The Closure-based Cueing Model: Cognitively-Inspired Learning and Generation of Musical Sequences

Publication Type:

Conference Paper


SMC Conference 2011 (2011)


In this paper we outline the Closure-based Cueing Model (CbCM), an algorithm for learning hierarchical musical structure from symbolic inputs. Inspired by perceptual and cognitive notions of grouping, cueing, and chunking, the model represents the schematic and invariant properties of musical patterns, in addition to learning explicit musical representations. Because the learned structure encodes the formal relationships between hierarchically related musical segments, as well as the within-segment transitions, it can be used for the generation of new musical material following principles of recombinance. The model is applied to learning melodic sequences, and is shown to generalize perceptual contour and invariance. We outline a few methods for generation from the CbCM, and demonstrate a particular method for generating ranked lists of plausible continuations from a given musical context.

smc2011_submission_189.pdf719.12 KB
SMC paper: