An Overview of Cross-Cultural Music Corpus Studies

Abstract

The past few decades have seen a rapid increase in the availability and use of large music corpora. However, most music corpus studies remain limited to Western music, limiting our ability to understand the diversity and unity of human music throughout the world. I argue for the potential of cross-cultural corpus studies to contribute to comparative musicological studies in domains including music classification, evolution, universals, and human history. I highlight and discuss a number of important cross-cultural corpora, including the Berlin Phonogramm-Archiv, CompMusic Project, Essen Folksong Collection, Cantometrics Project, Garland Encyclopedia of World Music, Natural History of Song, and Deep History of Music projects. In the process, I discuss the pros and cons of music notation vs. recordings, automatic vs. manual analysis, regional vs. global analysis, and associated challenges regarding choosing appropriate analysis methods that can allow meaningful comparison across cultures. I argue for the need for bigger and better global music corpora and more emphasis on integration within and beyond academia, including to domains such as the music industry and cultural heritage organizations.

Modeling enculturated bias in entrainment to rhythmic patterns

“…In this work we derived precise phase-based expectations from discretized (symbolic) interval patterns, as notated in sheet music, allowing us to derive highly stereotyped representations of rhythmic patterns. This approach is limited in that relatively few non-Western music corpora are available in a computational notated form [ 22 , 23 ], and typically a Western-trained musical expert is required to notate and/or validate such corpora, which might introduce biases. However, there is no reason why pPIPPET templates need to be configured with expected events related by perfectly discretized intervals.…”

“…These probabilistic models are reliant on symbolic sequence-learning algorithms [ 21 ], learning from (symbolic) score-based musical corpora. While there is a long tradition of cross-cultural music corpus studies [ 22 , 23 ], these models are limited by a coarse representation of rhythm, and cannot explain how people dynamically map—during entrainment—between auditory rhythms represented in a continuous space and discretized (symbolic) representations. As these models cannot serve as models of real-time behavior, there is considerable room for improvement in our computational understanding of how probabilistic temporal expectations continuously bias entrainment.…”

The Global Jukebox: A public database of performing arts and culture

“…For a fuller description of key differences between these three global datasets and other regional datasets, cf. [ 48 ]. (See also section 2.1 for more details on the Cantometrics sampling methodology).…”

Comparative musicology: The science of the world’s music

“…Many of these large collections of thousands of melodies have recently been digitized and analyzed using modern computational methods, as we will discuss in section 2.6. (Panteli et al, 2018;Savage, 2022;Savage, Passmore, et al, 2022;Schaffrath, 1995;Tierney et al, 2011;van Kranenburg et al, 2013van Kranenburg et al, , 2014.…”

“…Exceptions to this Western-centric tendency have increasingly begun to emerge (for review, see) (Gómez et al, 2013;Jacoby et al, 2020;Savage, 2022;Stevens, 2012;Thompson et al, 2019). Some of this may be spurred by the publication of Henrich et al's (Henrich et al, 2010) influential article highlighting the dangers of relying on research limited to "WEIRD" (Western, Educated, Industrial, Rich, Democratic) societies.…”