Abstract
This paper presents the preliminary analyses towards the development of a formal method for generating autonomous, dynamic ontology systems in the context of web-based audio signals applications. In the music domain, social tags have become important components of database management, recommender systems, and song similarity en- gines. In this study, we map the audio similarity features from the Iso- phone database [25] to social tags collected from the Last.fm online mu- sic streaming service, by using neuro-fuzzy (NF) and multi-layer percep- tron (MLP) neural networks. The algorithms were tested on a large-scale dataset (Isophone) including more than 40 000 songs from 10 different musical genres. The classification experiments were conducted for a large number of tags (32) related to genre, instrumentation, mood, geographic location, and time-period. The neuro-fuzzy approach increased the over- all F-measure by 25 percentage points in comparison with the traditional MLP approach. This highlights the interest of neuro-fuzzy systems which have been rarely used in music information retrieval so far, whereas they have been interestingly applied to classification tasks in other domains such as image retrieval and affective computing.