Title: Ontology Work Guide for Web Contents Principle and Methods proposal for new work item in ISO TC37
1Ontology Work Guide for Web Contents- Principle
and Methods(proposal for new work item in ISO
TC37)
- ISO TC37/SC4 N435
- Nov 12, 2007
- Presented by Miran Choi/ETRI
- Written by Jae Sung Lee/Chungbuk National Univ.
2- Contents
- Motivation
- Ontology Construction for Web Contents
- Related Standards and Ontology Work Scope
3Motivation
4Semantic search
- Limit of keyword search
- ambiguous keywords in query and docs
- retrieves too many redundant documents
- waste time on finding what we want
- Semantic search
- Ideally no ambiguity in query and text
- retrieve docs with more accuracy by use of
semantic information - need to solve technical problems
- Query method
- Semantic tagging
- Globally common ontology construction
5Various ontologies
- Various ontologies are being built
- in many areas and applications
- Classification of ontology is various
- Different levels of ontologies
- Lightweight ontology
- Simple metadata level, or concepts with
hierarchy, or small number of axioms - E.g. term list, MDR, ISOCat, Topic Map, SUMO
- Heavyweight ontology
- Delicate definitions and first order logic for
inference - E.g. BFO, DOLCE
6Evolving ontologies
- Ontologies are growing in their own ways
- Various collections of ontology
- Ontology library
- e.g. SWAG, DAML ontology library
- Ontology registry
- e.g. MDR, ISOCat
- Etc.
- For the global search
- Semantic harmonization methods are needed by
sharing standard principle and methods - Global coordination is needed for the various
standards
7Standardization of construction procedure
- Needs
- To make ontologies consistent and compatible each
other - To promote reuse of existing ontologies
- To avoid conflicting definitions
- To lessen confusion in semantic contents search
- The sooner, the less confusion!
8Ontology construction for web contents
9Web contents for semantic search
- Web contents characteristic
- Various topics and technical levels
- But, target to the general, less technical, not
too specific contents. - Ontology for semantic search
- Less strict meaning definition than the ontology
for interoperability - A little semantic difference is tolerable
- Lightweight ontology is enough
- Use controlled vocabulary or shared terms
- Include some basic relationship if needed
- But no complicated relationship
10Using other ontologies
- A guide is needed to use existing ontologies
- May need a ontology for ontologies to look up.
- May need to filter out only lightweight
features except other complex relations or
axioms. - Consulting existing ontologies
- ontology library
- E.g. Daml ontology library, Semantic Web
Agreement Group - ontology registry
- E.g. Meta data registry, Data category registry
(ISOcat) - ontology homepage/document
- E.g. SUMO
11Construction of new ontology
- General construction guide can be derived from
other guides - Terminology work(ISO 704)
- Building ontologies and knowledge elicitation
(Rector et al) - And others
- General steps of ontology construction
- Classify the target objects
- Define the concept of the objects
- Concepts are defined by only essential
characteristics - Define relationship between concepts
- Implementations and evaluation
12Ontology formats/languages
- Standard Ontology Language
- Choose one of the languages
- RDF(S), OWL, UML, SCL, DL etc
- Or, define mapping between them by using the
following standards - Ontology Definition Metamodel (by OMG)
- Metamodel Framework (MMF)
- Naming convention (designations)
- Terms
- sometimes ambiguity problem
- Numbers
- E.g. Published Semantic Indicators
- Concatenated terms (ISO/IEC 11179)
- E. g. PersonGivenName
- Harmonized method?
13Publicity
- Central ontology registry
- All new ontologies should be known to
applications and other users. - Central registry keeps the ontologeis or
reference points. - Central registry can be accessed by all.
- Need to keep all other compatible ontologies list.
14Related standardsand ontology work scope
15Semantic Web
- Represent semantic information using URI triples.
- A triple represents subject, property (relation)
and object. - Make inference based on the triples
-
- Notation 3 example
- ltgt lthttp/purl.org/dc/elements/1.1/creatorgt
_x0. - _x0 lthttp//xmlns.com/0.1/foaf/namegt Jae
Sung Lee. - Ontology is built by using URI triples.
- Interoperable among agents sharing ontology
- Problem various ontology metadata
- use their own triples with various ontology
(namespace) - interoperable in a domain, but not in a global
scale.
16Common Ontology
- One universe common ontology for interoperability
- Ideal but not practical now
- No global ontology
- Web is not static
- Needs semantic reconciliation because...
- agreed standard only address a small body of
knowledge - must accommodate prior resources to standard
- new work will have to go beyond standard
- Consensus ontology
- if have sufficient overlap under the same
universe of discourse - then reconcile ontologies
- Merging ontology is difficult
- concept mapping will happen 1-1, n-1, n-n
- value mapping is not always consistent
17W3C Ontology
- Ontology languages for standard format
- RDF
- represents graph for structure
- RDFS
- use specialized vocabulary and primitive classes
and properties - OWL
- more precise expressive than RDFS
- Evolutionary ontology
- Keep systems interoperable using partial
understanding and transformability - For common understandable terms, third party
databases are needed eg. SWAG (Semantic Web
Agreement Group) - SWAG keeps namespaces rdf, foaf, dc and etc.
- DAML ontology library keeps various ontologies.
18IEEE SUO/SUMO
- IEEE SUO
- limited to concepts that are meta, generic,
abstract and philosophical - general enough to address (at a high level) a
broad range of domain areas - provide a structure and a set of general concepts
upon which domain ontologies could be
constructed. - IEEE SUMO/MILO
- Largest formal public ontology
- Suggested Upper Merged Ontology
- MId-Level Ontology
- Communications, Countries and Regions,
distributed computing, Economy, Finance,
Engineering components, Geography, Government... - domain ontologies could be constructed based on
this.
source http//www.ontologyportal.org
19Topic Map
- For interoperability between ontologies
- Naive approach
- make a new vocabulary not practical
- PSI approach (Published Subject Indicators)
- maps equivalent terms to a unique ID.
- OASIS
- builds large PSI database for interoperability
source http//www.xml.com/pub/a/2002/09/11/topicm
aps.html
20ISOCat
- ISO TC37 standards
- DCR keeps all data categories.
- Each app DCS can select subset of fields.
- Each apps DCS can select subset data units.
21ISO/IEC 11179 and MDR
- MDR (meta data registry)
- stores data elements (both semantics and
representations) - The semantic areas describe precise definitions
- The representational areas define how the data is
represented in a specific format such as XML - ISO/IEC 11179 for data elements
- Registration guidelines
- Naming and Identification Principles
- Formulation of Data Definitions rules
- Classification Scheme
22Small Business Vocabulary and Business Rules
- SBVR is targeted at business rules and business
vocabularies - Model Driven Architecture
- SBVR meta model is automatically generated from
SBVR vocabularies - SBVR meta model provide standardized data
interfaces and data interchange. - A business vocabulary contains all the terms and
concepts in business.
source SBVR adopted specification
23Metamodel Framework (MMF)(ISO 19763)
- Software engineering approach
- Ontology is represented in application model
- Model driven ontology build
- Goal
- to harmonize metamodel technology and contents of
metamodel - provide interoperability by using reference
ontology - Core model of the MMF
- provides a mechanism for describing each
different metamodel in local registries - enables registration of those to the registry
- MMF model mapping
- register mapping rules
- enable the federation among different registries
24Metamodel Framework Architecture
source ISO/IEC CD 19763-01200x(E)
25OMG ODM(Ontology Definition Metamodel)
- Includes metamodels and defines mapping between
them - RDFS, OWL, UML, SCL, ER, TM, DL metamodels
- Concentrates on most widely applicable and most
readily achievable goals - use case analysis to 3 major clusters of apps
business, analytic, engineering app.
26Structure of ODM
source OMG/RFP ad/05-01-01
27Relation between ODM and MMF
- Each ODM metamodel is registered to MMF
source OKABE, Masao presentation in KIPONTO 2005
28Position of standards and our workscope
Meta level
MMF
ODM
SUO/SUMO MILO
TM
Ontology work guide for web contents
MDR
Domain specific
SBVR
ISOCat
General and language based
Application model driven
29Thank you very much! ?????!