Abstract
In this paper, we compare four-part harmonization produced using two different machine learning models: a Bayesian network (BN) and a recurrent neural network (RNN). Four-part harmonization is widely known as a fundamental problem in harmonization, and various methods, especially based on probabilistic models such as a hidden Markov model, a weighted finite-state transducer, and a BN, have been proposed. Recently, a method using an RNN has also been proposed. In this paper, we conducted an experiment on four-part harmonization using the same data with both a BN and RNN and investigated the differences in the results between the models. The results show that these models have different tendencies. For example, the BN’s harmonies have less dissonance but especially the bass melodies are monotonous, while the RNN’s harmonies have more dissonance but especially bass melodies are smoother.
Keywords
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsReferences
Allan, M., Williams, C.K.I.: Harmonising chorales by probabilistic inference. In: Advances in Neural Information Processing Systems, vol. 17, pp. 25–31 (2005)
Bahdanau, D., Cho, K., Bengio, Y.: Neural machine translation by jointly learning to align and translate. In: IEEE International Conference on Learning Representations (2015). http://arxiv.org/pdf/1409.0473v6.pdf
Buys, J., Merwe, B.V.D.: Chorale harmonization with weighted finite-state transducers. In: Twenty-Third Annual Symposium of the Pattern Recognition Association of South Africa, pp. 95–101 (2012)
Ebcioglu, K.: An expert system for harmonizing chorales in the style of J. S. Bach. J. Log. Program. 8(1–2), 145–185 (1990)
Elman, J.L.: Finding structure in time. Cogn. Sci. 14(2), 179–211 (1990). http://doi.wiley.com/10.1207/s15516709cog1402_1
Gers, F.A., Schmidhuber, J.: Recurrent nets that time and count. In: Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks, vol. 3, pp. 189–194 (2000)
Hadjeres, G., Pachet, F.: DeepBach: a steerable model for Bach chorales generation. arXiv:1612.01010 [cs.AI], pp. 1–20 (2016)
Hild, H.: HARMONET: a neural net for harmonizing chorales in the style of J. S. Bach. In: Advances in Neural Information Processing 4, pp. 267–274 (1992)
Johnson, M., et al.: Google’s multilingual neural machine translation system: enabling zero-shot translation. arXiv preprint arXiv:1611.04558, pp. 1–16 (2016)
Kingma, D., Ba, J.: Adam: a method for stochastic optimization. In: International Conference on Learning Representations, December 2015. http://arxiv.org/abs/1412.6980
Mikolov, T., Yih, W.t., Zweig, G.: Linguistic regularities in continuous space word representations. In: Proceedings of NAACL-HLT 2013, pp. 746–751. Association for Computational Linguistics, Atlanta (2013)
Pachet, F., Roy, P.: Formulating constraint satisfaction problems on part-whole relations: the case of automatic musical harmonization. In: Workshop at European Conference on Artificial Intelligence 1998. Constraint Techniques for Artistic Applications (1998)
Phon-Amnuaisuk, S., Wiggins, G.A.: The four-part harmonisation problem: a comparison between genetic algorithms and a rule-based system. In: Proceedings of the AISB 1999 Symposium on Musical Creativity, pp. 28–34 (1999)
Phone-Amnuaisuk, S., Smaill, A., Wiggins, G.: Chorale harmonization: a view from a search control perspective. J. New Music. Res. 35, 279–305 (2006)
Raczynski, S., Fukayama, S., Vincent, E.: Melody harmonisation with interpolated probabilistic models. Project-Team METISS (Research Report N 8110), France (2012)
Sutskever, I., Vinyals, O., Le, V.Q.: Sequence to sequence learning with neural networks. In: Neural Information Processing Systems 2014 (NIPS 2014) (2014)
Suzuki, S., Kitahara, T.: Four-part harmonization using Bayesian networks: pros and cons of introducing chord nodes. J. New Music. Res. 43(3), 331–353 (2014). https://doi.org/10.1080/09298215.2014.911917
United Church of Christ in Japan Hymn Committee (ed.): Hymn: Hymn the second edition Tomoni Utaou. The Board of Publications The United of Church of Christ, Tokyo (1982). (in Japan)
Yi, L., Goldsmith, J.: Automatic generation of four-part harmony. In: Conference on Uncertainty in Artificial Intelligence- Applications Workshop, vol. 268 (2007)
Acknowledgments
This work has been supported by MEXT Grant-in-Aid (No. 16H05878, 16K16180, 16H01744, 26280089, 26240025, 16KT0136, 17H00749).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer Nature Switzerland AG
About this paper
Cite this paper
Yamada, T., Kitahara, T., Arie, H., Ogata, T. (2018). Four-Part Harmonization: Comparison of a Bayesian Network and a Recurrent Neural Network. In: Aramaki, M., Davies , M., Kronland-Martinet, R., Ystad, S. (eds) Music Technology with Swing. CMMR 2017. Lecture Notes in Computer Science(), vol 11265. Springer, Cham. https://doi.org/10.1007/978-3-030-01692-0_15
Download citation
DOI: https://doi.org/10.1007/978-3-030-01692-0_15
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-01691-3
Online ISBN: 978-3-030-01692-0
eBook Packages: Computer ScienceComputer Science (R0)