This project is developing a general recommendation engine (GRE) that any NSDL system can integrate to provide recommendation services to its users. GRE selects the relevant materials for users, be they students doing assignments, teachers preparing for classes, or researchers trying to understand a new topic area. As GRE is implemented and incorporated with more NSDL collections, it will improve the information search of NSDL users, especially those who are not familiar with a subject area and its available resources. GRE integrates the three most dominant recommendation technologies - collaborative filtering (CF), content-based filtering (CB), and knowledge-based recommendation (KB). In order to further improve the recommendation accuracy, this project refines a preliminary proof-of-concept user task module by implementing an 'implicit user profiler' that learns the current user task from user behavior. This project is also investigating the optimal configurations of the three recommendation engines in different situations; e.g., for varying subject domains, user types, and user tasks. A final element of the project is a comprehensive investigation of the impacts of recommendation systems on digital library users.
