Akira Karasudani

Extracting document relationships by analyzing user's activity history

In order to reduce peoples workload of looking for valuable information from a large amount of available information, recommendation systems, task management systems, and so on are attracting considerable attention. Such systems are expected to allow easy access to information under the current work context. In the development of these systems, how to handle a users current work context to know what information he wants to use is a key point. We have developed a novel method to extract relationships among documents as the work context by analyzing the users activity history according to several viewpoints of a human being who memorizes and seeks information. We report the details of the proposed method, the evaluation result, and an application example.

ACM HotMobile 2013 poster: extraction algorithm of relationship between smartphone applications for recommendation

Users employing smartphones typically combine multiple applications to perform their tasks. It would be possible to be recommended the appropriate applications by acquiring the contexts of users who perform such tasks. And the contexts are composed by the relationship of applications used in the tasks. We present an algorithm that extracts the relationship between applications that the user intentionally uses in combination. At the end of the paper, we report the results of verification tests conducted on this algorithm.