Analyzing Byte-Pair Encoding on Monophonic and Polyphonic Symbolic Music: A Focus on Musical Phrase Segmentation

TL;DR

This paper investigates how Byte-Pair Encoding (BPE) affects symbolic music processing, especially in phrase segmentation, revealing its dependence on instrumentation and its variable effectiveness in monophonic versus polyphonic music.

Contribution

It provides a qualitative analysis of BPE's behavior on symbolic music and evaluates its impact on musical phrase segmentation for different musical textures.

Findings

BPE training depends heavily on instrumentation.

BPE captures abstract musical content effectively.

BPE improves phrase segmentation in polyphonic music.

Abstract

Byte-Pair Encoding (BPE) is an algorithm commonly used in Natural Language Processing to build a vocabulary of subwords, which has been recently applied to symbolic music. Given that symbolic music can differ significantly from text, particularly with polyphony, we investigate how BPE behaves with different types of musical content. This study provides a qualitative analysis of BPE's behavior across various instrumentations and evaluates its impact on a musical phrase segmentation task for both monophonic and polyphonic music. Our findings show that the BPE training process is highly dependent on the instrumentation and that BPE "supertokens" succeed in capturing abstract musical content. In a musical phrase segmentation task, BPE notably improves performance in a polyphonic setting, but enhances performance in monophonic tunes only within a specific range of BPE merges.