Third Project Review

1
Corporate Semantic Web
Acacia http//www.inria.fr/acacia INRIA Sophia
Antipolis
2
Corporate Semantic Web ?
Use Semantic Web approach for Corporate Memory
and Corporate Knowledge Management
3
Objectives
implement and trial a corporate memory management
framework based on agents and ontologies CoMMA
Corporate Memory Management with Agents
2 relevant scenarios have been chosen to
highlight the problem of information retrieval in
the company

• Enhancement of New Employee Insertion in the
  company,

• Performing process that detect, identify and
  interpret technology movements for matching
  technology evolutions with market opportunities
  to disseminate among employees innovative ideas
  related to Technology Monitoring activities

Objectives
4
Objectives
Corporate knowledge management aims at
facilitating creation, dissemination, transmission
and reuse of knowledge in an organisation

• propose an innovative solution based on
  integration of technologies
• ontologies or knowledge models
• multi-agent architecture of several co-operating
  agents
• meta-information (resource annotation) expressed
  in RDF format
• Machine Learning Techniques for user adaptability

Objectives
5
CoMMA Objectives
Objectives
6
CoMMA Consortium
European IST project 2000-2001 3 industrial
partners Atos Origin (F) CSTB (Centre
Scientifique et Technique du Batiment)
(F) T-Systems Nova (G) 3 academic
partners INRIA (F) LIRMM/CNRS (F) University
of Parma (I)
CoMMA Consortium
7
CoMMA What is it ?

• Corporate Memory
• An explicit, disembodied and persistent
  representation of knowledge and information in an
  organization, in order to facilitate their access
  and reuse by members of the organization, for
  their tasks.

8
How ?

• How ?
• Corporate memories are heterogeneousand
  distributed information landscapes
• Stakeholders are an heterogeneous and distributed
  population
• Exploitation of CM involves heterogeneousand
  distributed tasks

Multi-Agent System Modularity, Distributed,
Collaboration Machine Learning Adaptation,
Emergence
XML Standard, Structure, Extensible, Validate,
Transform RDF Annotation, Schemas
9
Overall Schema Ontology
(2)
(1)
(3)
10
Example of problem ambiguity

• The balance of our pharmaceutical project.

• Two concepts one term ambiguity
• Ontology object capturing relevant aspects of
  the meaning of concepts used in our application
  scenarios (example)

11
Building the ontology
12
Memory Structure
13
Illustration of the cycle
14
Model-based Annotated Memory

• Corporate Semantic Web
• RDF RDFS XML framework for Web resources
  descriptions ? Use it for Intranets
• Ontology in RDFS
• Description of the Situation in RDF
• User Profiles
• Organization model
• Annotations in RDF describing Documents

15
End-Users
16
Interfacing Users

• User Interfaces
• Annotating documents
• Querying the memory
• Hide complexity (ontology, agents,...)
• Present the results

• Push technology
• Improve information flowing
• Proactive diffusion of annotations
• Communities of interest

17
Profiles Learning

• Organizational model
• Users' Profiles
• Administrative Information (link to Org. model)
• Explicit preferences
• Favorite queries / annotations
• Characteristics derived from past use
• Learning techniquesRepresent, learn and compare
  current use profiles to improve future use.
• Learning during a login session
• Ranking results

18
Multi-agent Architecture
19
Principal interest of MAS in CoMMA

• One functional architecture leading to several
  possible configurations in order to adapt to the
  broad range of environments that can be found in
  a company
• Architecture Agent kinds and their relationship
  Fixed at design time
• Configuration Exact topography of a given MAS
  Fixed at deployment time
• Flexible distribution
• Locally adapt to resources and users
• Global capitalization through cooperation
• Integration of different technologies

20
Societies, Roles and Interactions
21
Conclusion
Done
22
Authors

Engineer

Archivist

(internal / informal sources)

(external sources)
DocsAnnotations


Area referent

Coordination Strategic orientation

ANNOTATION

Index card ,Synthesis,

PUSH



Query

RETRIEVAL

User

TECHNOLOGY
MONITORING


The diffusion of innovative ideas among employees
The Technology Monitoring scenario
23

• The actors of the Technology Monitoring scenario
  
• Archivist in charge of feeding the system -gt
  Author
• Engineer and Researcher
• watching his expertise Area -gt User
• feeding the system with new information -gt
  Author
• in charge of identifying correspondents and
  coordinating thematic groups -gt Area referent

The actors
24

• For the Authors
• Indexing information by annotating companies,
  people, documents...
• For the Area referents
• Identifying resources, skills about given
  business domains
• For the Users
• Being automatically informed about relevant
  information according to their profile (push
  mode)
• Querying the system (pull mode)

Examples of Supported tasks
25
NEI Scenario the insertion of new employees in
the company concerns the new employees who need
to handle a lot of new information about their
enterprise in a very short time, to be rapidly
efficient
26
The actors

• The NE who just arrived in his new company
• not familiar with the environment
• needing answers to many standard questions
• The tutor
• person responsible to support NEs during the
  first weeks
• with CoMMA responsible to fill the annotation
  base

27
CoMMA solution
The CoMMA Solution

• 5 major components
• An ontology (OCoMMA)
• A multi-agent system,
• A Semantic search engine (CORESE),
• A machine learning algorithm
• A GUI
• ? The CoMMA technical solution for the
  implementation of a Corporate memory.

28
CoMMA solution

• splitting resources / system

29
CoMMA solution

• splitting resources / system
• the document resources

30
CoMMA solution

• splitting resources / system
• the document resources
• the configuration resources

31
CoMMA solution

• splitting resources / system
• the document resources
• the configuration resources
• Ontology

32
CoMMA solution

• Ontology OCoMMA
• Dedicated to corporate memory,
• Represented in RDFS,

33
CoMMA solution

• rdfsClass for concepts of the ontology,
• Possibility to use class inheritance

34
CoMMA solution

• rdfProperty for relations of the ontology,
• specialization of properties
• director subPropertyOf manager
• director ? manager

35
CoMMA solution

• rdfslabel for synonyms and multi- language of
  the ontology,
• Use of stylesheet to filter terminology and
  multi-language.

36
CoMMA solution

• rdfscomment for natural language definition
• the link between definition and concept is kept
• ? ontology trackability

37
RDFS Example Class

• ltrdfsClass rdfID"Document"gt
• ltrdfssubClassOf rdfresource"Entity"/gt
• ltrdfssubClassOf rdfresource"EntityConcerningA
  Topic"/gt
• ltrdfssubClassOf rdfresource"NumberableEntity"
  /gt
• ltrdfscomment xmllang"en"gtEntity including
  elements serving as a representation of thinking.
• lt/rdfscommentgt
• ltrdfscomment xmllang"fr"gtEntite comprenant
  des elements de representation de la pensee.
• lt/rdfscommentgt
• ltrdfslabel xmllang"en"gtdocumentlt/rdfslabelgt
• ltrdfslabel xmllang"fr"gtdocumentlt/rdfslabelgt
• lt/rdfsClassgt

38
RDFS Example Property

• ltrdfProperty rdfID"Title"gt
• ltrdfssubPropertyOf rdfresource"Designation"/gt
• ltrdfsrange rdfresource"rdfsLiteral"/gt
• ltrdfsdomain rdfresource"Document"/gt
• ltrdfscomment xmllang"en"gtDesignation of a
  document.
• lt/rdfscommentgt
• ltrdfscomment xmllang"fr"gtDesignation du
  document.
• lt/rdfscommentgt
• ltrdfslabel xmllang"en"gttitlelt/rdfslabelgt
• ltrdfslabel xmllang"fr"gttitrelt/rdfslabelgt
• lt/rdfPropertygt

39
CoMMA solution

• splitting resources / system
• the document resources
• the configuration resources
• Ontology, Enterprise model

40
Enterprise Model

• ltcLegalCorporation rdfabout"http//www.inria.f
  r/"/gt
• ltcNationalOrganizationGroup rdfabout"http//w
  ww.inria.fr/"gt
• ltcDesignationgtInstitut National de Recherche
  en Informatique et Automatiquelt/cDesignationgt
• ltcHasForActivitygtltcResearch/gtlt/cHasForActivi
  tygt
• ltcIsInterestedBygtltcComputerScienceTopic/gtlt/c
  IsInterestedBygt
• ltcIsInterestedBygtltcMathematicsTopic/gtlt/cIsIn
  terestedBygt

41

• ltcLocalOrganizationGroup rdfabout"http//www-so
  p.inria.fr/"gt
• ltcDesignationgtUR Sophia Antipolis de l'INRIA
  Institut National de Recherche en Informatique et
  Automatiquelt/cDesignationgt
• ltcHasForActivitygtltcResearch/gtlt/cHasForActivi
  tygt
• ltcIsInterestedBygtltcComputerScienceTopic/gtlt/cIs
  InterestedBygt
• ltcIncludegtltcProjectGroup rdfabout"http//w
  ww.inria.fr/recherche/equipes/acacia.en.html"/gtlt/c
  Includegt
• ltcIncludegtltcProjectGroup rdfabout"http//w
  ww-sop.inria.fr/tropics/"/gtlt/cIncludegt
• ltcIncludegtltcProjectGroup rdfabout"http//w
  ww-sop.inria.fr/cafe/"/gtlt/cIncludegt

42
CoMMA solution

• splitting resources / system
• the document resources
• the configuration resources
• Ontology, Enterprise model, User profiles

43
User Profile Example

• ltcIndividualProfile rdfabout""gt
• ltcCreationDategtan 2000lt/cCreationDategt
• ltcTitlegtEmployee profile of Olivier
  Corbylt/cTitlegt
• lt/cIndividualProfilegt
• ltcEmployee rdfID "http//www-sop.inria.f
  r/acacia/personnel/corby/"gt
• ltcFamilyNamegtCorbylt/cFamilyNamegt
• ltcFirstNamegtOlivierlt/cFirstNamegt
• ltcHasForOntologicalEntrancePointgtltcKnowledgeMo
  delingTopic/gtlt/cHasForOntologicalEntrancePointgtlt
  cHasForOntologicalEntrancePointgtltcObjectProgramm
  ingTopic/gtlt/cHasForOntologicalEntrancePointgt

44
CoMMA solution

• splitting resources / system
• the document resources
• the configuration resources
• the multi agent system framework

45
CoMMA solution

• Gui building an annotation.

46
CoMMA solution

• Machine Learning technique

• use feedbacks to learn document relevancy
• feedback from one user can be generalized to
  users having the same fields of interest,
• is designed for both pull mode and push mode

47
CoMMA solution

• Multi-agent system
• document sub society

48
CoMMA solution

• CORESE a semantic search engine
• relies on RDF(S) and conceptual graph theory,
• use of the inheritance graph of RDFS
  (specialization and generalization),
• Inference mechanisms
• manage the annotation distribution
• Java API wrapped into an agent

• Multi-agent system
• document sub society

49
RDF Annotation

• ltcResearchReport rdfabout'http//www.inria.fr/r
  apports/sophia/RR-3819.html'gt
• ltcCreatedBygt
• ltcPerson rdfabout'http//www.inria.fr/nad
  a.matta'gt
• ltcFamilyNamegtMattalt/cFamilyNamegt
• ltcFirstNamegtNadalt/cFirstNamegt
• lt/cPersongt
• lt/cCreatedBygt
• ltcCreatedBygt
• ltcPerson rdfabout'http//www.inria.fr/oli
  vier.corby'gt
• ltcFamilyNamegtCorbylt/cFamilyNamegt
• ltcFirstNamegtOlivierlt/cFirstNamegt
• lt/cPersongt
• lt/cCreatedBygt

50
RDF Annotation

• ltcCreatedBygt
• ltcProjectGroup rdfabout
  'http//www.inria.fr/recherche/equipes/acacia.en.h
  tml'gt
• ltcDesignationgtAcacialt/cDesignationgt
• ltchasCreated rdfresource'http//www.i
  nria.fr/rapports/sophia/RR-3819.html'/gt
• lt/cProjectGroupgt
• lt/cCreatedBygt
• ltcCreationDategt11-1999lt/cCreationDategt
• ltcTitlegt Méthodes de capitalisation de
  mémoire de projet
• lt/cTitlegt

51
CoMMA solution

• Multi-agent system
• Interconnecting sub society

52
CoMMA solution

• A Distributed annotations management algorithm
• Relies on
• metrics that evaluate the semantic similiarity
  of annotations
• complex protocols between  connecting agents 
  and  document agents  to rebuild the splitted
  annotation.

• Multi-agent system
• Interconnecting sub society

53
CoMMA solution
The CoMMA Methodology
54
Other project results
Other project results

• O CoMMA ontology
• Extension of RDF(S) language for representing
  knowledge
• CORESE new inference mechanisms
• Techniques of categorization of RDF-annotated
  documents
• Multi-agent architecture for IR
• RDF-based JADE ontology content language
• Management of distribution of annotations and of
  queries.
• Machine learning techniques

55
Ontology O CoMMA

• Method Data collection, Terminological Phase ,
  Structuration, Validation, Formalization in RDFS

• Result 420 concepts, 50 relations, 630 terms,
  12 levels of depth

56
CORESE search engine
57
Relation properties
58
Relation properties
RDF CG
59
Relation properties
Relation properties

• Transitivity, Symmetry, Reflexivity, Inverse for
  RDF properties
• Annotations are augmented with new knowledge
  deduced from these properties
• Transitivity, symmetry and inverse are computed
  once and added to annotations
• Reflexivity is computed on the fly according to
  queries

60
Inference Rules for RDF
Inference Rules for RDF

• Augment the ontology with rules that enable to
  deduce and add new knowledge to annotations

IF a team participates to a consortium AND
a person is a member of the team THEN the
person participates to the consortium
IF Person?p-(member)-Team?t-(participates)-
Consortium?c THEN Person?p-(participates)-Co
nsortium?c

• Forward chaining inference engine

61
Inference Rules for RDF
RDF Rule Syntax
ltcosrulegt ltcosifgt ltcPerson rdfabout?pgt ltc
membergt ltcTeam rdfabout?tgt ltcparticipat
egt ltcConsortium rdfabout?c/gt lt/cparti
cipategt lt/cTeam lt/cmembergt lt/cPerson lt/cosi
fgt ltcosthengt ltcPerson rdfabout?pgt ltcpartic
ipategt ltcConsortium rdfabout?c/gt ltcpartic
ipategt lt/cPerson lt/costhengt lt/cosrulegt
62
Conclusion
Conclusion
The CoMMA system is implemented A Corporate
Semantic Web http//www.si.fr.atosorigin.com/sophi
a/comma tested at T Nova Systems (Deutsche
Telekom) and CSTB testbed for Corporate Semantic
Web technologies XML, Agents, Ontology,
Semantic metadata, Learning
63
Conclusion
Conclusion (2)
Corese semantic engine RDF(S) and Conceptual
Graphs tested at Renault on a design project
memory tested with the Gene Ontology