摘要：The first goal of this paper is to
introduce musicologists and music theorists to
the benefits offered by state-of-the-art pattern
recognition techniques. The second goal is to
provide them with a computer-based framework
that can be used to study large and diverse
collections of music for the purposes of
empirically developing, exploring and validating
theoretical models. The software presented
in this paper implements techniques
from the fields of machine learning, pattern
recognition and data mining applied to and
considered from the perspectives of music
theory and musicology. An important priority
underpinning the software presented here is
the ability to apply it to a much wider range of
art, folk and popular musics of the world than
is possible using the types of computer-based
approaches traditionally used in music research.
The tools and techniques presented
here will thus enable exploratory research that
can aid in the formation and validation of
theoretical models for types of music for
which such models have been elusive to date.
These tools will also allow research on forming
theoretical links spanning types of music
that have traditionally been studied as distinct
groups. A particular emphasis is placed on the
importance of performing studies involving
many pieces of music, rather than just a few
compositions that may not in fact be truly
representative of the overall corpus under
consideration.
关键词：Music information retrieval, machine
learning, musicology
