Major revision (1.01) of the Music Ontology
Frederick Giasson
I just published a draft of the next revision of the ontology. Many things
changed accordingly to all the insightful comments you posted on this mailing list.
The draft is is available here:
Please tell me what you think about that draft, and if you have any ideas
related to it.
I will write the documentation of the revision as soon as people seems to agree
that it is ok.
Take care,
Salutations,
Fred
Kingsley Idehen
Do you have a script for Indexing this Data in Virtuoso based on a
PostgreSQL DSN?
--
Regards,
Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen
President & CEO
OpenLink Software Web: http://www.openlinksw.com
Frederick Giasson
> Do you have a script for Indexing this Data in Virtuoso based on a
> PostgreSQL DSN?
Not now since the ontology is not froze for a first version (hope this revision
go in that direction). But as soon as people seems to agree that we have a good
starting base, I will develop such a script, and I will post it on this mailing
list with an explanation of how it works in case people want to import the
musicbrainz db for their own and want to query it using Virtuoso's triple store
using its SPARQL parser.
I will keep the mailing list updated vis-a-vis that in the next week, I hope.
Take care,
Fred
Ivan
first of all, my apologies if the only thing I do is to raise problems
without offering a solution right away:-)
I am looking at the ontology purely from a classical music point of
view. *If* (and that is an 'if') this ontology should also cover
classical music, then I see some missing features. Some come to my mind
based on my recent experience of having bought an iPod and putting a
large portion of my CD collection on it...
1. 'Album' you define that as "One or more track issued together.". I
guess the intention is that, if I buy a CD, *that* is an Album, right?
But if I buy a CD that contains two piano concertos of Bach, what is
the album? Is it the whole CD, or do we have something else that
contains two albums? Do we need an intermediate granularity for a
musical piece that contains a certain number of tracks and an album is
(or may be) a collection of those musical pieces? There is of course
the term 'Work' that you use, that may be a musical piece, but I would
expect this to appear in this ontology
2. I see that you have now the notion of 'key' which, I presume, would
cover things like 'D major'. But its domain is a track. However, in
most of the cases (not always...) a musical piece has a key, and it is
certainly that piece that is traditionally characterized by its key.
3. A large portion of the classical musical pieces have an
identification number. Op. 45, K456, etc. The denotation differ from
one composer to the other and, alas!, there may be several, competing
denotation systems for each composer. However, there should be some
slot here for a musical piece, that is for sure...
4. You do not seem to have a composer, only an artist. You do have a
'composed by', though. However, the perception is that the ontology
puts the emphasis *on the artist* (say, Eric Clapton) and not on the
composer. In the classical music world this is really the opposite. The
work itself is associated to Bach, that is the major person there, and
can be performed by one or several artists. This dual notion should be
reflected in the ontology.
5. I see that you refer to instruments, and that is fine. I hope that
this means I can list an artist with his/her name and the instruments
he/she plays. Note that one person may play several instruments; that
we need the term 'conductor', 'tenor', 'bass', etc, to properly
catalogue a musical piece...
I am sure there are more, and would be pleased to see the experiences
of others on that. Of course, we may decide that *this* ontology should
*not* cover classical music, recognizing that it is a very different
world, different traditions, etc, and we would have to do a separate
classical music ontology. This may very well be the way to go; I do not
really know.... But is is worth discussing!
Thanks for all this...
Ivan
Frederick Giasson
> first of all, my apologies if the only thing I do is to raise problems
> without offering a solution right away:-)
Hehehe, no problem, as long as I get feedbacks to improve the ontology :)
> I am looking at the ontology purely from a classical music point of
> view. *If* (and that is an 'if') this ontology should also cover
> classical music, then I see some missing features. Some come to my mind
> based on my recent experience of having bought an iPod and putting a
> large portion of my CD collection on it...
Yeah sure it should. In fact, I would like to develop that ontology in such a
way that people could use and extend it as they like. So developing the
foundation of a music ontology, and then people could create new genres,
properties, etc. (something like FOAF (and its relation with BIO, REL, etc. but
for music instead of people).
> 1. 'Album' you define that as "One or more track issued together.". I
> guess the intention is that, if I buy a CD, *that* is an Album, right?
An "Album" can be on a CD, streamed over the internet, etc. In fact, it is an
aggregation of tracks.
> But if I buy a CD that contains two piano concertos of Bach, what is
> the album? Is it the whole CD, or do we have something else that
Could it be a mo:Compilation?
> contains two albums? Do we need an intermediate granularity for a
> musical piece that contains a certain number of tracks and an album is
> (or may be) a collection of those musical pieces? There is of course
> the term 'Work' that you use, that may be a musical piece, but I would
> expect this to appear in this ontology
Hummm, would you have an example of intermediate granularity and its interaction
with the other classes?
An album (as we know them on shelves) can be a musical work. A symphony is a
musical work, a concerto is a musical work, etc.
But I agree that this is ambiguous and that we would need to clean these classes
(and their definition).
Would you have an idea?
> 2. I see that you have now the notion of 'key' which, I presume, would
> cover things like 'D major'. But its domain is a track. However, in
Exact
> most of the cases (not always...) a musical piece has a key, and it is
> certainly that piece that is traditionally characterized by its key.
What you mean by a "musical piece" here? Personally I see a "track" as a "mucial
piece".
In fact case, should I rename "track" for "Musical Piece"?
> 3. A large portion of the classical musical pieces have an
> identification number. Op. 45, K456, etc. The denotation differ from
> one composer to the other and, alas!, there may be several, competing
> denotation systems for each composer. However, there should be some
> slot here for a musical piece, that is for sure...
Each sure. So, how would you name and describe such a property? Would you have
some RDFS for me :)
> 4. You do not seem to have a composer, only an artist. You do have a
> 'composed by', though. However, the perception is that the ontology
> puts the emphasis *on the artist* (say, Eric Clapton) and not on the
> composer. In the classical music world this is really the opposite. The
> work itself is associated to Bach, that is the major person there, and
> can be performed by one or several artists. This dual notion should be
> reflected in the ontology.
Yeah I agree, but:
There are artists that perform (performers)
and
There are artists that compose (composers)
So, right now, for the ontology, both composer and performer are artists.
No?
> 5. I see that you refer to instruments, and that is fine. I hope that
> this means I can list an artist with his/her name and the instruments
> he/she plays. Note that one person may play several instruments; that
> we need the term 'conductor', 'tenor', 'bass', etc, to properly
> catalogue a musical piece...
Yeah sure you can. Check the domain of mo:instrument.
> I am sure there are more, and would be pleased to see the experiences
> of others on that. Of course, we may decide that *this* ontology should
Yeah me too ;)
> *not* cover classical music, recognizing that it is a very different
> world, different traditions, etc, and we would have to do a separate
> classical music ontology. This may very well be the way to go; I do not
> really know.... But is is worth discussing!
Definitely. But as I said above, the best would be that this Music Ontology
become the basement for the development of other, more specialized, musical
ontologies.
> Thanks for all this...
Thanks to you for these observations!
Take care,
Fred
Ivan
Let me give you one specific example for a CD, this may clarify most of
the issues I have (I hope:-).
I have one specific CD. The CD itself does *not* have a specific name,
but it contains four pieces of Schumann: Kreisleriana, Nachtstücke,
Gesänge der Frühe, and Geister-Variationen, played on the piano by
András Schiff.
I am making up some sort of a phony RDF here, I have not checked all
details with the current ontology and namespaces, sorry. But what I
would prefer to see is something along the lines of:
[
a mo:Album;
mo:consistsOf (_:p1, _:p2, _:p3, _:p4 )
]
_:p1 a MusicalWork;
# maybe a seaparate type here for a distinct piece?
mo:genre mo:Classical;
dc:title "Kreiserliana";
dc:creator: # or mo:composed?
[
a foaf:Person;
a mo:Composer;
foaf:name "Robert Schumann";
....
];
# I am afraid the 'opus' is so unspecified in practice, that we have
to stay by a string...
mo:opus "Op. 133";
# the key is not specified
mo:consistsOf (_:tr10,_:tr11,...,_:tr17);
mo:performed
[
a foaf:Person;
a mo:SoloMusicArtist;
foaf:name "András Schiff";
mo:plays mo:Piano;
].
_:tr10 a mo:Track;
dc:title "Äußerst bewegt";
# not much more in this case but there may be more. Maybe tempo in
some case?
_:tr10 a mo:Track;
dc:title "Sehr innig und nich zu rasch";
# not much more in this case but there may be more. Maybe tempo in
some case?
etc.
I call it now MusicalWork in reference to your blog, whether this is
the right name or not, I do not know. The important is the strict
differentiation between the album and the individual pieces which may
in turn consists of possibly several tracks.
Some more answers below:
On Jan 10, 8:43 pm, Frederick Giasson <f...@fgiasson.com> wrote:
> Hi Ivan!
>
> > first of all, my apologies if the only thing I do is to raise problems
> > without offering a solution right away:-)Hehehe, no problem, as long as I get feedbacks to improve the ontology :)
>
> > I am looking at the ontology purely from a classical music point of
> > view. *If* (and that is an 'if') this ontology should also cover
> > classical music, then I see some missing features. Some come to my mind
> > based on my recent experience of having bought an iPod and putting a
> > large portion of my CD collection on it...
> Yeah sure it should. In fact, I would like to develop that ontology in such a
> way that people could use and extend it as they like. So developing the
> foundation of a music ontology, and then people could create new genres,
> properties, etc. (something like FOAF (and its relation with BIO, REL, etc. but
> for music instead of people).
>
> > 1. 'Album' you define that as "One or more track issued together.". I
> > guess the intention is that, if I buy a CD, *that* is an Album, right?
> An "Album" can be on a CD, streamed over the internet, etc. In fact, it is an
> aggregation of tracks.
>
> > But if I buy a CD that contains two piano concertos of Bach, what is
> > the album? Is it the whole CD, or do we have something else that
> Could it be a mo:Compilation?
For me, a compilation is when I publish a CD with, say, a collection of
Beatles songs that were originally published for the first time in
various albums. This is more or less meaningless for classical
music....
>
> > contains two albums? Do we need an intermediate granularity for a
> > musical piece that contains a certain number of tracks and an album is
> > (or may be) a collection of those musical pieces? There is of course
> > the term 'Work' that you use, that may be a musical piece, but I would
> > expect this to appear in this ontology
> Hummm, would you have an example of intermediate granularity and its interaction
> with the other classes?
>
See above
> An album (as we know them on shelves) can be a musical work. A symphony is a
> musical work, a concerto is a musical work, etc.
>
> But I agree that this is ambiguous and that we would need to clean these classes
> (and their definition).
>
> Would you have an idea?
>
Yes, I think (as I jotted above), we have a MusicalWork as a
fundamental unit, and an Album (or a CD) *may* be the collection of
such MusicalWorks. I am not sure the word 'Album' is the right one, but
that might be etimology only. I guess my main point is that the focal
point is a MusicalWork (and this is what you put into your blog, in
fact), and the term 'Album' is much much more vague. Of course, in the
case of very classical pop music (sorry, I am an old guy, so my
examples are antiquitated...) the 'Sergent Pepper' is a Beatles Album
that may be considered as a MusicalWork but, there again, each track on
that Album may be a separate MusicalWork. My point is that these two
terms seem to have a different meaning in rock music.
> > 2. I see that you have now the notion of 'key' which, I presume, would
> > cover things like 'D major'. But its domain is a track. However, in
> Exact
>
> > most of the cases (not always...) a musical piece has a key, and it is
> > certainly that piece that is traditionally characterized by its key.
> What you mean by a "musical piece" here? Personally I see a "track" as a "mucial
> piece".
>
Forget. MusicalWork as I used it above, may be better.
> In fact case, should I rename "track" for "Musical Piece"?
>
> > 3. A large portion of the classical musical pieces have an
> > identification number. Op. 45, K456, etc. The denotation differ from
> > one composer to the other and, alas!, there may be several, competing
> > denotation systems for each composer. However, there should be some
> > slot here for a musical piece, that is for sure...
> Each sure. So, how would you name and describe such a property? Would you have
> some RDFS for me :)
>
As I commented above: this is *really* messy. It is very often a set of
letters and a number, something like 'Op. 45'. Sometimes it says 'Op.
posth.' for a particular piece.
For the time being I've put a string in my example. Maybe some general
structure containing a string plus an integer plus some other qualifier
may work out; I am happy to check with my collection if it comes to
that.
> > 4. You do not seem to have a composer, only an artist. You do have a
> > 'composed by', though. However, the perception is that the ontology
> > puts the emphasis *on the artist* (say, Eric Clapton) and not on the
> > composer. In the classical music world this is really the opposite. The
> > work itself is associated to Bach, that is the major person there, and
> > can be performed by one or several artists. This dual notion should be
> > reflected in the ontology.
> Yeah I agree, but:
>
> There are artists that perform (performers)
>
> and
>
> There are artists that compose (composers)
>
> So, right now, for the ontology, both composer and performer are artists.
>
> No?
>
Ah. So what you say is:
Artist a foaf:Person.
Composer rdfs:subClass Artist.
Performer rdfs:subClass Artist.
is that what you mean? That could work. If we use OWL, we could think
about further characterizations (is each artist a composer or a
performer or both?), but this can be done later...
> > 5. I see that you refer to instruments, and that is fine. I hope that
> > this means I can list an artist with his/her name and the instruments
> > he/she plays. Note that one person may play several instruments; that
> > we need the term 'conductor', 'tenor', 'bass', etc, to properly
> > catalogue a musical piece...
> Yeah sure you can. Check the domain of mo:instrument.
>
Great!
> > I am sure there are more, and would be pleased to see the experiences
> > of others on that. Of course, we may decide that *this* ontology should
> Yeah me too ;)
>
> > *not* cover classical music, recognizing that it is a very different
> > world, different traditions, etc, and we would have to do a separate
> > classical music ontology. This may very well be the way to go; I do not
> > really know.... But is is worth discussing!
> Definitely. But as I said above, the best would be that this Music Ontology
> become the basement for the development of other, more specialized, musical
> ontologies.
>
> > Thanks for all this...Thanks to you for these observations!
>
> Take care,
>
> Fred
Cheers
Ivan
P.S. I wonder whether the google group is the best place of discussing
these; though if, time to time, you make a compilation of things on a
blog, that could work.
Frederick Giasson
> Let me give you one specific example for a CD, this may clarify most of
> the issues I have (I hope:-).
>
> I have one specific CD. The CD itself does *not* have a specific name,
> but it contains four pieces of Schumann: Kreisleriana, Nachtstücke,
> Gesänge der Frühe, and Geister-Variationen, played on the piano by
> András Schiff.
Okay, I understand what you mean and what you want to describe.
> [
> a mo:Album;
> mo:consistsOf (_:p1, _:p2, _:p3, _:p4 )
> ]
Check my last proposition. In fact, an album would be nothing else than the
"type" given to a record (cd, etc) by the "industry". So, is it an Album, a
Composition, etc.
>
> _:p1 a MusicalWork;
> # maybe a seaparate type here for a distinct piece?
> mo:genre mo:Classical;
> dc:title "Kreiserliana";
> dc:creator: # or mo:composed?
dc:creator
> [
> a foaf:Person;
> a mo:Composer;
This is a good question.
Do I say that this person is a Composer? Or do I infer that it is a composer (at
some level) considering that he composed something?
In fact, I currently use the second way via the mo:composed property.
> foaf:name "Robert Schumann";
> ....
> ];
> # I am afraid the 'opus' is so unspecified in practice, that we have
> to stay by a string...
> mo:opus "Op. 133";
Should possibly add this mo:opus property to define a mo:MusicalExpression?
> # the key is not specified
> mo:consistsOf (_:tr10,_:tr11,...,_:tr17);
In fact, as I cited in my last proposition, I would relate a mo:Track to a
mo:Medium (that is the super class for mo:CD, mo:Md, etc).
And a mo:Track would be the record of a MusicalExpression on a mo:Medium.
> I call it now MusicalWork in reference to your blog, whether this is
> the right name or not, I do not know. The important is the strict
> differentiation between the album and the individual pieces which may
> in turn consists of possibly several tracks.
Read my latest proposition.
> For me, a compilation is when I publish a CD with, say, a collection of
> Beatles songs that were originally published for the first time in
> various albums. This is more or less meaningless for classical
> music....
I agree. The last proposition should fix that problem.
> Yes, I think (as I jotted above), we have a MusicalWork as a
> fundamental unit, and an Album (or a CD) *may* be the collection of
> such MusicalWorks. I am not sure the word 'Album' is the right one, but
In fact, with my proposition, a CD is a collection of Tracks, and Tracks are the
physical record of a MusicalExpression.
And an Album, is just a type of MusicalManifestation given by the musical
industry to market their things.
> that might be etimology only. I guess my main point is that the focal
> point is a MusicalWork (and this is what you put into your blog, in
> fact), and the term 'Album' is much much more vague. Of course, in the
> As I commented above: this is *really* messy. It is very often a set of
> letters and a number, something like 'Op. 45'. Sometimes it says 'Op.
> posth.' for a particular piece.
>
> For the time being I've put a string in my example. Maybe some general
> structure containing a string plus an integer plus some other qualifier
> may work out; I am happy to check with my collection if it comes to
> that.
Good.
> Ah. So what you say is:
>
> Artist a foaf:Person.
> Composer rdfs:subClass Artist.
> Performer rdfs:subClass Artist.
Yeah, but I don't define them. In fact, I infer that someone one is a composer
is he mo:composed something. Then, we can probably find its "level of
composition" by checking the number of composition he has.
> is that what you mean? That could work. If we use OWL, we could think
> about further characterizations (is each artist a composer or a
> performer or both?), but this can be done later...
Yeah.
> P.S. I wonder whether the google group is the best place of discussing
> these; though if, time to time, you make a compilation of things on a
> blog, that could work.
Dunno if it is the best, but it is certainly simple and it is working fine for
the SIOC ontology :)
Tell me what you think of my latest proposition. I think that all these
discussion are really reviling the true essence of what should be the Music
ontology.
Take care,
Fred