Home // CONTENT 2025, The Seventeenth International Conference on Creative Content Technologies // View article

The Generation of Piano Music in the Style of Johannes Brahms Using Neural Network Architectures

Authors:
James Doherty
Brendan Tierney

Keywords: Intelligence; music generation; neural network architecture; Brahms.

Abstract:
Neural network architectures currently are only able to employ music generation tasks to similar levels of human composers when the music is at a basic compositional standard as they struggle with the complex motifs and harmonic structures of Western Classical Music. This study aims to determine if various data preprocessing and augmentation techniques can train a neural network model to generate pieces of piano music to a similar level of musicality and emotion as Romantic Period composer Johannes Brahms. Quantitative experimentation involving Music Information Retrieval was conducted, as well as a quantitative survey with respondents consisting of only professional musicians, composers, and conductors. Analysis of the results demonstrated that Transformer models using various attention mechanisms generated statically similar results to the original piano works of Brahms and that survey participants struggled to distinguish between the pieces generated by Brahms and the models. The results indicate that various data preprocessing and augmentation methods do have an impact on model accuracy resulting in the ability to generate longer sequences of music containing the composite motifs and harmonic structures of romantic period piano music.

Pages: 7 to 12

Copyright: Copyright (c) IARIA, 2025

Publication date: April 6, 2025

Published in: conference

ISSN: 2308-4162

ISBN: 978-1-68558-262-3

Location: Valencia, Spain

Dates: from April 6, 2025 to April 10, 2025