Enforcing Meter in Finite-Length Markov Sequences

By Pierre Roy and François Pachet.

In the Proceedings of the 27th AAAI Conference on Artificial Intelligence (AAAI 2013), Bellevue, Washington, July 2013.

Read the paper (PDF)
Publisher record

About the paper

Sequences generated from a Markov model can reflect the style of a corpus. This paper addresses an additional requirement: the generated sequence must also fit a specified meter.

Abstract summary

The authors express metrical rules as cumulative costs over successive elements. Their meter constraint combines these rules with the transitions allowed by the Markov model and propagates restrictions on which partial sequences can lead to a valid result.

Related work

Enforcing Structure on Temporal Sequences: The Allen Constraint (CP 2016) connects event indexes to temporal positions, enabling structural constraints across sequences whose events have varying durations. PDF

See the Markov Constraints project, the application to song lyrics in Markov Constraints for Generating Lyrics with Style, and the later CP 2015 paper on exact sampling.

Back to Publications