Library Book Recommendations Based on Latent Topic Aggregation

During recent years, how to provide personalized services has become an important research issue in library services. The libraries provide more and more personalized services such as customized web interface and reading suggestions. In the traditional approaches, the features of the books that a reader likes are used to construct the profile of the reader to support recommendation of books such as query keywords. But with the fact of the huge holdings in the libraries, the librarians need to effectively help the readers to find the books of interest. Collaborative filtering (CF) is a way to make it possible by use patron’s circulation logs which contain their borrow history as favorite readings. In this paper, we first use Latent Dirichlet Allocation to find the latent topics existing in the circulation logs, then we combine patron reading histories with the generated latent topics to produce a suggestion list for the patron. With the elaborated experiments demonstrated in this paper, it showed good results from the volunteers’ feedback.