Daniel Field – MPhil.

This research applies the general logic and sequence used when teaching jazz improvisation to complete beginners as the basis for automatic improvisation. This project is building an algorithm that approaches the generation of jazz solos in the same way that a student of jazz would do, by treating the machine as a complete beginner and teaching it the fundamentals. It is hoped that the algorithm would be able to achieve a good passing grade in jazz improvisation on typical ‘beginner’ and ‘intermediate’ jazz song lists, alongside jazz students at second-year tertiary level, with high reliability and with melodic variety from one run to the next. The key differentiator in this study is the attempt to directly use the logic and sequence applied when teaching jazz improvisation to students, quite apart from any particular computing concept. While of course the study cannot escape the rules and constraints of computer science, nevertheless the focus will be on emulating taught human thinking to the greatest possible extent.

An approach to automatic music improvisation