techniques have been developed. Current recommendation techniques rely on similarity between documents. In our work, recommendations are made based on three factors: similarity between documents, information amount, and information novelty. With the introduction of degree of interest, users interests can be better characterized. Theoretical analysis and experimental evaluations demonstrate that our techniques can improve both the recommendation recall and recommendation precision.