An improved architecture of item-based collaborative filtering system for Chinese texts

Chinese papers have become a major resource for Chinese researchers to learn about the status of their fields. Information filtering technique can help people to find useful information among these resources online. An efficient information filtering approach is needed to prioritize Chinese papers so that Chinese researchers can spend less time searching for papers of their interest. Since traditional collaborative filtering algorithm has the problems of sparse matrix and low accuracy, which will influence the results of prediction. This paper presents the architecture for Chinese texts filtering. An algorithm based on the similarity of information items is also proposed. It can solve the sparse matrix problem and improve the efficiency of prediction. Preliminary experiment shows the efficiency of this algorithm.