An Expert System Devoted to Automated Music Identification and Recognition - Intelligent Music Information Systems: Tools and Methodologies - page 82

IntroductIon

have some knowledge of the meaning of the
“identifiers” included in Table 1. The interaction
between users and experts improves, in succes-
sive steps—beginning with knowledge. This ES
can be useful for cult music fans but, mainly, for
both learning and teaching history of music and
of composition (Fernández de Sevilla & Hilera,
2001; Fernández de Sevilla & Gutiérrez de Mesa,
2002; Fernández de Sevilla, 2001; Fernández
de Sevilla, 2002; Fernández de Sevilla, 2003;
Fernández de Sevilla & Fernández del Castillo,
2002; Fernández de Sevilla, 2005; Balaban &
Ebcioglu, 1992).
It has to be emphasized at this point, that our
purpose is to suggest and adapt a computational
method illustrated with the corresponding imple-
mentation, rather than to present a selection of
identifiers perfectly describing the characters of
the different musical periods considered. If other
experts judged that the identifiers should be chosen
another way, only the input data to the system had
to be altered, but not the rest of the system, that
would be kept unchanged (obviously, the results
given by the system could change). If a larger
number of identifiers had to be considered, there
would be no problem in dividing the system into
subsystems, and to gather the partial results of
those subsystems in a new one that would provide
the final result.
Any ES (once constructed) must go through
two processes: verification and validation. To
verify the ES means to check that, syntactically,
it does not contain inconsistencies. To validate
the ES means that, after it has been verified, its
outputs are in agreement with the “real” world, in
our case, that the ES outputs reliable music styles
identifications. Although the process of validation
may improve our ES, the ES is very strong regard-
ing verification because its IE is based on a sound
computational method: Gröbner Bases.
The objectives of the chapter are the follow-
ing:

This work describes a logic and computer alge-
bra-based Knowledge Based Expert System, to
be hereinafter denoted as “ES”, for cult music
styles identification of the period that lasts from
the XVII century to the beginnings of the XX
century. By introducing into the ES some mu-
sic-related identifiers, that the user has selected
while or after analyzing a score, the ES identifies
both its style and, consequently, the epoch when
it was composed.
Expert Systems have at least a “knowledge
base”, hereinafter denoted as “KB”, which col-
lects and represents formally expert knowledge,
an “inference engine”, hereinafter denoted as
“IE”, which extracts new knowledge from the
KB, and a graphic user interface, hereinafter
denoted as “GUI”. An explanation of what KB,
IE and GUI are in any ES shall be the topic of the
section “Generalities about Rule-Based Expert
Systems”.
In order to build the KB of the ES, a list of
characteristics, which are called in this chapter
“identifiers”, of the music styles of the epoch
under consideration must be previously defined.
The list used in this chapter is the list in Table
1. The user, during or after analyzing a score,
introduces into the ES the identifiers that the user
has selected from Table 1: the ES is supposed to
output the identification and recognition of the
score style.
A simple method builds, in three “levels” or
steps, the formulae that constitute the KB, called
“production rules” (the necessary definitions and
concepts will presented in their adequate places
in the chapter). For this purpose, tables that as-
sign music related characteristics of an epoch to
different consequents of the production rules are
designed. This leads to the construction of a com-
pact and coherent system of production rules.
The nonexpert users (an example of this type
of user is a first year student of a conservatory
or a cult music fan) of our ES are supposed to