This paper was presented at the charming Sysmus 2011 in Cologne. The paper is going to be published soon at ePosMusic! This is the short version. Please contact the authors for a copy and further information.
Authors: Daniel Wolff and Tillman Weyde
Abstract: We present an approach to automatically adapting a computational model for music similarity to users’ cultural contexts, indicated by parameters like location and age. Using user ratings of perceived similarity, we aim to model variations in these ratings associated with culture-based variables of the users. The resulting similarity measure meets its useful application in music information retrieval, particularly personalised music recommendation, but relations of cultural variables and music perception, which are represented in the adapted similarity models, are also of generic musicological interest. This paper outlines the general framework envisaged for this task and first experimental results on the feasibility of learning musical similarity from relative user ratings.