MusiCog: A Framework to Assess Music Grouping Systems Using Cognitive Models

Abstract

Music recommendation is rich field with many systems coming into existence on web, which work on different principles with various underlying criteria. But it has been difficult to assess and learn from the feedback, generally, of users of such systems. Not all of them have an advantage of large user bases or quick access to them, like last.fm. For such systems, especially in the early stages, it is time consuming to wait and collect the data from a large number of users to verify and cross check their metrics to optimize them. We present here a general framework, MusiCog, which researchers could play with to tweak performances of their recommendation system modeling it in terms of user’s cognition. The user survey once done will be useful through many development cycles of system. The system has been promising so far, it’s being applied on Indian film music in an ongoing experiment.


Back to Saturday Posters