On the Characterization of Expressive Performance in Classical Music: First Results of the Con Espressione Game

A piece of music can be expressively performed, or interpreted, in a variety\nof ways. With the help of an online questionnaire, the Con Espressione Game, we\ncollected some 1,500 descriptions of expressive character relating to 45\nperformances of 9 excerpts from classical piano pieces, played by different\nfamous pianists. More specifically, listeners were asked to describe, using\nfreely chosen words (preferably: adjectives), how they perceive the expressive\ncharacter of the different performances. In this paper, we offer a first\naccount of this new data resource for expressive performance research, and\nprovide an exploratory analysis, addressing three main questions: (1) how\nsimilarly do different listeners describe a performance of a piece? (2) what\nare the main dimensions (or axes) for expressive character emerging from this?;\nand (3) how do measurable parameters of a performance (e.g., tempo, dynamics)\nand mid- and high-level features that can be predicted by machine learning\nmodels (e.g., articulation, arousal) relate to these expressive dimensions? The\ndataset that we publish along with this paper was enriched by adding\nhand-corrected score-to-performance alignments, as well as descriptive audio\nfeatures such as tempo and dynamics curves.\n