The Integration of Ontologies Using OWL

1

• The Integration of Ontologies Using OWL
• Presented by Steven Arnett, Chief
• U.S. National Codification Bureau

2
Overview

• DLIS is sponsoring a research project being
  carried out by Lehigh University
• The purpose of the project is to evaluate the use
  of the Web Ontology Language (OWL) as an exchange
  platform between different taxonomies
• Objective To align the Federal Catalog
  System/NATO Codification System (FCS/NCS)
  taxonomy with supplier taxonomies so as to
  provide DLIS and NATO with fuller access to
  suppliers and their products
• Work began in October 2007 and will be conducted
  over an 18 month period

3
Project Team and Responsibilities

• Lehigh University
• Project Management
• Ontology Development
• Taxonomy Integration
• Compiler Development
• ECCMA
• Development of terms and definitions
• Inclusion of taxonomy terminology in eOTD
• Facilitation of OWL output from eOTD
• CTC
• Technical guidance on data modeling, eOTD core
  model, and ISO 22745
• ISO/IEC JTC1/SC32 metadata standards
• ISO TC37 terminology standards

4
The Problem

• DLIS and NATO must be able to integrate their
• item taxonomies with commercial systems
• Commercial cataloging systems have
  domain-specific taxonomies that are not
  interconnected
• Information is stored in different, often
  proprietary, and incompatible formats
• Data elements have to be translated and
  transferred manually among different
  systems
• The result is a significant cost of time and
  money

5
Other Taxonomies

• United Nations Standard Products and Services
  Code (UNSPSC)
• Standardized Material and Service Classification
  (eCl_at_ss)
• Common Procurement Vocabulary (CPV)
• Parts Library (PLIB)

6
Approach

• Use the ECCMA Open Technical Dictionary (eOTD) as
  the basis for the integration of taxonomies
• The eOTD is the industry version of the FCS/NCS
  and seeks recognition as the international
  standard for e-catalogs via the ISO 22745
  designation

7

Interactions with Other Classifications

• Mapping keywords is insufficient
• one-to-one correspondences arent always possible
• overlapping classes
• functional vs. compositional taxonomies
• e.g. grinding machinecutlery (0161-101-0070711)
  vs. grinding machinecarbide tool bit
  (0161-101-0070981)
• Need deeper analysis
• Focus on systemic classification of attributes
  that connects communities of information
• This connection can be established with the
  Semantic Web

8
Limitations of XML

• XML has limited semantics
• in schema comments
• in external English definitions
• implied by names
• Integration of different XML schemas cannot be
  easily automated
• how do I know if ltdiametergt means inside
  diameter or outside diameter?

9
The Semantic Web

• Definition
• The Semantic Web is not a separate Web but an
  extension of the current one, in which
  information is given well-defined meaning, better
  enabling computers and people to work in
  cooperation. (Berners-Lee et al., Scientific
  American, May 2001)
• Key International Standards
• World Wide Web Consortium (W3C) recommendations
• Resource Description Framework (RDF)
• Web Ontology Language (OWL)
• same status as HTML / XML

10
Comparison of Database and Semantic Web

• SEMANTIC WEB
• Organizing knowledge using
• Ontology Language (OWL)
• Relating Objects with inheritance and properties
• Semantic Query
• Benefit
• Reflects the actual model of objects and their
  relationships
• Ontology Integration (loosely)

• DATABASE
• Organizing Data
• - Tables
• Relating Data with keys
• - Primary and Foreign keys
• Syntactic Query
• Problem
• - Does not reflect the actual model of Objects
  and their relationship
• - Data Integration (tightly)

11
RDF and RDF Schema
ltrdfProperty rdfIDnamegt ltrdfsdomain
rdfresourceVehiclegt lt/rdfPropertygt ltrdfsCl
ass rdfIDWarVehiclegt ltrdfssubclassOf
rdfresource http//schema.org/genVehicl
egt lt/rdfsClassgt
rdfsClass
rdfProperty
rdftype
rdftype
gVehicle
rdftype
rdfsdomain
rdfssubclassOf
ltrdfRDF xmlnsghttp//schema.org/gen
xmlnsmhttp//schema.org/milgt
ltmWarVehicle rdfIDbradleygt
ltgnamegtBradley M2A3lt/gnamegt
lt/mWarVehiclegt lt/rdfRDFgt
mWarVehicle
gname
rdftype
gname
Bradley M2A3
bradley
12
OWL

• Web Ontology Language
• data is linked to ontologies that provide
  semantics
• syntax based on RDF
• but richer semantics
• can describe more complicated relationships
• semantics based on description logic (DL)

• A ToolKit is a subset of the set of objects which
  only have Tools as parts

ltowlClass rdfIDToolKitgt
ltrdfssubClassOfgt ltowlRestrictiongt
ltowlonProperty rdfresourcehasPart /gt
ltowlallValuesFrom resourceTool /gt
lt/owlRestrictiongt lt/rdfssubClassOfgtlt/owlCla
ssgt
13
Description Logic (DL)

• Form of knowledge representation
• useful for formally defining classes
• studied extensively in 1990s
• mature reasoning software
• FaCT, RACER, etc.
• Benefits
• optimized computation of subsumption
• calculate implicit subClassOf relations
• ontology integration
• if two ontologies use class expressions to define
  their vocabularies in terms of a third ontology,
  then subsumption can be used to compute an
  integrated ontology

14
OWL Class Constructors
15
OWL Axioms
16
Overall Project Plan

• Select target domains and systems
• Create an OWL ontology for eOTD
• Create an OWL ontology for the FCS/NCS
• Create OWL ontologies for the target systems
• Demonstrate ontology integration
• Create OWL wrappers for the target systems
• Develop ontology translation compiler

17
Task 1 - Select Target Domain

• Select a limited number of domains for the
  prototype

• Selected Domains
• Microcircuits, Digital
• Batteries
• Bushings/Bearings
• Fasteners
• O-Rings

• Selected Target Taxonomies
• UNSPSC
• eCl_at_ss
• CPV
• PLIB

18
Task 2 eOTD Ontology

• Create eOTD ontology
• Extract relevant subset of the dictionary
• Normalize standard item names
• Write program to translate to RDF/OWL
• Automatically produce RDF/OWL version
• Augment ontology with formal definitions
• Try edit guides and manual approach using
  definitions

19
Task 2 eOTD Ontology

• Automatically convert eOTD to RDF
• Standard Item Names ? RDF classes
• Standard Attribute Names ? RDF properties
• Automatically create a taxonomy
• e.g., BUSHINGAUTOMOTIVE (FSC 5365)
• BushingAutomotive rdfssubClassOf Bushing
• note this will require human review
• consider ASPIRIN,ALUMINA,AND MAGNESIA
  TABLETSUSP (0161-101-0205151, FSC 6505 Drugs
  and Chemicals)

20
Task 3 FCS/NCS Ontology

• Automatically translate FCS/NCS to RDF
• FSGs/FSCs/AINs ?RDF Classes
• rdfsubClassOf relations between FSG and 1 FSC
• Automatically describe FSC using eOTD standard
  item names that have one or more FSCs listed in
  the eOTD
• Manually augment with detailed OWL descriptions
  using the eOTD ontology for primitives
• limited to domain selected in Task 1

21
Task 4 Target Systems Ontologies

• Automatically translate to RDF classes/properties
• exact method is system dependent
• e.g., given a relational database, each table
  maps to a class and attribute maps to property
• Create OWL definitions for each class
• use eOTD ontology for primitive terms
• scoped to just domain identified in Task 1

22
Task 5 Integration Demonstration

• Use a DL reasoner to integrate FCS/NCS ontology
  with the target ontologies
• compute which target classes are implicit
  subclasses of FCS/NCS classes (subsumption)
• automatically merges two taxonomies
• Intermediate validation
• quality of integration depends on quality of
  definitions in Tasks 2 4
• may lead to refinement of earlier DL descriptions

23
Task 5 Example
eOTD Ontology
eotdYoke eotdMainRotorAssembly
UNSPC Ontology
FCS/NCS Ontology
FCS/NCSclass1615 ? eotdRotorBlade
eotdMainRotorAssembly
DL Reasoner
24
Task 6 Wrappers for Targets

• Goal automatically integrate a supplier catalog
  into the FCS/NCS
• Wrapper
• software module that provides a standard
  interface to a legacy software program
• Create wrappers that convert actual output to
  corresponding RDF/OWL
• since our original ontologies are close in
  structure to the system, this is straightforward

25
Task 7 Translation Compiler

• Automatically generates translation code
• input source and target ontology
• output translator that converts source data to
  target data
• uses DL reasoning to determine relationships
• Compiler can be rerun any time source / target
  ontologies change
• infrequent, so speed isnt critical
• Translator is fast
• can be run as frequently as data changes

26
System Architecture
eOTD Ontology (OWL)
extends
extends
Ontology Translator Compiler
Supplier Ontology (OWL)
FCS/NCS Ontology (OWL)
commits to
commits to
Supplier Catalog
Supplier Catalog (OWL)
Supplier-to-FCS/NCS Translator
Translated Catalog(OWL)
Supplier Wrapper
27
Expected Benefits

• Creating ontologies for the FCS/NCS and
  commercial taxonomies provides semantically
  correct mappings for interoperability
• Sharing of information via ontologies provides
  greater item visibility
• Automatic translation from FCS/NCS ontology to
  target ontology expands supplier base
• Catalog consistency is promoted by committing
  to an ontology

28
Conclusion and Expectations

• This is a proof of concept project
• focuses on a small portion of FCS/NCS
• only considers taxonomy integration
• many possibilities for follow-on projects
• Its not magic, its good engineering!
• domain modeling
• international standards
• proven technology

29
Data is the DNA

• Data is the DNA of materiel management
• Acquisition
• Financial management
• Hazardous material
• Freight and packaging
• Maintenance
• Sustainability
• Disposal
• Demilitarization

In-Transit
In-Theater
In-Storage
Development
Disposal
NSN
UID
RFID