Research Of Digital Library Collaborative Filtering And GPU Computing Technology

Digital Library is concerned by many countries in recent years. It has achieved rapid development, and becomes an important approach for people's access to knowledge and information. Personalized Recommendation is one of important value-added services of digital libraries. This paper proposes an in-depth research to use the user's access logs of digital library, mining user preference, and conduct collaborative filtering recommendation. It makes digital library system to help the reader to find out more useful knowledge and information in the massive resources.The traditional digital library services are built based on the user's explicit demand, by the queries of the user's information needs associated with a specific digital resource, which is a passive information retrieval services that just meets the basic needs of users, and cannot provide personalized services according to the user's interest of reading.The main contribution of this paper is as follow: First, propose the method of the establishment of the user's preference list of digital resources mining from CADAL's logs, calculation of users' similarity through preference list, and recommendation by collaborative filtering. Second, design preference list based on collaborative filtering algorithms on general purpose GPU programming environment. Third, evaluate related performance of the algorithms, including the selection of parameters and performance analysis. Forth, complete the implementation of the algorithm, integrate it into the existing CADAL Digital Library Portal personalization module and achieve better recommendation effectiveness.