Corpus and machine-learning research on music relies on high-quality datasets that often involve annotations and analysis by experts. In creating these annotations, it is important to ensure formal consistency and machine readability, but also a high musical expressivity. Annotation workflows can either rely on general music notation tools such as the Verovio Humdrum Viewer (Ricciardi, 2020; Rodin & Sapp, 2010) or MuseScore (Hentschel et al., 2021), or work with dedicated tools for creating specific types of annotations (Ericson et al., 2023; Giraud et al., 2018; Tomašević et al., 2021).
This paper presents a set of tools for working with protovoice analyses, a formalism introduced by Finkensiep & Rohrmeier (2021). In the protovoice model, the structure of a piece is described as a derivation, the execution trace of a generative process that produces the piece from a small set of operations. During this process, two types of generic relations are tracked: horizontal connections between notes that belong to the same "protovoice", and the vertical organisation of notes into "slices". At every point in the generation process, the current state of the piece is represented as a sequence of slices and transitions, where the slices contain notes and the transitions contain edges connecting these notes.