Our study compares three approaches mode in medieval plainchant from the Cantus database. The eight modes are the central tonalities around which the repertoire is organized. The classical approach classifies melodies to modes using the final note and the range of the melody. The profile approach looks at pitch (class) profiles. Finally, the distributional approach represents melodies as weighted vectors of motif frequencies (tf–idf vectors). We try all sorts of motifs, or units, including three ‘natural’ units: notes that form so-called neumes, syllables or words. Overall, the distributional approach works best, and of all units that can be used with it, natural units work best. In fact, it works surprisingly well, even if we throw away the actual pitches and only use its contour. Could this mean that, just like a sentence is made by stringing together words, a chant melody is made together by concatenating small musical motifs?
You can read more about this project in chapter 4 of my PhD dissertation.
Bas Cornelissen, Willem Zuidema, and John Ashley Burgoyne (2020). Mode classification and natural units in plainchant. Proceedings of the 21st International Society for Music Information Retrieval Conference, Montréal, Canada, 2020
📖 ISMIR Paper (supplements) • 🎬 ISMIR Video
Cornelissen, B., Zuidema, W., & Burgoyne, J. A. (2021). Musical Modes as Statistical Modes: Classifying Modi in Gregorian Chant. Proceedings of the 6th International Conference on Analytical Approaches to World Music.
📖 AAWM Paper • 📜 AAWM Poster
Cornelissen, B., Zuidema, W., & Burgoyne, J. A. (2022). Understanding automatic mode classification in Western plainchant. 50th Medieval and Renaissance International Music Conference (MedRen 2022), Uppsala, Sweden.
🖼 MedRen22 Slides