Artificial Music Critics

This paper proposes a framework for the simplification of the development of Artificial Art Critics. We provide two basic elements: an architecture that consists of two main modules for the pre-processing and classification of an artwork, and a validation methodology that consists of several stages, such as the objective evaluation of an artwork (with targets like author or style identification) and a dynamic evaluation that implies the integration of the Artificial Art Critic into a multi-agent environment. We also present some experimental results concerning the first stage of the validation methodology. The results show the ability of the system to identify the author of a musical piece and its adaptive capacity to determine the relevant features of the musical piece.