A Framework for Comparing SATB Harmonization Approaches
| Tipo de publicação: | Inproceedings |
| Citação: | |
| Publication status: | Accepted |
| Booktitle: | ACM GECCO 2026 (GECCO ’26 Companion). |
| Ano: | 2026 |
| Publisher: | ACM |
| Resumo: | Four-part harmonization (SATB) is a complex problem, characterized by a large search space and numerous harmonic constraints. Despite the many approaches proposed, direct comparison between results remains difficult due to the lack of a standardized evaluation framework. In this work, we propose a practical, reproducible approach based on the Sharpmony platform that integrates a rule-based correction system and an evolutionary environment for harmonizing melodies. The proposed solution enables the evaluation of different approaches using uniform criteria, thanks to its compatibility with standard score formats. As a case study, we compare a classical evolutionary approach with a modern evolutionary system that employs local search, directed mutation, and leverages data from Sharpmony students. The results show significant improvements, with error-free harmonizations and average times of about 14 minutes. |
| Palavras-chave: | 4-part harmonization, Computational Creativity, Directed Mutation, Genetic Algorithm, Local Search |
| Autores | |
| Adicionado por: | [] |
| Total mark: | 0 |
|
Anexos
|
|
|
Notas
|
|
|
|
|
|
Tópicos
|
|
|
|
|