Research | 2025 (Journal submitted)

Journal submitted to the Transactions of the International Society for Music Information Retrieval (co-author).
Conducted case studies on the BPS-motif dataset, analyzing motifs and visualizing results to assess structural patterns, style, and method effectiveness.

Abstract

Motif discovery in polyphonic symbolic music data is an important yet challenging task in music processing. In this paper, we propose a novel motif discovery method by combining the traditional rule-based repeated pattern discovery algorithms with a machine learning-based model which performs the task of motif note identification, i.e., identifying whether or not a note belongs to a motif. More specifically, the motif note identification model extracts motif notes for subsequent repeated pattern discovery. Removing non-motif notes can reduce the unwanted outputs in repeated pattern discovery and thereby improve the performance. With limited amount of training data, motif note identification can be implemented by fine-tuning a pre-trained model for symbolic music using pseudo-labels. The results demonstrate the feasibility of applying data-driven methods to motif discovery, even when labeled training data is scarce.

Authors: Jun-You Wang, Yu-Chia Kuo, and Li Su