| This work is aimed at developing a more comprehensive approach to information filtering---the Comprehensive Information Filtering System (CIFS). CIFS is designed as a distributed, personal filtering system for mobile users. It uses trust and context information to supplement content-based relevance ratings, in order to provide richer, adaptive filtering capabilities. The underlying theory is that there exists an intuitive link between the relevance of information, the trustworthiness of the source and the circumstances under which the information is received. Therefore, the user profile combines the message content, context, and contact information into an indivisible scenario with a common relevance weight. CIFS is also designed to learn the user's profile elements over time, as a result of implicit feedback. Simulation data and the results of a preliminary user study (involving 5 users) of CIFS show that the combined scenario-based filter is more effective than individual content or contact-based filters. As well, the results show that the learning algorithm used shows great promise as do the implicit measures of reading time and user activity monitoring. |
