Automatic stylistic composition of Bach chorales with deep LSTM

Liang, Feynman; Gotham, Mark; Johnson, Matthew; Shotton, Jamie

Abstract

This paper presents "BachBot": an end-to-end automatic composition system for composing and completing music in the style of Bach's chorales using a deep long short-term memory (LSTM) generative model. We propose a new sequential encoding scheme for polyphonic music and a model for both composition and harmonization which can be efficiently sampled without expensive Markov Chain Monte Carlo (MCMC). Analysis of the trained model provides evidence of neurons specializing without prior knowledge or explicit supervision to detect common music-theoretic concepts such as tonics, chords, and cadences. To assess BachBot's success, we conducted one of the largest musical discrimination tests on 2336 participants. Among the results, the proportion of responses correctly differentiating BachBot from Bach was only 1% better than random guessing.

Citation

Liang, F., Gotham, M., Johnson, M., & Shotton, J. (2017). Automatic stylistic composition of Bach chorales with deep LSTM. In S. J. Cunningham, Z. Duan, X. Hu, & D. Turnbull (Eds.), Proceedings of the 18th International Society for Music Information Retrieval Conference, ISMIR 2017 (449-456)

• Conference Name: ISMIR 2017: 18th International Society for Music Information Retrieval Conference
• Conference Location: Suzhou
• Start Date: Oct 23, 2017
• End Date: Oct 27, 2017
• Publication Date: 2017
• Deposit Date: Feb 29, 2024
• Pages: 449-456
• Book Title: Proceedings of the 18th International Society for Music Information Retrieval Conference, ISMIR 2017
• ISBN: 9789811151798
• Public URL: https://durham-repository.worktribe.com/output/2273232
• Publisher URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85055992374&partnerID=40&md5=eebbfebf781bab11de68bc83996ad11f