Evaluation of n-Gram-Based Classification Approaches on Classical Music Corpora

摘要

The paper deals with evaluation of various n-gram-based composer classification algorithms. Our analysis has a broad scope: We have analyzed three labelled corpora, five similarity measures, several feature extraction methods, the influence of forced balanced training and an extensive range of n-gram lengths. We found that most of the approaches we analyzed, when properly parametrized, can give very good results, on par with other state-of-the art data mining techniques and greatly outperforming humans in composer recognition.