Shizuku2.0: Cooperative reading support system / Mao Tsunekawa, Haruki Ono, Kyoji Konishi... [et.al].

The paper aims to propose cooperative reading, which is a reading support technique that allows library users to help each other. To achieve cooperative reading, it is necessary for a user to discover others with similar interests. Therefore, this paper also aims to develop and evaluate a recommendation function that recommends similar users using Nippon Decimal Classification (NDC) Tree Profiling. Is the user recommendation using NDC Tree Profiling effective in finding similar users? Which parameter of NDC Tree Profiling method is the most effective expression of users‘ interests? We developed the Shizuku2.0 system to support the creation of a library user community in which users help each other efficiently and mutually. We also designed and developed NDC Tree Profiling, which enables the creation of library user profiles, for the purposes of the user recommendation mechanism. To verify the effect of the user recommendation mechanism, we performed an experiment with 37 student users to calculate recall and precision. We found that the recommendation using NDC Tree Profiling is more effective than a random recommendation. However, we also recognized that there is room for improvement relative to a past information recommendation technique. Moreover, we found the second level of the NDC code could be the most effective expression of users‘ interests. In the discussion of the optimization of parameters, we propose a new way of implementing the NDC Tree, based on the second division of NDC, which is expected to improve creation of user profiles.