Designing and Using an Audio-Visual Description Core Ontology - - PowerPoint PPT Presentation

designing and using an audio visual description core
SMART_READER_LITE
LIVE PREVIEW

Designing and Using an Audio-Visual Description Core Ontology - - PowerPoint PPT Presentation

Designing and Using an Audio-Visual Description Core Ontology Antoine Isaac & Raphal Troncy Friday 8 th of October, 2004 Outline Motivations Methodology and content Focusing on domain needs Focusing on upper-level


slide-1
SLIDE 1

Designing and Using an Audio-Visual Description Core Ontology

Friday 8th of October, 2004

Antoine Isaac & Raphaël Troncy

slide-2
SLIDE 2

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

1

Outline

  • Motivations
  • Methodology and content

– Focusing on domain needs – Focusing on upper-level considerations – Reconciliation

  • Use
  • Conclusion
slide-3
SLIDE 3

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

2

Uses of AV Document Descriptions

  • Archival and description of documents from a

cultural heritage point of view: INA

  • Exchanging program identification and

characterization for interactive TV: TV-Anytime

  • Diffusion of program information (news

agencies): ProgramGuideML

  • Storing and sharing AV content descriptions

(automatic extraction results): MPEG7 standard

⇒ Development of standard vocabularies, syntactic specifications

slide-4
SLIDE 4

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

3

Meaning problem

  • Description deep meaning cannot be accessed

and processed by systems

– Knowledge is often implicit (labels and comments in natural language) – Formal specifications are mostly syntactic

⇒ Need for a formal ontology to better manipulate AV content

  • Formal semantics should be interesting

– Reasoning with AV document descriptions – Interoperability with formal domain-specific

  • ntologies, allowing to mix AV and domain-related

reasoning

slide-5
SLIDE 5

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

4

Can we find an AV core ontology?

  • There are many common needs amongst
  • bserved applications

– Characterization of programs and sequences – Decomposition of programs and sequences – Ability to introduce description of the activities that constitute the context of AV documents (roles of people involved, way production and broadcast are achieved)

  • These concepts are close to a "neutral"

archival viewpoint

slide-6
SLIDE 6

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

5

Outline

  • Motivations
  • Methodology and content

– Focusing on domain needs – Focusing on upper-level considerations – Reconciliation

  • Use
  • Conclusion
slide-7
SLIDE 7

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

6

Methodology

  • Grounding conceptualization by observed

purposes and domain initiatives

⇒ justification of the C.O. by making it compliant with shared views on the domain

  • Articulation with an upper-level ontology

⇒ justification of the C.O. by making it compliant with shared views on high-level categories and axiomatizations

⇒ Get a fully shareable and interoperable C.O.

slide-8
SLIDE 8

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

7

Outline

  • Motivations
  • Methodology and content

– Focusing on domain needs – Focusing on upper-level considerations – Reconciliation

  • Use
  • Conclusion
slide-9
SLIDE 9

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

8

MPEG 7 and AV C.O.

  • Large effort
  • Existing formal ontologies adaptations

– [Hunter, SWWS'2001] (RDFS) – [Tsinaraki, CAISE'2004] (OWL)

  • MPEG7 main features

– Descriptors focused on the physical features of the AV signal – Higher-level description schemes rather centred on grammatical specifications

⇒ More "conceptual" DSs need some development to catch core domain needs

slide-10
SLIDE 10

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

9

C.O. content

  • Concerning AV objects:

– distinction sequence/program – decomposition and qualification of those objects – link to external world themes and entities (content description)

  • Underlying use patterns for elicited categories

Theme-related Entities Channel, Live, Daily... Program broadcastRelation [hasBroadcaster, hasPeriodicity...] thematicRelation [hasTheme, shows, refersTo...] Sequence partOf partOf Productor, Zoom... productionRelation [hasAuthor, hasFilmingFeature...]

slide-11
SLIDE 11

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

10

Example:

Upper-level categorization of sequences

slide-12
SLIDE 12

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

11

Towards Formal Semantics

  • Formal definitions of concepts (NC, SC)
  • Relational axioms (composition)
  • An OWL example:

<owl:Class rdf:ID="DialogSequence"> <rdfs:subClassOf rdf:resource="#SpokenSequence"/> <rdfs:subClassOf> <owl:Restriction> <owl:onProperty> <owl:ObjectProperty rdf:about="#hasParticipant"/> </owl:onProperty> <owl:minCardinality rdf:datatype="&xsd;int">2</owl:minCardinality> </owl:Restriction> </rdfs:subClassOf> </owl:Class>

slide-13
SLIDE 13

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

12

Outline

  • Motivations
  • Methodology and content

– Focusing on domain needs – Focusing on upper-level considerations – Reconciliation

  • Use
  • Conclusion
slide-14
SLIDE 14

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

13

Upper-level foundations

  • Chosen framework:

– DOLCE [Gangemi, EKAW 2002] – Description & Situation extension [Gangemi, ODBASE 2003]

  • Provides:

– Upper-level concepts and relations – Ontological design pattern

⇒ Both of them can be specialized to match domain needs

slide-15
SLIDE 15

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

14

D&S pattern specialization

subsumption link conceptual relation D&S pattern Parameter Role Course

  • f

Events BroadcastTime, Audience Broadcaster, Channel, BroadcastedProgram, Receiver Broadcast- CouseOf Event Date, Rate Organization, Person, AVDocument Emission, Reception Region Endurant Perdurants valuedBy playedBy sequences modality f or has f or requisite has for requisite participant in location location X

slide-16
SLIDE 16

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

15

Outline

  • Motivations
  • Methodology and content

– Focusing on domain needs – Focusing on upper-level considerations – Reconciliation

  • Use
  • Conclusion
slide-17
SLIDE 17

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

16

Articulation with domain needs

  • Do foundationally motivated choices really fit

domain needs?

– Some notions are too abstract – Some relational paths are too long

  • To be usable in the domain, core notions

have to be adapted to domain uses

– Goal:

  • Articulation between upper-level AV pattern and use

patterns

– How?

  • With formal rules allowing KBS to deal simultaneously

with both forms of knowledge

⇒ descriptions may be far from domain concerns

slide-18
SLIDE 18

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

17

Relational shortcut example (1)

BroadcastTime BroadcastedProgram BroadcastCourseOfEvents Date Program valuedBy plays modality for has for requisite wasBroadcastedAt existing relation inferred relation X X

slide-19
SLIDE 19

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

18

<ruleml:imp> <ruleml:_body> <swrlx:classAtom> <owlx:Class owlx:name="Program" /> <ruleml:var>prgm</ruleml:var> </swrlx:classAtom> <swrlx:classAtom> <owlx:Class owlx:name="BroadcastedProgram" /> <ruleml:var>bcPrgm</ruleml:var> </swrlx:classAtom> <swrlx:classAtom> <owlx:Class owlx:name="BroadcastCourseOfEvents" /> <ruleml:var>bcCOE</ruleml:var> </swrlx:classAtom> <swrlx:classAtom> <owlx:Class owlx:name="BroadcastTime" /> <ruleml:var>bcTime</ruleml:var> </swrlx:classAtom> <swrlx:classAtom> <owlx:Class owlx:name="Date" /> <ruleml:var>date</ruleml:var> </swrlx:classAtom> <swrlx:individualPropertyAtom swrlx:property="&dolce;plays"> <ruleml:var>prgm</ruleml:var> <ruleml:var>bcPrgm</ruleml:var> </swrlx:individualPropertyAtom> <swrlx:individualPropertyAtom swrlx:property="&dolce;modality-for"> <ruleml:var>bcPrgm</ruleml:var> <ruleml:var>bcCOE</ruleml:var> </swrlx:individualPropertyAtom> <swrlx:individualPropertyAtom swrlx:property="&dolce;has-for-requisite"> <ruleml:var>bcCOE</ruleml:var> <ruleml:var>bcTime</ruleml:var> </swrlx:individualPropertyAtom> <swrlx:individualPropertyAtom swrlx:property="&dolce;valued-by"> <ruleml:var>bcTime</ruleml:var> <ruleml:var>date</ruleml:var> </swrlx:individualPropertyAtom> </ruleml:_body> <ruleml:_head> <swrlx:individualPropertyAtom swrlx:property="wasBroadcastedAt"> <ruleml:var>prgm</ruleml:var> <ruleml:var>date</ruleml:var> </swrlx:individualPropertyAtom> </ruleml:_head> </ruleml:imp>

Relational shortcut example (2)

slide-20
SLIDE 20

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

19

Outline

  • Motivations
  • Methodology and content

– Focusing on domain needs – Focusing on upper-level considerations – Reconciliation

  • Use
  • Conclusion
slide-21
SLIDE 21

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

20

How to use a core AV ontology?

  • Domain extension (and restriction)

– Complementary vocabulary: roles, kinds of AV creation processes and effects, etc. – Focusing choices: for some sub-domains, no need for complex description of specific AV actions (broadcast)

  • Application extension

– Fine-grained vocabulary and reasoning knowledge customization – Articulation with ontologies describing "world" domains (with formal knowledge involving concepts and relations from both ontologies)

slide-22
SLIDE 22

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

21

Applications

  • TV-Anytime

– We can now create formal descriptions referring to TV-Anytime vocabulary

  • [Troncy, ISWC 2003]

– AV ontology has been used in conjunction with a domain ontology (cycling) to formally describe structure and content of sports-related AV documents

  • OPALES project

– Similar experiment, characterization of sequences and their AV features from a pedagogical viewpoint

slide-23
SLIDE 23

10/08/2004

  • A. Isaac & R. Troncy - CORONT'2004

22

Conclusion

  • Dual legitimacy for a core ontology

– Domain relevance (user needs) – Upper-level compliance (Dolce)

  • Limitations and problems

– Time-consuming effort (adaptation, rules, etc.) – Is full-scale reasoning feasible? – Limits between core and domain conceptualizations – TV bias (publication instead of broadcast?)

⇒ Reasoning knowledge as reconciliation