Adaptive User Modeling in a Content-Based Music Retrieval System
Pierre‐Yves Rolland · 2001
Such data can indeed be utilized by a CBMR system to bias its search results, hopefully making the latter more accurate. Based on user data the system builds expectations which are used to filter candidate results. For instance, if a song belonging to the user’s most usually retrieved music genre matches the user’s query, it would be advantaged by the system over a matching song of a genre the user never retrieved before. Similarly for songs belonging or not to a genre declared by the user as being among his/her favorite. A number of methods have been proposed for collecting, representing and updating these more traditional kinds of user data. These are out of the scope of this paper.