Using Discovered, Polyphonic Patterns to Filter Computer-generated Music

A metric for evaluating the creativity of a music-generating system is presented, the objective being to generate mazurka-style music that inherits salient patterns from an original excerpt by Frédéric Chopin. The metric acts as a filter within our overall system, causing rejection of generated passages that do not inherit salient patterns, until a generated passage survives. Over fifty iterations, the mean number of generations required until survival was 12.7, with standard deviation 13.2. In the interests of clarity and replicability, the system is described with reference to specific excerpts of music. Four concepts—Markov modelling for generation, pattern discovery, pattern quantification, and statistical testing—are presented quite distinctly, so that the reader might adopt (or ignore) each concept as they wish. 1 Aim and Motivation A stylistic composition (or pastiche) is a work similar in style to that of another composer or period. Examples exist in ‘classical’ music (Sergey Prokofiev’s Symphony No. 1 is in the style of Joseph Haydn) as well as in music for film and television, and in educational establishments, where stylistic composition is taught ‘as a means of furthering students’ historical and analytical understanding’ (Cochrane 2009). If a computational system produces successful stylistic compositions (‘successful’ in the sense of the ‘indistinguishability test’ of Pearce and Wiggins (2001) for instance), then it is capable of a task that, in the human sphere, is labelled creative. The creativity metric presented below is intended as preliminary to (not a replacement of) an ‘indistinguishability test’. This paper relates ongoing research on a computational system with the aim of modelling a musical style. The motivation for the system is as follows. Cope (2005, pp. 87-95) describes a data-driven model that can be used to generate passages of music in the style of Johann Sebastian Bach’s chorale harmonisations. His model can be cast as a first-order Markov chain and we have replicated this aspect of his model, with some modifications (see Sect. 2 for details). Our method is applied to a database consisting of Frédéric Chopin’s mazurkas. This choice 1 See Pearce et al. (2002) on how computational modelling of musical styles constitutes one motivation for automating the compositional process.