Taking Digital Humanities to Guatemala, a Case Study in the Preservation of Colonial Musical Heritage

The goal of this project is the preservation and dissemination of Guatemala’s colonial musical heritage by applying music information retrieval tools to a group of Guatemalan manuscript sources while maintaining the original sources in their homeland. This group of sources is written in mensural notation, a music notation style used in Europe throughout the Late Middle Ages and the Renaissance. The only known Guatemalan mensural repertoire is contained in two extant manuscript sources: (i) the Santa Eulalia manuscripts (US-BLI), copied in the late sixteenth century; and (ii) the Metropolitan Cathedral choirbooks (GCA-Gc), copied during the sixteenth to eighteenth centuries. The Santa Eulalia manuscripts, a collection of fifteen books originally found in a north-west region of Guatemala, are no longer in the country. Therefore, the set of choirbooks kept in the Metropolitan Cathedral of Guatemala City contains the only mensural repertoire that remains in the country, and so it is of utmost importance to treasure and preserve them for future access. In this paper, we will explain the different steps that will be taken to guarantee the preservation and accessibility of the Guatemalan Cathedral choirbooks.
The musical heritage from the colonial period of Latin American countries is not well known around the world. Even though colonial music is rooted in the early Western music traditions (i.e., Medieval, Renaissance, and Baroque music), little is known about how these traditions and repertoires evolved in the Spanish colonies of the Americas. The lack of high-quality digitization of music documents from these countries and era limits access to people able to visit the site that has the original sources. Furthermore, the use of old music notation restricts performance of the music to the knowledgeable experts. In this paper, we will use recent music encoding technologies to address the accessibility issues of the mensural colonial repertoire of the Americas.
The barriers to accessibility in the body of Guatemalan mensural music can be summarized into: (i) the lack of high-quality digital images, (ii) the notation style, and (iii) its layout. Firstly, high-quality images are necessary for the music to be readable. While low-quality images can still be read by humans, high-quality images are necessary to encode the music in an automated way using techniques such as Optical Music Recognition (OMR). Furthermore, high-quality images allow for digital restoration, enhancing details nearly invisible in the original sources. Secondly, mensural notation differs significantly from our current music notation system in that the duration of notes is not absolute, but rather context-dependent. This characteristic makes it impossible for non-experts to accurately read this music. Lastly, the parts corresponding to each voice (i.e., soprano, alto, tenor, bass) are written in separate areas of the page instead of being vertically aligned as in a modern score format. Together the context-dependent nature of the notation and the separate-parts layout hinder the appreciation of the polyphonic texture of the music. It is only until musicians acquainted with the notation sing the various parts together or until an expert transcribes the music into a modern score that these textures can be really perceived and enjoyed. The layout of the original music deters its study even for experts, because it is hard to visualize the vertical relationships between notes sung simultaneously in two different voices, something that only becomes clear when the music is presented in score format.
In this paper, we will present the step-by-step process that will unravel the accessibility barriers of this repertoire, resulting in the digitization and encoding of the repertoire as musical scores in a machine-readable file format. This procedure involves the use of the following tools:

A Do-It-Yourself (DIY) book scanner. Given their antiquity, these books cannot be digitized face-down using a common flatbed scanner. On the contrary, they need to be digitized from the top using a “book scanner” and, ideally, using a v-shaped book cradle to lower the stress on the book spine. Currently, there is no book scanner in Guatemala that has the appropriate dimensions to digitize these books. Since buying a large book scanner is out of our possibilities, we will build our own by emulating their book cradle, lights, and camera configuration.

Optical Music Recognition (OMR) software applications. Just as Optical Character Recognition (OCR) allows the machine to read the characters written in the page, OMR systems read the music symbols from a music document. OMR software that will be used to read and encode the sources include:
Pixel.js (Saleh et al., 2017),
Gamera (Droettboom et al., 2003) and a machine learning model developed by Pacha and Calvo-Zaragoza (2018) that was trained with Spanish mensural music. The resulting data from the OMR process will be encoded in the MEI (Music Encoding Initiative) format (Roland, 2002).

The mensural notation
Scoring-Up Tool (Thomae, 2017). This set of scripts convert the part representation of the score into a modern rendering, providing a more conventional and modern representation to musicians and lay people.

We hope that the digitization and encoding of the music documents with the aforementioned techniques will allow for the preservation and dissemination of Guatemalan mensural music—a historic musical heritage that otherwise may be lost, forgotten, or damaged. The result should facilitate its dissemination, study by musicologists, and appreciation by the general public, especially for repertoires that are virtually unknown abroad. Furthermore, we expect this research to be used as a model for the digitization of the mensural repertoire of other countries that were once part of the colonial past, such as the Spanish Colonial Empire.

Music Manuscripts

US-BLI: Bloomington, Indiana University, Lilly Library MS 1–15.
GCA-Gc: Guatemala City, Catedral, Archivo Capitular 1–4.

Bibliography

Droettboom, M., MacMillan, K. and Fujinaga, I. (2003). The Gamera framework for building custom recognition systems.
Symposium on Document Image Understanding Technologies. pp. 275–86.

Pacha, A. and Calvo-Zaragoza, J. (2018). Optical Music Recognition in Mensural Notation with Region-Based Convolutional Neural Networks.
Proceedings of the 19th International Society for Music Information Retrieval Conference (ISMIR). Paris, France, pp. 23–27.

Roland, P. (2002). The Music Encoding Initiative (MEI).
Proceedings of the 1st International Conference on Musical Application Using XML, vol. 1060. pp. 55–59.

Saleh, Z., Zhang, K., Calvo-Zaragoza, J., Vigliensoni, G. and Fujinaga, I. (2017). Pixel.js: Web-based Pixel Classification Correction Platform for Ground Truth Creation.
Proceedings of the 12th IAPR International Conference on Graphics Recognition. Kyoto, Japan: Springer LNCS.

Thomae, M. (2017). Automatic Scoring up of Mensural Music Using Perfect Mensurations, 1330-1550 Montreal: McGill University Master’s Thesis.