description logics in rte

38
Description Logics in RTE Kilian Evang Introduction SHOIN(D) Individual Names Roles Concepts Terminological Axioms Assertions Concrete Domains Comparison Reasoning for Concepts for Knowledge Bases [Bedaride, 2003] T and H Background Knowledge ABox Saturation Subgraph Detection Back Matter Description Logics in RTE Kilian Evang 2009-07-20

Upload: kilian-evang

Post on 18-Dec-2014

364 views

Category:

Technology


0 download

DESCRIPTION

 

TRANSCRIPT

Page 1: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Description Logics in RTE

Kilian Evang

2009-07-20

Page 2: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Description Logics

I a family of logics

I origins in research on knowledge representation systems

I widely used in practice, notably in Semantic Webtechnology

I address expressivity-tractability tradeoff: adequateknowledge representation, useful inferencing

I basic standard DL called ALI degree of expressivity of a DL can be expressed in terms

of additional constructs added to AL

Page 3: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Individuals, Concepts, Roles

[Horridge et al., 2007], p. 13

Page 4: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

SHOIN (D)

I chosen here because the XML description languageOWL DL is based on it

I OWL DL and its subset OWL Lite widely used inSemantic Web technology

I extends ALC of [Bedaride, 2003] by several constructs

Page 5: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Expressions in SHOIN (D)

I individual namesI example: paulI denote individuals aka objects

I concepts (aka classes)I example: PersonI denote sets of individuals

I roles (aka properties)I example: hasChildI denote binary relations between individuals, i.e. sets of

ordered pairs of individuals

I formulasI terminological axiomsI assertions

Page 6: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Interpretations

An interpretation I consists of

I a domain ∆I of individuals andI an interpretation function ·I that maps

I individual names to elements of ∆I

I concept descriptions to subsets of ∆I

I role descriptions to subsets of ∆I ×∆I

Page 7: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Individual Names

Syntax: aSemantics: aI ∈ ∆I

Example: paulUnderstand: “the individual named paul”

Unique name assumption: an interpretation assigns eachindividual name a different individual.

Page 8: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Atomic Roles

Syntax: RSemantics: RI ⊆ ∆I ×∆I

Example: hasChildUnderstand: “the set of all parent-child pairs”

Example: isChildOfUnderstand: “the set of all child-parent pairs”

Page 9: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Inverse Roles

Syntax: R−

Semantics: {(x, y) | (y, x) ∈ RI}

Example: hasChild−

Understand: “the set of all child-parent pairs”

Example: isChildOf−

Understand: “the set of all parent-child pairs”

Page 10: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Atomic Concepts

Syntax: ASemantics: AI ⊆ ∆I

Example: PersonUnderstand: “the set of all persons”

Page 11: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Conjunction

Syntax: C u DSemantics: (C u D)I = CI ∩ DI

Example: Person u FemaleUnderstand: “the set of all female persons”

Page 12: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Disjunction

Syntax: C t DSemantics: (C t D)I = CI ∪ DI

Example: Doctor tGardenerUnderstand: “the set of all doctors and gardeners”

Page 13: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Negation

Syntax: ¬CSemantics: (¬C )I∆I \ CI

Example: ¬FlowerUnderstand: “the set of all individuals that aren’t

flowers”

Page 14: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Exists Restriction

Syntax: ∃R.CSemantics: (∃R.C )I = {x | ∃y((x , y) ∈ RI ∧ y ∈ CI)}

Example: ∃hasChild.PersonUnderstand: “the set of all individulals that have a

child which is a person”

Page 15: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Number Restrictions

Syntax: > nP, 6 nPSemantics: (> nP)I = {x | |{y | (x , y) ∈ PI}| > n}

(6 nP)I = {x | |{y | (x , y) ∈ PI}| 6 n}

Example: > 3hasChildUnderstand: “the set of all individuals with at least

three children”

Page 16: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Value Restriction

Syntax: ∀R.CSemantics: (∀R.C )I =

{x | ∀y((x , y) ∈ RI → y ∈ CI)}

Example: ∀hasChild.FemaleUnderstand: “the set of all individuals all of whose

children are female (including allindividuals without any children)”

Page 17: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Nominals

Syntax: {o1, . . . , on}where o1, . . . , on are individual names

Semantics: {o1, . . . , on}I = {oI1 , . . . , oIn }

Example: {china, france,russia,uk,usa}Understand: “the set of the permanent members of

the UN security council”

Page 18: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

The Universal Concept and the Bottom Concept

Syntax: >Semantics: >I = ∆I

Syntax: ⊥Semantics: ⊥I = ∅

Page 19: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Inclusions

Syntax: C v D (R v S)Semantics: An interpretation I

satisfies C v D (R v S)iff CI ⊆ DI (RI ⊆ SI).

Example: Apple v FruitUnderstand: “Every apple is a fruit.”

Example: hasTopping v hasIngredientUnderstand: “Having something as a topping also

means having it as an ingredient.”

Page 20: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Equalities

Syntax: C ≡ D (R ≡ S)Semantics: An interpretation I

satisfies C v D (R v S)iff CI = DI (RI = SI).

Example: SpicyPizza ≡Pizza u ∃hasTopping.SpicyTopping

Understand: “A SpicyPizza is defined to be a pizzawith a spicy topping.”

Example: isChildOf ≡ hasChild−

Understand: “isChildOf is defined to be the inverserole of hasChild.”

Page 21: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Transitive Roles

Syntax: R ∈ R+

Semantics: RI = (RI)+

Example: isPartOf ∈ R+

Understand: “If A is a part of B and B is a partof C, then A is also a part of C.”

I important for part-whole descriptions

I allows for defining concepts that have no finite model[Sattler, 1996]

Page 22: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Concept Assertions

Syntax: C (a)Semantics: An interpretation I satisfies C (a) iff

aI ∈ CI .

Example: Father(peter)Understand: “Peter is a father.”

Page 23: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Role Assertions

Syntax: R(a, b)Semantics: (a, b)I ∈ RI

Example: hasChild(mary,paul)Understand: “Paul is a child of Mary.”

Page 24: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Concrete Domains

Rouhgly and intuitively, concrete domains are a languageextension that allows for “importing”

I “individuals” such as 18,√

2, "Zwolf Boxkampfer",or "Zwo"

I “roles” such as greaterThan or startsWithfrom worlds such as arithmetic or string manipulation intothe logic. OWL DL uses this to assignnumeric/string/date/... properties to individuals.

Page 25: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Comparison of Four DLs

construct AL ALC S SHOIN (D)

atomic negation X X X Xconjunction X X X Xuniversal quantification X X X Xexistential quantification limited X X Xdisjunction X X Xtransitive roles X Xnumber restrictions Xrole hierarchies Xinverse roles X

Page 26: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Knowledge Bases

I a knowledge base is a set of formulas (explicitknowledge)

I sometimes divided up into two subsets:I TBox

I contains only terminological axiomsI provides a general terminology

I ABoxI contains only assertionsI provides a specific world description

I also contains implicit knowledge

I implicit knowledge can be made explicit by reasoning

Page 27: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

An Example Knowledge Base

TBox

Woman ≡ Person u Female

Man ≡ Person u ¬Woman

Mother ≡ Woman u ∃hasChild.Person

Father ≡ Man u ∃hasChild.Person

Parent ≡ Father tMother

Grandmother ≡ Mother u ∃hasChild.Parent

MotherWithManyChildren ≡ Motheru > 3hasChild

MotherWithoutDaughter ≡ Mother u ∀hasChild.¬Woman

Wife ≡ Woman u ∃hasHusband.Man

ABoxhasChild(mary, paul), Father(paul)

An example piece of implicit knowledge

Grandmother(mary)

Page 28: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Modelhood

An interpretation I is a model of (satisifies)

I a formula φ iff it satisfies φ.

I a TBox T iff it is a model of every terminological axiomin T .

I an ABox A iff it is a model of every assertion in A.

I an ABox A with respect to a TBox T iff it is a modelof both A and T .

I a concept C iff CI is nonempty.

Page 29: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Reasoning Tasks for Concepts

Let C ,D concepts and T a TBox (e.g. see above).I C is satisfiable wrt. T iff C and T have a common

model.I e.g. not satisfiable: Man uWoman

I C is subsumed by D wrt. T iff CI ⊆ DI for everymodel I of T .

I e.g. Mother is subsumed by WomanI C and D are equivalent wrt. T iff CI = DI for every

model I of T .I e.g. ∃hasChild.Person is equivalent to FathertMother

I C and D are disjoint wrt. T iff CI ∩ DI = ∅ for everymodel I of T .

I e.g. Man and Woman are disjoint

Page 30: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Reasoning Tasks for Knowledge Bases

Let K a knowledge base.I consistency checking: K is consistent iff it has a

model.I e.g. above KB is consistent, adding Mother(paul)

would make it inconsistent

I instance checking: Given a concept C and anindividual name a, K entails C (a) iff K ∪ {¬C (a)} isinconsistent.

I e.g. Grandmother(mary) is entailed by above KB

I retrieval problem: Given a concept C , find allindividual names a such that K entails C (a).

I e.g. the result for ∃hasChild.Person would be {mary}I realization problem: Given an individual name a, find

the most specific concepts C such that K entails C (a).

I ...

Page 31: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

[Bedaride, 2003]: RTE in Four Steps

I RTE in four steps:

1. represent T and H as two ABoxes2. make a TBox with background knowledge3. saturate ABoxes with TBox4. subgraph-detect ABox H in ABox T

I Example T/H pair:I T: “John buys a cat at the pet shop for 50 euros.”I H: “A shop sells an animal to John.”

Page 32: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Step 1: Represent T and H as Two ABoxes

I ABox T = {CommercialTransaction(ct1), John(j1),PetShop(ps1),Cat(c1), 50Euros(p1), buyer(ct1, j1),seller(ct1,ps1), goods(ct1,c1),money(ct1,p1)}

I ABox H = {CommercialTransaction(ct2), John(j2),Shop(s2),Animal(a2),buyer(ct2, j2),seller(ct2, s2), goods(ct2,a2)}

I Note:I FrameNet frames and frame elements represented as

individuals, characterized by concept assertionsI connected via frame-specific rolesI no difference made between common/proper,

definite/indefinite, singular/plural NPI each ABox has its own set of individual names

Page 33: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Step 2: TBox with Background Knowledge

I ABox T = {CommercialTransaction(ct1), John(j1),PetShop(ps1),Cat(c1), 50Euros(p1), buyer(ct1, j1),seller(ct1,ps1), goods(ct1,c1),money(ct1,p1)}

I ABox H = {CommercialTransaction(ct2), John(j2),Shop(s2),Animal(a2),buyer(ct2, j2),seller(ct2, s2), goods(ct2,a2)}

I TBox BK = {PetShop v Shop,Cat v Animal}I Note:

I atomic concepts mapped to WordNet synsets (how –WSD?)

I for each pair (Sh,St) of synsets from H and T, check ifthere is a relation and if so,

I add the appropriate axiom(s) to the TBox: Sh v St forhyponymy, St v Sh for hypernymy, Sh v St andSt v Sh for synonymy, Sh v ¬St and St v ¬Sh forantonymy

Page 34: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Step 3: Saturate ABoxes with TBox

I TBox BK = {PetShop v Shop,Cat v Animal}I ABox T ′ = {CommercialTransaction(ct1), John(j1),

PetShop(ps1),Cat(c1), 50Euros(p1), buyer(ct1, j1),seller(ct1,ps1), goods(ct1,c1),money(ct1,p1),Shop(ps1),Animal(c1)}

I ABox H ′ = {CommercialTransaction(ct2), John(j2),Shop(s2),Animal(a2),buyer(ct2, j2),seller(ct2, s2), goods(ct2,a2)}

I Note:I T ′ (H ′) is T (H) saturated with BK , i.e. containing

every assertion entailed by BK ∪ T (BK ∪ H)

Page 35: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

Step 4: Subgraph-Detect H ′ in T ′

I Let σ = {ct2/ct1, j2/j1,a2/c1, s2/ps1}I ABox T ′ = {CommercialTransaction(ct1), John(j1),

PetShop(ps1),Cat(c1), 50Euros(p1), buyer(ct1, j1),seller(ct1,ps1), goods(ct1,c1),money(ct1,p1),Shop(ps1),Animal(c1)}

I ABox H ′σ = {CommercialTransaction(ct1),John(j1),Shop(ps1),Animal(c1),buyer(ct1, j1),seller(ct1,ps1), goods(ct1,c1)}

I Note:I We detect entailment iff we can find a individual name

substitution σ such that H ′σ ⊆ T ′, i.e. all informationin H ′ is also in T ′.

Page 36: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

References

Franz Baader, Diego Calvanese, Deborah L. McGuiness,Daniele Nardi and Peter F. Patel-Schneider (2003)The description logic handbook: theory, implementation,and applicationsCambride University Press

Paul Bedaride (2003)Using Description Logics for Recognising TextualEntailmentIn: Proceedings of the Twelfth ESSLLI Student Session

Page 37: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

References

Matthew Horridge, Simon Jupp, Georgina Moulton,Alan Rector, Robert Stevens and Chris Wroe (2007)A Practical Guide to Building OWL Ontologies UsingProtege 4 and CO-ODE Tools, Edition 1.1

Ulrike Sattler (1996)A concept language extended with different kinds oftransitive rolesSpringer

Page 38: Description Logics in RTE

Description Logicsin RTE

Kilian Evang

Introduction

SHOIN (D)

Individual Names

Roles

Concepts

TerminologicalAxioms

Assertions

Concrete Domains

Comparison

Reasoning

for Concepts

for Knowledge Bases

[Bedaride, 2003]

T and H

BackgroundKnowledge

ABox Saturation

Subgraph Detection

Back Matter

RteClassMember v ∃thanks−.{kilian}