Using Probabilistic Parsers

to Support Salsa Music Composition. 2017/06/26

Autores

Brayan Rodríguez, Raúl Gutiérrez de Piñérez

Fecha de publicación

2017-06-19

Conferencia

International Conference on Mathematics and Computation in Music, pages 361-372

Editor

Springer, Cham

Abstract

Salsa is a long-established music genre. It has been used as a way to define, identify and express social beliefs. Due to the limited computational study of this genre, we consider relevant to identify and analyze the musical features of this music genre. Thus, we train a corpus with Grupo Niche songs for generating the production rules for an induced probabilistic context-free grammar through a probabilistic parser. In addition, we implement a web-based tool to support musical composition and generate automatic Salsa songs. In this work, we also compare three automatic songs using cross-validation on the corpus. We show the stability of the grammar because the precision of the generated songs compared to corpus’ songs is close to those that are not in the corpus.

PDF