ABSTRACT
An unfinished work of art is eternally provocative. Unfinished musical works, particularly those by important composers, are rarely left incomplete: midwife musicians learn to recreate the style of a composer in order to finish them. The recent availability of musical texts in machine-readable form allows us to apply methods of statistics, machine learning, and artificial intelligence to the formerly exclusive domain of historically minded composers and musicologists. The scientific approach, while it may begin with the same goal as traditional inquiry-i.e., How do we finish the piece?—leads to new questions and points in new directions.
