Music Ontology Specification

Abstract

The Music Ontology Specification provides main concepts and properties fo describing music (i.e. artists, albums, tracks, but also performances, arrangements, etc.) on the Semantic Web. This document contains a detailed description of the Music Ontology.

Music Ontology Classes Hierarchy

There are the schemes of the hierarchy of the Music Ontology classes. These schemes show the interaction between the Music Ontology classes and other ontologies classes.

RDF Document Examples

Examples of SPARQL queries against these RDF graphs

Background

The Music Ontology is an effort led by ZitGist LLC. and the Centre for Digital Music to express music-related information on the Semantic Web.

The Music Ontology is mainly influenced by:

• The FRBR Final Report
• The Event ontology
• The Timeline ontology
• The ABC ontology from the Harmony project
• The FOAF Project

More detailed are available on the Wiki publication page.

The Music Ontology Description

This specification serves as the Music Ontology "namespace document". As such it describes the Music Ontology and the terms (RDF classes and properties) that constitute it, so that Semantic Web applications can use those terms in a variety of RDF-compatible document formats and applications.

This document presents the Music Ontology as a Semantic Web vocabulary or Ontology. The Music Ontology is straightforward, pragmatic and designed to allow simultaneous deployment and extension, and is therefore intended for widescale use.

Evolution and Extension of the Music Ontology

Revisions and extensions of Music Ontology are conducted through edits to the namespace document, which by convention is published in the Web at the namespace URI.

The properties and types defined here provide some basic concepts for use in Music Ontology descriptions. Other vocabularies (e.g. the Dublin Core metadata elements for simple bibliographic description, FOAF, etc.) can also be mixed in with the Music Ontology terms, as can local extensions. The Music Ontology is designed to be extended, and modules may be added at a later date.

Music Ontology Modules

Time, and TimeLine, Event

The parts of the Music Ontology related to the production process of a particular piece of music (composition, performance, arrangement,...) as well as the parts dealing with time-related information are based on three external ontologies. The Music Ontology provides RDFS wrappers for the main classes, properties and individuals of these three ontologies.

The first ontology is: OWL-Time. Three terms of this ontology are used by the Music Ontology: TemporalThing, Instant and Interval.

However the kind of temporal information we may want to express goes a bit beyond OWL-Time, so we use an extension of it, developped in the Centre for Digital Music, Queen Mary, University of London: the TimeLine ontology. Indeed, we may want to express instants and intervals on multiple "timelines" (a timeline being a coherent backbone for temporal things): the one backing a particular audio file, the one behind an audio/video stream, or the physical one, backing a musical performance. Two classes of timelines are defined: PhysicalTimeLine (an instance of it being universaltimeline, which is the one on which we may address "the 13th of october, 2006"), and RelativeTimeLine (instances of this class may back audio signals, and we may address things such as "between 2 and 3 seconds" on them).

There is only one way of addressing temporal things per class of timeline. On a physical time line, a point is identified by a xsd:dateTime -- through the beginsAtDateTime property, and a duration by a xsd:duration -- through the durationXSD property. On a relative time line, a point P is identified by the duration of the interval [0,P], and this duration is identified by a xsd:duration -- through the beginsAtDuration property. A duration is identified by durationXSD.

In order to express knowledge about the production process of a piece of music, we use the Event ontology, also developped at the Centre for Digital Music. Events are seen as a way to arbitrary classify a space/time region. We have the possibiliy to attach to these events: agents (active participants to the event, like a performer, a sound engineer, ...), factors (passive things having a role in the event, like a musical instrument, a musical score, ...) and products (things produced by the event, such as a sound, a musical work, ...). A key feature of this ontology is also to allow "splitting" of events, through the sub_event transitive property. Using events, we may express: this musician was playing this instrument at this given time.

In the current version of the Music Ontology, the main sub-classes of Event are: Performance, Recording, Arrangement, Composition. However, given its abstract definition, we can describe lots of other things using this class: results of feature extraction, beat tracking, segmentation of songs...

Music Creation Workflow

In order to describe music-related events, we consider describing the workflow beginning with the creation of a musical work to its release on a particular record. This is our main description
paradigm, and was first used in the music production ontology developed at C4DM.

In the "easy" case (non-electronic music), We can describe this workflow within two boundaries: the simplest one and the most expressive one.

The simplest one consider the existence of 4 concepts within this workflow: MusicalWork (the musical work itself), Performance (the event corresponding to an actual performance
of the work), Signal (recording the performance as a signal), and MusicalManifestation (the release of this signal on a particular record).

The most expressive one consider the existence of 7 concepts: Composition (the event leading to the creation of a musical work), MusicalWork, Performance, Sound (the physical sound
produced by the performance), Recording (the event representing the transduction from a physical sound to a signal, through the use of a microphone), Signal, and MusicalManifestation.

Thus, we could imagine other ontologies plugged on top of MO, in order to represent the cognition of a sound (related to Sound), the different types of microphones that can be used (related to Recording), and so on.

In order to switch from the simplest workflow to the most expressive one, we define a single shortcut property: recorded_as, directly linking a Performance to a Signal. This property MUST be present in every case, in order to be able to do a simple query for accessing simple information.

The Music Ontology and Standards

It is important to understand that the Music Ontology as specified in this document is not a standard in the sense of ISO Standardisation, or that associated with W3C Process.

The Music Ontology depends heavily on W3C's standards work, specifically on XML, XML Namespaces, RDF, and OWL. All the Music Ontology documents must be well-formed RDF/XML documents.

This specification contributes an ontology, the "Music Ontology ", to the Semantic Web, specifying it using W3C's Resource Description Framework (RDF). As such, the Music Ontology adopts by reference both a syntax (using XML), a data model (RDF graphs) and a mathematically grounded definition for the rules that underpin the RDF design.

The Music Ontology and RDF

Why does the Music Ontology use RDF?

The Music Ontology is an application of the Resource Description Framework (RDF) because the subject area we're describing – music: artists, albums and tracks -- has so many competing requirements that a standalone format would not capture them or would lead to trying to describe these requirements in a number of incompatible formats. By using RDF, the Music Ontology gains a powerful extensibility mechanism, allowing Music-Ontology-based descriptions to be mixed with claims made in any other RDF vocabulary.

This is for that reason that the RDF document examples at the beginning of this document suggested to use the FOAF ontology to describe the Person, and its relations with other Person (artist or non-artist), and to use the Music Ontology to specify that that Person is also an Artist. That way, we can re-use other ontologies such as the Relationship ontology, in conjunction with the FOAF ontology, to describe different relationship between Artists.

The Music Ontology as an ontology cannot incorporate everything we might want to talk about that is related to music: artists (people), albums and tracks. Instead of covering all topics within the Music Ontology itself, we describe the basic topics and build into a larger framework - RDF - that allows us to take advantage of work elsewhere on more specific description vocabularies.

RDF provides the Music Ontology with a way to mix together different descriptive vocabularies in a consistent way. Vocabularies can be created by different communities and groups as appropriate and mixed together as required, without needing any centralized agreement on how terms from different vocabularies can be written down in XML or N3.

Check the Ontology namespaces referenced section to find some ontologies that ca be use in conjonction with the Music Ontology.

This mixing happens in two ways: firstly, RDF provides an underlying model of (typed) objects and their attributes or relationships. mo:Album is an example of a type of object (a "class"), while mo:compilation_of and mo:has_track are examples of a relationship and an attribute of an mo:Album; in RDF we call these "properties". Any vocabulary described in RDF shares this basic model, which is discernable in the syntax for RDF, and which removes one level of confusion in understanding a given vocabulary, making it simpler to comprehend and therefore reuse a vocabulary that you have not written yourself. This is the minimal self-documentation that RDF gives you.

Secondly, there are mechanisms for saying which RDF properties are connected to which classes, and how different classes are related to each other, using RDF Syntax and OWL. These can be quite general (all RDF properties by default come from an rdf:Resource for example) or very specific and precise (for example by using OWL constructs). This is another form of self-documentation, which allows you to connect different vocabularies together as you please.

In summary then, RDF is self-documenting in ways which enable the creation and combination of vocabularies in a devolved manner. This is particularly important for an ontology which describes communities, since online communities connect to many other domains of interest, which it would be impossible (as well as suboptimal) for a single group to describe adequately in non-geological time.

RDF is usually written using the XML or N3 syntaxes. If you want to process the data, you will need to use one of the many RDF toolkits available, such as Jena (Java) or Redland (C).

More information about RDF can be found in the RDF Primer.

The Music Ontology cross-reference: Classes, Properties and Individuals

The Music Ontology introduces the following classes, properties and individuals. See the Music Ontology namespace document in RDF/XML for more detail.

Classes, Properties and Individuals (full detail)