Existing semantic representations of music analysis encapsulate narrow sub-domain concepts and are frequently scoped by the context of a particular MIR task. Segmentation is a crucial abstraction in the investigation of phenomena which unfold over time; we present a Seg- ment Ontology as the backbone of an approach that models properties from the musicological domain independently from MIR implementations and their signal processing foundations, whilst maintain- ing an accurate and complete description of the relationships linking them. This framework provides two principal advantages which we explore through several examples: a layered separation of concerns that aligns the model with the needs of the users and systems that consume and produce the data; and the ability to link multiple analyses of differing types through transforms to and from the Segment axis.
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Admire2011 slides
1. The Segment Ontology:
Bridging Music-Generic
and Domain-Specific
Ben Fields Kevin Page, David De Roure
Electronics and Computer Science Oxford e-Research Centre
University of Southampton University of Oxford
United Kingdom United Kingdom
(Now at Musicmetric)
Tim Crawford
Department of Computing
Goldsmiths, University of London
United Kingdom
Friday, July 15, 2011
2. or
How to make your data
useful?
Friday, July 15, 2011
3. or
How to make your data
useful?
(for more than the person
at the desk next door)
Friday, July 15, 2011
4. Overview
• Introduction and background
• Foundations
• Ontology and approach
• Worked examples
• Conclusions
Friday, July 15, 2011
Sorry about an ontology talk at before 10 in the morning, but we’ll
all do our best...
15. model
Friday, July 15, 2011
More specifically, by modelling the relationships and links
between all these disparate things, we make the tools cheap
and easy.
17. Foundations
1. Domain-specific musicology
2. Domain-specific MIR tasks
3. Music-generic
4. High-level Relationships
Friday, July 15, 2011
18. Foundations
1. Domain-specific musicology: concepts,
in our use case, from musicology, and the
human interpretation of music and sound.
Friday, July 15, 2011
Domain-specific musicology concepts are elements of form, such as intro, verse, chorus,
bridge; sonata, minuet and trio, fugue. These are likely to be applied to sections of the
signal, for example “this section is a bridge”.
**Mythical music taxonomy**
19. Foundations
2. Domain-specific MIR tasks: parts of the
model that relate to an MIR task, such as
the elements extracted by a feature
extractor, common labels from a classifier,
or distance metrics.
Friday, July 15, 2011
Domain-specific MIR tasks encompass artifacts of the structural segmentation task, for
example a classifier might identify (and potentially label) sections that are similar; a contributor
task might identify chords. Again, these concepts are likely to be applied to sections of signal.
**Audio Feature ontology or related***
20. Foundations
3. Music-generic: common concepts that
transcend domain-specific such as
Intervals, Segments, etc.
Friday, July 15, 2011
Music-generic concepts are common to different tasks and applications. Here the segments would
be those annotated using the domain-specific concepts and the alignments and relationships
between them (e.g. the segment labelled as a chorus follows the segment labelled as a verse; that
one chord follows another).
21. Foundations
4. High-level Relationships: the absolute
and relative relationships between music
generic elements, TimeLines and SegmentLines;
and the maps between them.
Friday, July 15, 2011
Finally high-level relationships capture mappings between the musicologically
labelled segments and the MIR task derived segments. We expand upon this
example in Figure 6.
22. Foundations
1. Domain-specific musicology
2. Domain-specific MIR tasks
3. Music-generic
4. High-level Relationships
Friday, July 15, 2011
Our ontology will deal with 3 and 4 so 1 and 2 can interact more easily (hopefully. with tools.)
24. Ontology and approach
owl:
equivalent SubClassOf
tl:Interval Segment
Class
OWL-time:
tl:onTimeLine onSegmentLine
Interval
RatioMap
SubClassOf
e
tl:TimeLine SegmentLine
to from NonLinearMap
tl:domainTimeLine
SubClassOf isA
tl:TimeLineMap SegmentLineMap NonSequentialMap
Friday, July 15, 2011
25. Ontology and approach
owl:
equivalent SubClassOf
tl:Interval Segment
Class
OWL-time:
tl:onTimeLine onSegmentLine
Interval
RatioMap
SubClassOf
e
tl:TimeLine SegmentLine
to from NonLinearMap
tl:domainTimeLine
SubClassOf isA
tl:TimeLineMap SegmentLineMap NonSequentialMap
Friday, July 15, 2011
a tl Interval with the addition of a “label” expressing an association (SIM) that can be “placed” upon Time-
Lines (TL) and SegmentLines.
26. Ontology and approach
owl:
equivalent SubClassOf
tl:Interval Segment
Class
OWL-time:
tl:onTimeLine onSegmentLine
Interval
RatioMap
SubClassOf
e
tl:TimeLine SegmentLine
to from NonLinearMap
tl:domainTimeLine
SubClassOf isA
tl:TimeLineMap SegmentLineMap NonSequentialMap
Friday, July 15, 2011
SegmentLine: an AbstractTimeLine and a relative complement to the
temporal TimeLine.
27. Ontology and approach
owl:
equivalent SubClassOf
tl:Interval Segment
Class
OWL-time:
tl:onTimeLine onSegmentLine
Interval
RatioMap
SubClassOf
e
tl:TimeLine SegmentLine
to from NonLinearMap
tl:domainTimeLine
SubClassOf isA
tl:TimeLineMap SegmentLineMap NonSequentialMap
Friday, July 15, 2011
SegmentLineMap: a means to express a high-level relation- ship between
SegmentLines or with TimeLines; can imply relationships between Segments
on SegmentLines and TimeLines; similarly a SegmentLineMap can be used to
infer properties between Segments.
28. Ontology and approach
StartAtSegment
4/4Bar isA
ContainsSegment ContainsSegment Segment
Beat1 isA Beat2 isA beat3 isA Beat4 isA
EndAtSegment
sim:element !
Segment Segment Segment Segment
sim:element _:Similarity a _:Similarity a
sim:Similarity sim:Similarity
sim:element sim:element
Beat Bar Phrase Tadem
Mythical Music TS4/4 isA
!
Taxonomy SegmentLineMap
Friday, July 15, 2011
How about somethings (a bit) less abstract: 4 beats in 1 bar of 4/4
music.
29. Ontology and approach
StartAtSegment
4/4Bar isA
ContainsSegment ContainsSegment Segment
Beat1 isA Beat2 isA beat3 isA Beat4 isA
EndAtSegment
sim:element !
Segment Segment Segment Segment
sim:element _:Similarity a _:Similarity a
sim:Similarity sim:Similarity
sim:element sim:element
Beat Bar Phrase Tadem
Mythical Music TS4/4 isA
!
Taxonomy SegmentLineMap
Friday, July 15, 2011
Red for MIR researcher [alice]
Green for Musicologist [bob]
31. Worked example
Labelled structured segmentation
• Bob’s domain knowledge
• Alice’s MIR processes
• Mediated through the segment ontology
Friday, July 15, 2011
(you remember Alice and Bob, yes?)
32. Worked example
onSegmentLine SL1
isA SegmentLine
R
onSegmentLine
segmentAfter
S0 isA S1 isA S2 isA S3 isA tl:Interval
Segment Segment Segment Segment
M a
segmentBefore
mo:Signal
mo:recordedAs
_:Similarity a _:Similarity a _:Similarity a
sim:Similarity sim:Similarity sim:Similarity
D
_:Similarity a _:Similarity a _:Similarity a _:Similarity a
sim:Similarity sim:Similarity sim:Similarity sim:Similarity
"Intro" "Verse" "Refrain"
a label a label a label
"Help"
segmentAfter mo:MusicalWork
Help Intro isA Help Verse1 Help Refrain Help Verse2
Segment isA Segment isA Segment isA Segment
segmentBefore
Intro Verse Refrain Bridge
Mythical Music
Taxonomy
Friday, July 15, 2011
help! I need somebody [to explain linked data]
help! Not just anybody [eloquent speaker prefered]
help! (etc...)
[explain the graph]
33. Worked example
onSegmentLine SL1
isA SegmentLine
R
onSegmentLine
segmentAfter
S0 isA S1 isA S2 isA S3 isA tl:Interval
Segment Segment Segment Segment
M a
segmentBefore
mo:Signal
mo:recordedAs
_:Similarity a _:Similarity a _:Similarity a
sim:Similarity sim:Similarity sim:Similarity
D
_:Similarity a _:Similarity a _:Similarity a _:Similarity a
sim:Similarity sim:Similarity sim:Similarity sim:Similarity
"Intro" "Verse" "Refrain"
a label a label a label
"Help"
segmentAfter mo:MusicalWork
Help Intro isA Help Verse1 Help Refrain Help Verse2
Segment isA Segment isA Segment isA Segment
segmentBefore
Intro Verse Refrain Bridge
Mythical Music
Taxonomy
Friday, July 15, 2011
Alice’s domain-specific contribution
34. Worked example
onSegmentLine SL1
isA SegmentLine
R
onSegmentLine
segmentAfter
S0 isA S1 isA S2 isA S3 isA tl:Interval
Segment Segment Segment Segment
M a
segmentBefore
mo:Signal
mo:recordedAs
_:Similarity a _:Similarity a _:Similarity a
sim:Similarity sim:Similarity sim:Similarity
D
_:Similarity a _:Similarity a _:Similarity a _:Similarity a
sim:Similarity sim:Similarity sim:Similarity sim:Similarity
"Intro" "Verse" "Refrain"
a label a label a label
"Help"
segmentAfter mo:MusicalWork
Help Intro isA Help Verse1 Help Refrain Help Verse2
Segment isA Segment isA Segment isA Segment
segmentBefore
Intro Verse Refrain Bridge
Mythical Music
Taxonomy
Friday, July 15, 2011
Bob’s domain specific contribution
35. Conclusions
• The link between MIR researcher and
musicologist user is critical, yet difficult
• By moving complexing to a intermediary
tools to aid this link becomes cheap
• Make more tools!
• And the Mythical Music Taxonomy!
Friday, July 15, 2011
We have presented the Segment Ontology, that uniquely enables post-hoc integration across other datasets and
domain structures through a separation of concerns between segmentation, structure, and musical domains. The next
stage of our research is to deploy the ontology across the collections of ground truth and computational analysis
discussed in section 1, integrated as a part of a Linked Data API through which researchers can access, study, and
manipulate results. Completion of the Mythical Music Taxonomy is left as an exercise for the reader.
36. The End
Resources and more info:
http://www.linkedmusic.org/segment/
ISMIR Tutorial:
http://ismir2011.linkedmusic.org/
Friday, July 15, 2011
37. Questions?
Resources and more info:
http://www.linkedmusic.org/segment/
ISMIR Tutorial:
http://ismir2011.linkedmusic.org/
These slides:
http://slideshare.net/bfields
Friday, July 15, 2011