Recommending audio mixing workflows

Abstract

This paper describes our work on Audio Advisor, a workflow recommender for audio mixing. We examine the process of eliciting, formalising and modelling the domain knowledge and expert's experience. We are also describing the effects and problems associated with the knowledge formalisation processes. We decided to employ structured case-based reasoning using the myCBR 3 to capture the vagueness encountered in the audio domain. We detail on how we used extensive similarity measure modelling to counter the vagueness associated with the attempt to formalise knowledge about and descriptors of emotions. To improve usability we added GATE to process natural language queries within Audio Advisor. We demonstrate the use of the Audio Advisor software prototype and provide a first evaluation of the performance and quality of recommendations of Audio Advisor. © 2013 Springer-Verlag.