A piece of music derives coherence with repetition or looping, motivic development, and reference. These aspects of music are also the most difficult for current statistical and machine learning approaches to music generation because they require a representation power beyond even context-free grammars. Transformational approaches to music generation, which represent and conserve the structure of a template piece, are a possible solution to this problem.
This talk will present the concept of semiotic patterns and show how they can be derived from template pieces and instantiated to generate pieces with similar structure but completely new music material. Examples of music generation by transformation will be presented from several genres, with particular attention to trance.
Darrell Conklin is a research professor at the Department of Computer Science and Artificial Intelligence at the University of the Basque Country, Spain. In music informatics he investigates knowledge representation and machine learning methods for computational music analysis and music generation. He is a Principal Investigator on the EU FP7 project Lrn2Cre8 which studies the generation of music from models learned from symbolic music corpora.