Exploring Softly Masked Language Modelling for Controllable Symbolic Music Generation
This is an incremental approach for researchers in AI music generation, focusing on constrained symbolic music tasks.
The paper tackles controllable symbolic music generation by applying Softly Masked Language Modelling (SMLM), a generalization of masked language modelling, to allow elements to be known, unknown, or partly known, and demonstrates results using a transformer encoder architecture.
This document presents some early explorations of applying Softly Masked Language Modelling (SMLM) to symbolic music generation. SMLM can be seen as a generalisation of masked language modelling (MLM), where instead of each element of the input set being either known or unknown, each element can be known, unknown or partly known. We demonstrate some results of applying SMLM to constrained symbolic music generation using a transformer encoder architecture. Several audio examples are available at https://erl-j.github.io/smlm-web-supplement/