Including Latin American Music in MIR
Magdalena Fuentes
In recent years, Music Information Research (MIR) has made remarkable advances
in developing computational tools for analyzing, generating, and understanding
music. However, there is still a long way to go for MIR tools developed for
Latin American music. This talk will examine some examples of the unique
challenges that Latin American music presents to MIR research, ranging from data
scarcity to complex rhythmic, timbral, and improvisational qualities that defy
conventional methods. Specifically, I will highlight the need for culturally
sensitive datasets and models, addressing issues like polyphonic textures with
overlapping timbres, genre-specific rhythmic nuances, and timing (e.g., rubato
and microtiming). Additionally, I will discuss the social importance of adapting
MIR tools to underrepresented traditions, including the impact on cultural
preservation, educational resources, and creative exploration for Latin American
musicians and researchers. I will conclude with a look at open challenges in the
field, proposing new directions that include low-data learning approaches,
dataset curation, and light-weight models, with the aim to inspire future
research for a multicultural MIR landscape that truly reflects the diversity of
global music.
Overview
Program