A nalysing E motions in S chubert ' s E rlkönig : a C omputational A pproach
Tuomas Eerola · Music Analysis · 2010
ABSTRACT The present article provides an outline of the computational prediction of emotions expressed by music. A recent model based on the extraction of acoustic features derived from film music is used to demonstrate each phase of the model's construction, and the musical relevance and predictive accuracy of such models are discussed. This model, as well as a stylistically more appropriate model based on piano performances, is applied to a segmented analysis of Schubert's LiedDer Erlkönig. The predictions are compared to analytical insights provided by Spitzer in this issue. The two approaches, music analytical interpretation and computational analysis of expressed emotions, yield often coinciding emotional nuances, particularly when the stylistically appropriate computational model based on piano performances is used. The potential synergies, especially those related to a more comprehensive analysis of timbral characters via computer tools as well as the more refined interpretations that only a music analyst is able to draw, are discussed.