ByHelen Howell

2

Outline

What is interoperation? Why? Problems Approach to Interoperations ONION Cyc

3

What is Interoperation?

Interoperation may be defined as a loose coupling across information sources, semantic metadata descriptions and ontologies

4

Why?

Share information

5

Domain Differences

Naming attribute items differently payroll - EMP; personnel - PEOPLE

Scope payroll may include suport for student

benefits for employee’s children; personnel - employees only

6

Domain Differences (cont.)

Encoding ssn with and without hypens

Attribute scopes-Semantic hetergeneity

hot has different meaning in weather domain than in truck-engine domain

7

Integration Problems

Data that is available is subject to implicit assumptions which means that it cannot be reused out of context.

Data is often duplicated and labeled under different terms.

e.g., Address verses Addr

8

Approaches

Direct Translation

Single Shared Ontology

Multiple Shared Ontology

9

Direct Translation

For each ontology a set of translating functions is provided to allow mappings with the other ontologies.

10

Direct Translation

Only feasible if a few ontologies OBSERVER - mappings between a

term in a source ontology and its synonym in a destination ontology.

Set up relationships e.g., Volcano effects Environment

11

Single Shared Ontology

Often standards are not convenient to use since they have to be suitable for all potential uses. So a single subset is formed as a standard

12

Multiple Shared Ontologies

Locates shared knowledge in multiple but smaller shared

ontologies

13

Multiple Shared Ontologies

Referred to as Resource Clustering or Ontology Clustering

Approach more flexible and scalable No longer have to commit to a

comprehensive ontology

14

Multiple Ontologies

Normally organized into a hierarchy Top level provides general definitions

share by all resources Lower levels provide definitions of

concepts that extend and characterize upper levels

15

Mediation-Based Information Systems

Mediators provide intelligent middleware services- matching resources to applications- resources are autonomous and

heterogeneous

16

Mediation Based Information Systems

Identify most relevant resources for user’s needs

Retrieve the relevant data Translate it to a common

representatione.g., Nevada, NV, NV USA

Interoperation of Information Sources via Articulation of Ontologies

Prasenjit Mitra, Gio Wiederhold

Stanford University

Supported by AFOSR- New World Vistas Program

18

Interoperation of Information Sources

Compose information Multiple independent, heterogeneous

sources Reliability, scalability Semantic heterogeneity - same term different semantics - different term same semantics

19

Onion System

20

Articulation Generator

Driver

Ont1 Ont2

Phrase Relator

Thesaurus

StructuralMatcher

SemanticNetwork

Human Expert

Context-basedWord Relator

OntA

21

Preliminaries: Ontology

Ontology - hierarchy of terms and specification of their properties.

Modeled as a directed labeled graph + set of rules.

Ont = ( V, E, R) V - set of nodes(concepts) E - set of edges(properties) R - set of rules involving V,E

22

Articulation Rules

Articulation rules - logic-based rules that relate concepts in two ontologies:

- Binary Relationships (O1.Car SubClassOf O2.Vehicle) (O1.Buyer Equ O2.Owner) (O2.LuxuryCar SubClassOf O1.Car)

23

Example

Car

H123

Instance

List_PriceVID

PMitra

NameUnit

$ 40k CA

AddressAmount

Vehicle

LuxuryCar

PMitra

BuyerInstance

SubClass

Instance

VIN

H123

Owner Retail_Price

CA

Name Addr

$35k

SubClass

SubClass

BuyerOwner

Attribute

Attribute

Equ

24

Articulation Rules (contd.)

Horn Clauses - (O1.Car O1.Instance X), (X O1.Price Y), (Y > $30000) => (O2.LuxuryCar Instance X) - (V O2.Retail_Price P), (C O1.List_Price

L), (L Unit U), (L Amount A), (V Equ C) => (P Equ concat(U,A))

25

Contributions

Articulation Generation Toolkit - produce translation rules semi-

automatically - a library of reusable heuristic methods - a GUI to display ontologies and interact with the expert Ontology Algebra - query rewriting and planning

26

Articulation Generation Methods

Non-iterative Methods - Lexical Matcher - Thesaurus-based Matcher - Corpus-based Matcher - Instance-based Matcher Iterative Methods - Structural Matcher - Inference-based Matcher

27

Lexical Methods

Preprocessing rules. -Expert-generated seed rules. e.g., (O1.List_Price Equ O2.Retail_Price) -Context-based preprocessing directives. e.g., (O1.UK_Govt Equ O2.US_Govt) -Stop-word Removal & Stemming -Word match (full or partial) -Phrase match e.g., (O1.Ministry_Of_Defence Equ O2.Defense_Ministry) 0.6

28

Thesaurus-based methods

Consult a dictionary/thesaurus to find synonyms, related words

Generate a similarity measure or relatedness measure

- words that have similar words in their definitions are similar

Get more semantically meaningful relationships from WordNet (syn, hyper)

29

Candidate Match Repository

Term linkages automatically extracted from 1912 Webster’s dictionary *

* free, other sources . being processed.

Based on processing headwords definitions Notice presence

of 2 domains: chemistry, transport

30

Corpus-based

Collect a set of text documents preferably from same domain

- search using keywords in google Build a context vector (1000-character

neighbourhood) for each word Compute word-pair similarity based on the

cosine of the vectors Use word-pair similarity to find similarity

among labels of nodes/edges

31

Structural Methods

Uses results of lexical match If x% of parent nodes match & y% of

children nodes match Special relations (AttributeOf) match

32

Tools to create articulations

Graph matcherforArticulation- creatingExpert

Vehicle ontology

Transport ontology

Suggestionsfor articulations

33

continue from initial point

Also suggest similar terms for further articulation:

• by spelling similarity,• by graph position• by term match repository

Expert response:1. Okay2. False3. Irrelevant to this articulation

All results are recorded

Okay ’s are converted into articulation rules

34

An Ontology Algebra

Operations can be composed

Operations can be rearranged

Alternate arrangements can be evaluated

Optimization is enabled

The record of past operations can be

kept and reused when sources change

35

Binary Operators

A knowledge-based algebra for ontologies

The Articulation Function (ArtGen), given two ontologies, supplies articulation rules between them.

Intersection create a subset ontology keep sharable entries

Union create a joint ontology merge entries

Difference create a distinct ontology remove shared entries

36

Intersection: Definition

O1 = (V1, E1, R1) O2 = (V2, E2, R2) OI = ( O1 IntArtGen O2) = (VI, EI, RI)

Arules = ArtGen( O1, O2 ) VI = Nodes( Arules ) EI = Edges( Arules ) + Edges(E1, VI.V1) +

Edges( E2, VI.V2) RI = Arules + Rules( R1, VI.V1) + Rules( R2,

VI.V2)

37

Intersection: Example

ARules = { (O1.Car SubClass O2.LuxuryCar), (O1.MSRP Equ O2.Price)} NI = ( O1.Car, O1.MSRP, O2.LuxuryCar, O2.Price ) EI = Edges(ARules) + {(O1.Car Attribute O1.MSRP), (O2.LuxuryCar Attribute Price)}

InexpCar

Car LuxuryCar

MSRP Price LuxuryTax

SubClass

Equ

O1 O2OI

38

Intersection: Properties

Commutative? OI12=(O1 IntArtGen O2) = (O2 IntArtGen O1)=OI21

VI12 = VI21, EI12 = EI21, RI12=RI21

ARules12 = ARules21

ArtGen(O1, O2) = ArtGen(O2, O1)

IntArtGen is commutative iff ArtGen is commutative

39

Semantically Commutative

Example: ArtGen(O1,O2) : ( O1.Car SubClassOf O2.Vehicle) ArtGen(O2,O1) : ( O2.Vehicle SuperClassOf O1.Car)

Defn: ArtGen is Semantically Commutative iffArtGen(O1,O2) <=> ArtGen(O2,O1)

Operands to intersection can be rearranged if ArtGen is semantically commutative

40

Associativity Example

Example: ArtRules(O1, O3) : (O1.Car

SubClassOf O3.Vehicle) ArtRules(O2, O3) : (O2.Truck

SubClassOf O3.Vehicle) ArtRules(O1, O2) : null Intersection is not associative!

41

Associativity

(O1.O2).O3 = O1.(O2.O3) iff ArtGen is consistent and

transitively connective

consistent - given two nodes it generates the same relationship irrespective of relationships between other nodes

42

Associativity (contd.)

ArtGen relates A and B: (A Rel B) = (A R1 B) or (B R2 A)

Transitively connective: If ArtGen generates (A Rel B), (B Rel C) then it also generates (A Rel C’) where A O1, B O2, C,C’ O3

43

Using the Articulation Rules

Q(X) :- (LuxuryCar Instance X), (X Owner Y), (Y Addr `CA’)

Translated using Articulation Rules Q(X) :- (Car Instance X),(X List_Price Y) (Y Unit ‘$’),(Y Amount A), (A > 30000), (X Buyer Z), (Z Address ‘CA’)

44

Conclusion

Introduced an Ontology Interoperation System.

Interoperation based on articulations that bridge the semantic gap.

Graph-oriented data model, logical rules Founded on Ontology Algebra

Thanks to Stefan Decker, Alex Carobus, Jan Jannink, Sergey Melnik, Shrish Agarwal

DAMLOntologyInteroperationUsing Cyc

Steve ReedCycorp

46

DAML Excitement at Cycorp

Promoting DAML ontology interoperability through Cyc, as a reference ontology.Cyc now imports and exports DAML, paraphrases and reasons about it.

Cyc lexical tools assist the mapping of imported DAML ontologies into Cyc.

Both UDDI categorization schemes imported as DAML into Cyc:

–UNSPSC (via KSL) and NAICS

Cycorp is featuring DAML in it’s upcoming release of OpenCyc (this fall).

47

DAML Challenges at Cycorp

How to automate mapping and translation of DAML ontologies.

How to get the user community to help map numerous DAML terms -- interest growing at OpenCyc.Importation of large DAML schema is daunting but underway, WordNet, OpenDirectory, UNSPSC, NAICSProvide an inference engine for the Semantic Web - subsumption, classification, constraints, well-formed statement validation, diagnostics, justifications.What web services interface? SOAP likely.

DAML Ontology Interoperation

CycOntology

UNSPSCOntology

NAICSOntologyMapping 1:

UNSPSC to CycMapping 2:

NAICS to Cyc

User

Search for: “waterproof plywood”

Search for:UNSPSC/7311

Hardwood Int.… marine plywood …

UDDI Web Services

Return: marine plywood by Hardwood Int.

49

Ontology Mapping

From Doerr, Semantic Problems of Thesaurus Mapping

DAML Import

50

Ontology Mapping

From Doerr, Semantic Problems of Thesaurus Mapping

DAML Import

Cyc reference

51

UNSPC/NAICS Mapping

Universal Standard Products and Services Classification, North American Industrial Classification SystemBoth business taxonomies are monohierarchies, entities should have one code assigned for them.

Cyc: #$MonohierarchyClassificationSystem implies that each collection is #$partitionedInto its narrower terms.

Mapping is targeted at facets of the reference Cyc terms via functional composition.

52

Mapping to Semantic Facets

(#$isa (#$BusinessFacetFn #$HardwoodInternational #$Hardware) #$NAICS-Hardware-Store))

• Hardwood International is an online store, selling mainly lumber but offering hardware as well, but NAICS is a monohierarchy.

Hardwood International

44413

Hardware

(#$isa (#$BusinessFacetFn #$HardwoodInternational #$Hardwood) #$NAICS-Other-Building-Materials-Dealers))

44419

Hardwood

Business facets

53

Conclusion

Ontology Interoperation is not without it problems

Hetergeneous data can work together

ByHelen Howell