Retrieval model of social books based on multi-feature fusion

Social book has not only traditional descriptions added by the expert editors, but also contains a wealth of user-generated data defined as social information such as user comments, custom labels. There are some limitations in retrieving social books with traditional retrieval methods. So, this paper presented a retrieval method of social books based on multi-feature fusion. In this method, we extract multiple social features of books and fuse them as one single similarity matrix. Using it, the K nearest neighbors of each book can be found. Then, we re-rank the initial ranking retrieved by traditional model using the fused multi-feature similarities of one book´s K nearest neighbors, so as to improve the effectiveness of traditional retrieval method on social book search. Using Amazon/LT social books collection provided by INEX, compared with traditional retrieval method, we conduct some experiments. The results show that feature extraction methods and the social re-ranking model presented in this paper can effectively improve the performance of traditional retrieval method.