Pr

1
ONTOLOGIES beyond fashion A short introduction to
ontologies and the Semantic Web
Michel SIMONET TIMC-IMAG research
laboratory Université Joseph Fourier Grenoble -
France
2
ONTOLOGIES beyond fashion A short introduction

• FASHION the Semantic Web
• 1998 Tim Berners-Lee
• Machine and machine man
  and machine can communicate
• Communication / Understanding
• based on ONTOLOGIES

3
ONTOLOGIES A short introduction

• An example through Information Retrieval
• History, definition, examples
• Ontologies and data integration
• Ontologies and Information systems
• Ontology as the starting point of an Information
  System
• From ontology to database and software
• Gennere example

4
ONTOLOGIES A short introduction

• An example through Information Retrieval
• History, definition, examples
• Ontologies and data integration
• Ontologies and Information systems

5
The Medical Information Gap
Heterogeneous Medical Literature Databases and
the Internet
Medical Professionals Users
TOXLINE
CancerLit
EMIC
MEDLINE
Current Information Interfaces
Hazardous Substances Databank

• Aronson AR, Rindflesch TC. Query Expansion
  Using the UMLS Metathesaurus. In AMIA Annual
  Fall Symposium 1997 1997. p. 485-89.

6
Information searchon the Internet

• Google
• Easy to use Natural language
• Quality of result
• Time-consuming
• Medline - Pubmed
• High quality of scientific content
• Ease of use Controlled vocabulary (MeSH
  thesaurus)
• Time-consuming
• Objective Concept-based rather than
  word-based search

7
Word-based search
Mastectomy is a Breast Removal Synonym
8
Concept-based search
Breast removal Ablation du sein
Pertinent
Mastectomy Mastectomie
Pertinent
Searching
Breast removal
MASTECTOMY
Concept
Mammectomy Mammectomie
Pertinent
RADICAL MASTECTOMY
Radical mastectomy Mastectomie radicale
Pertinent
9
What is an ontology?Graphical representation
TREATMENT
IS-A
RADIOTHERAPY
CHEMOTHERAPY
SURGERY
ABLATION
followed_by
TUMORECTOMY
MASTECTOMY
HUMAN_BODY
Remove
Part-Of
ORGAN
BREAST
RADICAL MASTECTOMY
10
Benefits of Concept-based Information Retrieval

• Search is automatically extended to synonyms
• E.g., query  breast removal  ?
  mastectomy, mammectomy,
• Independence from query language
• E.g., query in French  mastectomie 
  answer documents in any language (e.g.,
  English, French, Spanish, German,
  Greek, Chinese )
• Query expansion using the concept hierarchy
• Result presentation using the Ontologys
  organization
• General orientation of the Semantic Web

11
ONTOLOGIES A short introduction

• An example through Information Retrieval
• Definition, examples, history
• Ontologies and data integration
• Ontologies and Information systems

12
What is an ontology?

• The Origins Plato and Aristotle
• A need to organize knowledge
• History in Computer Science
• Various definitions
• Consensus definition
• Example
• W3C hierarchy of languages
• Ontology usages

13
The Origins Plato and Aristotle

• Aristotle the study of beings insofar as they
  exist
• Reality Individuals Vs Concepts
  Plato, John Human
• What is universal, beyond particular
  representations?
• Categories of being - Physical objects -
  Minds - Classes - Properties -
  Relations
• Porphyry (3rd century) Porphyrys trees
• Categorization by identity and difference
• The basis of contemporary ontologies

14
History a need to organize knowledge

• Classifications in biology
• Linné (1707-1778)
• Thesaurus in Information Retrieval
• Consensus about names and structure

I.A.
Philosophy

• - Gruber (1990) Stanford KIF, Ontolingua
• Sowa Conceptual Graphs
• Description Logics (DL)
• Semantic Web

• Russell, Wittgenstein, Frege, Husserl, Peirce
• Nicola Guarino
• Barry Smith

15
History in Computer Science

• Semantic Networks (Shank - 1968)
• Concepts and relationships
• Confusion between concepts and individuals
• STUDENT IS-A PERSON
• John IS-A PERSON
• Conceptual Graphs (Sowa 1980)
• Formalization of semantic networks
• First-Order Logic
• Gruber (1990 Stanford)
• KIF Knowledge Interchange Format
• Ontolingua a language and a platform for
  ontology exchange
• 1st use of the term Ontology in Computer Science

16
Consensus definition

• CONCEPTS
• RELATIONSHIPS between concepts
• IS-A relationships (generic/specific)
• part-of relationships
• other relationships
• VOCABULARY preferred term for a concept
• DEFINITIONS
• informal, in natural language
• formal (eg., Description Logics)

17
Example Breast Cancer (1)

• CONCEPTS BREAST SURGERY MASTECTOMY
• RELATIONSHIPS between concepts
• LEFT MASTECTOMY IS-A BREAST SURGERY MASTECTOMY
• ENTIRE LEFT BREAST is-proper-material-part-of
  ENTIRE LEFT THORAX
• LEFT MASTECTOMY has-theme ENTIRE LEFT BREAST
• VOCABULARY preferred term for a
  concept breast surgery mastectomy

From the INFACE Ontology by Language and
Computing (www.landc.be)
18
Example Breast Cancer (2)

• CONCEPTS e.g., MASTECTOMY
• RELATIONSHIPS between concepts
• MASTECTOMY IS-A ABLATION
• ORGAN part-of HUMAN BODY
• TUMORECTOMY followed-by RADIOTHERAPY
• VOCABULARY preferred term
• mastectomy, mammectomy, breast removal,
  mastectomie, mammectomie, ablation du sein,
  µaste?t?µ?
• DEFINITIONS
• Surgical removal of the breast

From a Patient-oriented ontology by Radja
Messai TIMC (UJF)
19
What is an ontology?Graphical representation
20
W3C hierarchy of languages

• XML (eXtensible Markup Language)
• XML Schema (XSD)
• RDF (Resource Description Framework)
• RDF Schema
• OWL (Web Ontology Language)

In the framework of the Semantic Web
21
ONTOLOGIES A short introduction

• An example through Information Retrieval
• History, definition, examples
• Ontologies and data integration
• Ontologies and Information systems

22
ONTOLOGIES Data Integration
23
Problem

• Various types of data
• Structured databases
• Informal Texts
• Semi-structured XML
• Heterogeneity
• Query languages
• Structure, vocabulary,

24
Query through an Ontology
ConceptAnatomo-fonctionnel
ConceptAnatomique
Conceptfonctionnel
User
Hidbrain
Midbrain
Correspondence
Adaptor1
Adaptor2
Adaptor3
.. ..
. .
Texts
Databases
XML
25
ONTOLOGIES A short introduction

• An example through Information Retrieval
• History, definition, examples
• Ontologies and data integration
• Ontologies and Information systems

26

Ontologies
Information Systems
Ontologies and Information Systems in the
Health Field

• Common understanding? / Consensus ?
• Communication / Sharing

ONTOLOGIES
CONCEPTS DEFINITIONS
WHAT we speak about?
HOW do we speak of it?
VOCABULARY
27
Information Sytems

• The Information System is a support for
  communication inside the enterprise and between
  the enterprise and its environment 
• G. Panet R. Letouche Modèles techniques
  Merise avancés.
• The Real Organization
• se transforme, agit
• communicates
• memorise
• The system which is built to REPRESENT
• Actions
• Communication
• Memorisation

HUMAN
Software
Ontologies
DatabasesData warehouses
28

29
What is an ontology?

• CONCEPTS
• RELATIONSHIPS between concepts
• ISA relationships (generic/specific)
• part-of relationships
• other relationships
• VOCABULARY preferred term for a concept
• DEFINITIONS
• informal, in natural language
• formal

CONSENSUS
30
Example in mycology

ISA concepts hierarchy
Formal definitions of constraints
? Object classification /
identification
Champignon à lames
? Detection of inconsistent definitions
Russule
Ammanite
Couleurblanc,crèmeChairgrenue, cassante
Knowledge Base Diagnosis aid
Virescens
Cyanoxantha
Couleur blancChair grenueCouleurChair blanc
CouleurChair rosé
31
Formal Ontologies and Knowledge Bases
INSTANCE CLASSIFICATION
CONCEPT CLASSIFICATION

32
Formal Ontologies and Knowledge Bases

DefConcept PERSON ? nameSTRING and
ageINT DefConcept ADULT PERSON and
age18 DefConcept SENIOR PERSON and age
65 DefConcept SENIOR1 SENIOR and agelt 60

• Formal Ontology
• representation in a logical formalism
• E.g. Description Logic (DL)
• Inférence- Subsomption Concept
  Classification Consistency -
  Instance Classification

SUBSOMPTION SENIOR ? ADULT SENIOR ?
ADULT SENIOR subsumed by ADULT
CONSISTENCY SENIOR1 ? EMPTY_CONCEPT
33
Database Design
First Step
SEX

NHS
NAME
1st Name

• Identify the concepts of the domain

PERSON
HOSPITAL
CC
SPEC

• Determine relationships and their
  cardinalities

DOCTOR
PATIENT
prescrit
paie
DISEASE
ACT
coding
Micro-ontology of the domain
Ontological schema
34
Database Design

Ontological Schema
Name things and their relationships
Choice - model constraints (associations, ) -
cultural choices - object / value
E-R, UML Schema classes, methods
Choic - model constraints - atomic attributes
- normalization
Relational Schema tables
Choice - index - buffers - DBMS specific
features
Physical level files
Optimize
35
Database Design

• Evolution level ? Ontological level
• Corrective evolutive maintenance (80 cost)
• ? Loss of initial semantics
• Maintain links between levels
• ? Understanding, Mastering

36
GENNERE Project achievements and perspectives

• GENNERE database and software for 2 fields
• Nephrology (ESRD End-Stage Renal Disease)
• Rheumatology (RA Rhumatoid Arthritis)
• User testing and validation is ongoing at Rui Jin
  hospital
• Perspectives
• Data Warehouse for epidemiological studies
• Geographical Information systems
• Improve tools and methods for genericity

37
Genericity Achievements and Limits

• Genericity
• Common core ontology PATIENT FOLLOW-UP
  TREATMENT
• Common schema concepts and attributes
• Common set of events New Patient, New Treatment,
  Patient Transfer, Decease,
• Automatic generation of database (ISIS CASE tool)
• Database access through views (with limitations
  due to DBMS)
• Intensive use of metadata
• Domain values Comorbidities,
• Multilingualism (UTF8) GUI items, domain values
• Standard medical classifications (thesaurus)

38
Genericity Achievements and Limits

• Limits
• Disease-specific data
• e.g., vaccinations for RA
• core concepts (DISEASE, TREATMENT) have to be
  derived according to each disease
• Rheumatology TREATMENT is more complex
• Country-specific data
• Culture and health care system are different
• Patient identification, addresses
• No standard multilingual version of ICD10
• Thesaurus translation into Chinese
• No framework for automatic GUI generation

39
Conclusions

• Genericity was partly achieved
• Gain around 50 for 2nd disease (RA) although
  more complex
• Multilingualism Chinese, English, French
• Extend the ISIS CASE tool
• To deal explicitly with Generic and Specific
  concepts
• Through and XML (OWL?) description of the domain
• To perform automatic GUI generation
• So as to ease a strong interaction with users

40
Contact ????

• Michel SIMONET
• Michel.Simonet_at_imag.fr
• Didier GUILLON
• Didier.Guillon_at_agduc.com
• Dr Haijin YU ???
• Kidney_rj_at_medmail.com.cn

41
GENNERE partners
Paris - NECKER
Grenoble TIMC IMAG
Paul LANDAIS
Michel Ana SIMONET Didier GUILLON
Belgium - RAMIT
Grenoble - AGDUC
Georges de MOOR
Michel FORET Philippe GAUDIN
Shanghai RUI JIN
Nan CHEN
Wen ZHANG
42
Results (1)
43
Results (2)