Soprano voices in opera seria: dataset of musical features and metadata for 1,682 arias

This dataset contains symbolic musical features and metadata extracted from 1,682 soprano arias drawn from operas based on the five most widely set drammi per musica by Pietro Metastasio: Didone abbandonata (1724), Alessandro nell'Indie (1730), Artaserse (1730), Adriano in Siria (1732), and Demofoonte (1733). Each row corresponds to a single aria and includes: (1) musical features of the soprano part, extracted with the Python library musif from MusicXML scores (e.g. ambitus, interval size and type, rhythmic values, dynamics, and vocal presence, among others); and (2) metadata derived from the librettos, including the character, the character's dramatic gender (masculine/feminine), the name of the singer who premiered the aria (when known), and the singer's biological sex (male/female), inferred from their given name. This dataset provides the empirical basis for the article "Soprano voices in opera seria: a corpus-based inquiry into eighteenth-century vocal types", which applies supervised classification models (ridge and lasso logistic regression, random forest) to investigate how vocal writing in opera seria encodes characters' dramatic gender and singers' biological sex. The data were produced as part of the DIDONE project (ERC Grant Agreement No. 788986), funded by the European Research Council under the European Union's Horizon 2020 research and innovation programme.