Research On Personalized Book Information Service Technology

With the widly application and development of digital library service system in university library, personalized book information service technology is being paid more and more attention. It collects information of books for reader based on different disciplines, after filtering data, recommending useful information of books to reader. There are two key technologies in personalized book information service, which is reader interest modeling and personalized book recommendation technology. With the development of the number of users in the digital library service system and the rapid increase in the amount of library collections, the personalized book information service technology faces enormous challenges.Aiming at the problems and challenges of the personalized book information service technology, this thesis focuses on analysis of two key technologies. This work mainly includes the following aspects:First of all, in view of shortcomings which existing vector space model can not fully describe the user's interests, reader interest model with hierarchy and categories concept is proposed in this thesis. What's more, in view of shortcomings which existing personalized recommendation technology solves on the strategy for user's interested changes, an updating algorithm for interest model with feedback is put forward in the thesis. Then the comparative experiment of prediction accuracy is done among the proposed algorithm, forgetting strategy and time window strategy. The results show that performance of the proposed algorithm is better than the other two schemes. Finally, aiming at sparse matrix of traditional collaborative filtering algorithm, the thesis introduces an improved collaborative filtering algorithm which is based on reader interest model. The algorithm includes two personalized book recommendation processes, the recommendation which is based on subject word and the process to recommendation the new subject word. Finally, the comparative experiment of precision, recall and the value of F-measure are done on proposed algorithm and traditional collaborative filtering algorithm. Experiments show that performance of the algorithm is superior to the traditional collaborative filtering algorithm.