Subsequence kernel-based Chinese recommender system

With the rapid growth of Chinese websites in several years, how to get the useful information for users on Chinese websites has become a critic issue. This paper proposes the subsequence kernel-based Chinese recommender system, and subsequence kernel provides a flexible method to make the Chinese recognition easier, which overcomes the weakness of efficiency degrading for large scale users. Subsequence based on Kernel Matrix Factorization (KMF) is used to indicate that a phrase/word occurs contiguously within another phrase, and subsequence to allow the possibility that gaps separate the different characters resulting in a noncontiguous occurrence within the phrase. Experiments show that this method outperforms the others.