Smart Radio - a proposal

Abstract

There is significant value in having predictions for an item before deciding whether to invest time or money in consuming that item. In a web based scenario where the items are multimedia items such as audio, recommendations can be made to users based on an understanding of their previous consumption or their indications of likes and dislikes. We examine two types of recommendation: content based and non-content or collaborative recommendation. We then apply out thinking to the area of new internet services such as online radio, and propose an architecture for an intelligent music radio system. We then suggest the efficacy of using conceptual clustering techniques in such a paradigm.