Musical Grammars

This page contains data intended to accompany a paper entitled ``Statistical Learning of Harmonic Movement'', published in the Journal of New Music Research (full publication details forthcoming).

The files linked below contain information defining Finite State Transition Networks expressing information learned by computer about harmonic progression in 17th Century Sarabande forms.

The data from which these grammars were learned is divided into 9 sections, each of which has its own grammar in the table below. For more details, see the paper. The data for each grammar is expressed as a probabilistic Finite State Transition Network which can be loaded directly into the Prolog programming language. The format is as follows:

where Start is the start state of the arc (represented as a tripartite object, the details of which are explained in the paper), End is the end state of the arc (represented in the same way), Symbol is the symbol which is read or generated by the arc when the FSTN is used, and Probability is the probability of this symbol appearing in this context in the input data - it can be used for stochastic generation of music. Symbol represents chords in conventional I, IV, i, iv, etc) notation, except where a natural represetation is not possible in this form. Where this is the case, see the paper for details of the representation.

Please note that this information is meant to be used in conjunction with the paper referred to above.

The Data

This information may be freely downloaded and used for research and other non-commercial purposes. It is a condition of download that, if resulting work is published in any form or fashion, an acknowledgement of the source should be made.

(c) 1999, D. Ponsford, G. Wiggins, C. Mellish, M. Walker