OPALES Demonstration

1
OPALES Demonstration
Antoine Isaac

• CWI Feb. 22nd 2006

2
OPALES

• French Project (RIAM, 2001-2003)
• Partners
•  Developers 
• INA (Coordinator)
• LIRMM (IHM, GC)
• CS
•  Users 
• CNRS (video library)
• MSH (humanities, communication)
• CNDP (educational material)

3
 Tools to share knowledge in educational/scientif
ic communities 

• Goals
• Sharing and exploiting AV resources
• Fitting the needs of various specific user
  communities
• Approach
• Accessing document via annotations
• Similar to classical indexing approach
• Interpretation of AV content used for
  exploitation
• Ensuring the relevance of annotation for
  communities and applications
• Using points of view

4
Points of view

• Exploitation of annotation/indexes needs semantic
  continuity
• Shared interpretation/coherent manipulation for
  indexer/searcher/system
• OPALES points of view
• Mirror specific application/community needs
• Embody annotation policies
• Gather users who choose them for creating or
  filtering annotations
• Thorough experiments with 2 points of view
• Ethnology-oriented analysis of videos regarding
  childhood
• Educational exploitation of geographic
  information films

5
Points of view and annotation/search policies

• Free-text annotation and search
• Natural interpretation of AV content
• Form-based annotation and search
• Predefined structure
• Basic control (value lists)

6
OPALES Forms
7
Points of view and annotation/search policies

• Free-text annotation and search
• Natural interpretation of AV content
• Form-based annotation and search
• Predefined structure
• Basic control (value lists)
• Ontology-based annotation and search

8
Ontology-based annotation and search

• Explicit and flexible structuring of index
  elements
• Added value of relations for rich indexing
• Conceptual Graphs formalism

9
Ontology-based annotation
10
Ontology-based annotation and search

• Explicit and flexible structuring of index
  elements
• Control ensured by formal specifications
• Inference
• Taking into account (viewpoint-specific) implicit
  knowledge to link queries to indexes
• Cogitant CG inference engine

11
Problems of ontology-based approach

• Complex indexing process
• Difficult to get accustomed to
• Especially for people not used to KR
• Coupling design of complex ontological knowledge
  with use
• Legitimate conceptualisation w.r.t. application
  and practice
• Specification of formal inference knowledge

12
DOE ontology editor

• Semi-informal semantics Bachimont
• Differential principles inter-defining notions by
  similarities and differences
• Meaning naturally linked to an application domain
• Compatible with extraction of ontology elements
  from textual resources
• Véronique Malaisé
• Compatible with further formalization

13
DOE

• http//opales.ina.fr/public Bachimont, Isaac,
  Troncy
• (now http//homepages.cwi.nl/troncy/DOE/ !)

14
Indexing patterns

• Determine relational template recurrent in aimed
  information practices
• Dependent on point of view
• Propose those structures to the indexer as a
  starting point for description

15
Indexing patterns example (1)
English
16
(No Transcript)
17
Indexing patterns example (3)
English
18
Patterns and reasoning

• Comparison index/query
• Pattern as an implicit pivotal structure,
  compensating for description/query variations
• Inferring facts from pattern structure
• Inferring pattern-fitting facts from different
  structure

19
Conclusion

• Points of view as materialization of indexing and
  exploitation strategies
• Knowledge-based tools
• Expressivity, control and inference
• Experimentation (two qualitative evaluations with
  domain users) shows that there are obstacles
• No familiarity of targeted users with formal
  techniques
• Difficulty for the knowledge workers to design
  proper ontological resources for each point of
  view

20
Conclusion

• Experimentation shows that approach is made
  feasible by
• Ontologies linked in quite a natural way to
  meanings and practices targeted by points of
  views
• Differential principles help ontology designers
  work and final users understanding
• Formalizing users needs by Indexing Patterns
• Both for description and query
• Also legitimates ontology designers work
• Turning to relational reasoning for
  compensating variations
• Remark complexity goes now for KR engine
• Techniques to ensuring semantic continuity in
  ontology-based indexing and search

21
THANK YOU !
22
Childcare use pattern
Back
23
Specialised index
Back