Interactive music score completion

The autocompletion paradigm is applied to computer-aided music composition. Integrated in an interactive score editor prototype, a data-driven model is trained on an initial corpus defined by the composer. When the completion function is called, the model suggests possible continuations of musical phrases by superimposing them on the score. The user can then choose or avoid a quotation from the corpus, then modify it. Based on an interactive and online automatic learning, the modifications made, once validated, are reinserted in the learning corpus. Progressively, the completion suggestions are more and more personalized, according to the curatorial and editorial choices made by the composer. The score editor becomes a personal companion of the composer in classical tasks such as composition, harmonization, arrangement and orchestration