Medieval Chants & Open-Source Tools: A Pilot Case of Text Mining and Analysis

At present, data mining and visualization tools are gaining momentum at the research and university libraries across the world. Often, the tools are used as part of the libraries’ research services to educate patrons and provide new perspectives for scholars within the field of Digital Humanities. Equally important, the tools are also put to work on the libraries’ own collections. As a result, library staff may obtain a better understanding of the library’s holdings – and researchers in academia and other library users may suddenly find the library materials primed and accessible for complex textual studies. With the data mining and visualization techniques, comes a new way of thinking: For example, old materials may be approached in previously unexplored ways. Concerning the latter, we would like to present a pilot case involving one of our medieval liturgical manuscript fragments. This fragment is part of the binding of a 16th century first edition (print) of the famous Danish astronomer Tycho Brahe’s Epistolarvm astronomicarvm libri, Vranibvrgi 1596. The fragment which can be dated to c. 1250-1350 contains a small number of chants that were common in the liturgy of the Roman Catholic Church. That is why it has been possible in previous research at the University Library of Southern Denmark to identify the fragment’s chants using the Cantus Index – Online Catalogue for Mass and Office Chants (hosted by the University of Waterloo). Subsequently, in this pilot study, we are analyzing the selection of chants from the Cantus Database via the Voyant Tools environment, a web-based open source text mining and analysis tool. The Voyant Tools environment makes it possible to conduct studies of word frequency on the Cantus data which may or may not be attributed to specific genres of chant. With this pilot study we would like to show that the text analysis and visualization of the chant material illustrated in collocation graphs, in theory, may contribute to our understanding of the complex behavioral routines that formed the religious practice in the Middle Ages, as seen in the persistent use of recognizable textual patterns – with some morphological, syntactic and phraseological structures having greater frequency than others. The question is whether a full quantitative analysis of the presence of singular words, associated words, and strings of words / phrases can draw forth the Cantus corpus’ key components. Potentially, data mining and visualization methods may build bridges across multiple scientific disciplines, combining data science and statistics with linguistics, philology, genre analysis, hymnology, cultural history, and much more. We believe that future data mining developments, in combination with tools developed to apply optical character recognition (OCR) on handwritten texts, will make it possible to ask even more complex research questions and gain new insights into otherwise well-known library collections.