Music Corpus Analysis Using Unwords

摘要

Discovering patterns reoccurring within a collection of pieces is a fundamental type of music corpus analysis. The inverted task is to discover patterns that are surprisingly infrequent in a corpus, including completely absent patterns or unwords. The key issue in mining unwords is evaluating whether a specific pattern is merely statistically absent from the corpus, or is prohibited in the style exemplified by the corpus. This paper describes a statistical method for evaluating unwords and applies it to reveal interesting unwords for counterpoint and chord sequences.