Presentation is loading. Please wait.

Presentation is loading. Please wait.

Bart Gajderowicz, Ryerson University, Dept. of Computer Science

Similar presentations


Presentation on theme: "Bart Gajderowicz, Ryerson University, Dept. of Computer Science"— Presentation transcript:

1 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Introduction Motivation for Description Logic (DL) DL and First Order Logic (FOL) What is Description Logic? DL Architecture Reasoning Structures Applications of DL March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

2 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Motivation Any area with information overload. Where amount of declarative information to be processed manually exceeds that of human abilities. Even if a large number of resources could be found, it would take an impractical number of human hours to process. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

3 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Motivation Need for greater push for standardization of terminologies: Semantic Web Health Services Bio-informatics March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

4 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Motivation We would like to model complex concepts using modern modeling structures. OWL, Web Ontology Language, is a computer processible syntax for expressive Description Logic. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

5 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Motivation Propositional Logic (PL) has well defined operators and axioms creating atomic formulas. LHS and RHS Equivalences, such as: Commutativity ϕ ∨ ψ Ξ Ψ ∨ϕ Tautology ϕ ∨Τ Τ Unsatisfiability ϕ ∧  Implication ϕ -> ψ  ϕ ∨ ψ March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

6 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Motivation We can already do a lot with PL But … No access to the structure of atomic statements. Consider Statements Mary is female John is male Mary and John are siblings Consider Propositions Mary-is-female John-is-male Mary-and-John-are-siblings March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

7 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Motivation First Order Logic (FOL) The language of Predicate Logic Atomic formulas are sets of predicates interpreted as relations between elements of a domain. Consider Predicates Female (Mary) Male (John) Siblings (Mary, John) March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

8 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Motivation Predicate Logic and FOL allows for formulas, arguments, free and bounded variables, and constants. Allows for literals, atomic formulas (relations), and negation. Well formed formulas, truth connectives, quantifiers over bounded variables. Quantifiers: Existential x. ϕ Ξ For some x, ϕ is true Universal x. ϕ For all x, ϕ is true March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

9 Description Logic Motivation Represent relationships between objects.
More complex relationships can be defined in terms of more basic relationships or objects. Woman Ξ Person Π Female Man Person Π Woman verses Woman-is-a-Person-who-is-a-Female Man-is-a-Person-who-is-not-a-Woman March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

10 Description Logic … and First Order Logic
Description Logic (DL) is derived from FOL. DL Relationships are limited to 1-argument predicates called concepts and 2-argument predicates called roles. FOL Predicates DL Relationships Description unary concepts Woman( x ) binary roles Woman( x ).hasChild( y , x ) March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

11 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic … and First Order Logic Concept C and its translation π(C)(x) are equivalent in the interpretation I = (ΔI, . I ) and all a  ΔI , where Δ is the domain, and I is the interpretation. Consider: a  CI iff I |= π(C)(a). a is an instance of an interpretation of WOMAN iff a FOL translation of this concept is true for "a" in the interpretation. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

12 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic … and First Order Logic Translation of roles in FOL takes the form of a predicate with two arguments. Roles relate two concepts to each other. Roles cannot occur by themselves; only as a part of a complex concept. This puts restrictions on syntax of FOL formulas that are allowed. Important to guarantee decidability and good computational properties of DLs. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

13 Description Logic … and First Order Logic
Translations require mappings on two variables : π(x) and π(y) πx(R:C) = y:R(y; x) ∧ πy(C) Mother(x) is true iff x truth values match those of the definition of Woman with some y a child. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

14 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic … and First Order Logic Unary Concept translations Binary Role translation Description Logic First Order Logic πx(A) A(x) πx(C Π D) πx(C) ∧ πx(D) πx(C  D) πx(C) ∨ πx(D) πx(R.C) y.R(x; y) ∧ πy(C) πx( R.C) y.R(x; y)  πy(C) Description Logic First Order Logic πy(A) A(y) πy(C Π D) πy(C) ∧ πy(D) πy(C  D) πy(C) ∨ πy(D) πy(R.C)  y.R(y, x) ∧ πx(C) πy(R.C)  y.R(y, x)  πx(C) March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

15 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic … and First Order Logic Counting Quantifier translations Some x with at least n y’s A simple existential quantifier is insufficient: Description Logic First Order Logic πx(≥ n R)  ≥n y.R(x, y) πx(≤ n R)  ≤n y.R(x, y) March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

16 Description Logic What is Description Logic?
Sub-language of First Order Logic Model hierarchical and complex relationships. Collection of these statements is called TBox. Man Ξ Person Π Woman Father (Person Π Woman) Π hasChild:Person Father Ξ (Man) Π hasChild:Person March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

17 Description Logic What is Description Logic?
The TBox is known as an ONTOLOGY (from Greek: science of beings, or science of things that exist). Ontology used to be a sub-discipline of philosophy. It is now at forefront of real world applications. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

18 Description Logic What is Description Logic?
Models systems which are complex through size and cyclic definitions. “Man who has Only Sons” (Mos) Mos Ξ Man Π hasChild.Man “Man who has Only Male Offspring” (Momo) Momo Ξ Man Π hasChild.Momo March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

19 Description Logic What is Description Logic?
DL handles cyclic definitions through fixpoints Least fixpoint is the lowest x value where f(x) = x Greatest fixpoint is the greatest x value where f(x) = x Halting depends on the domain of x Example: a binary tree’s least-point are the nodes with only terminal nodes at the branches. BinaryTree Ξ Tree Π (≤ 2 has-branch) Π (has-branch.BinaryTree) BinaryEndTree Tree Π (≤ 2 has-branch) Π (has-branch.BinaryTree) March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

20 Description Logic What is Description Logic?
DL has variable-free syntax variables are implicitly understood There can be only two of them: x or y, but they can be reused. DL concepts denote sets of individuals. For concept C, and translation I. a  CI iff I |= π(C)(a). March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

21 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic DL Architecture Basic description language AL AL = attribute language A and B are atomic concepts C and D are concept descriptions R represents atomic roles Other Languages in this family are extensions of AL. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

22 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic DL Architecture Basic description language AL C, D  A atomic concept Τ universal concept bottom concept A atomic negation C Π D intersection  R. value restriction  R.Τ Limited existential quantification March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

23 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic DL Architecture Atomic Concepts and Roles Person Π Female Person who is not a female Person Π hasChild.Τ Person with at least 1 child Person Π hasChild. Person with no children Person Π hasChild.Female Person with at least one daughter Person Π hasChild.Female Person with only daughters Mother  Father Person who is a Parent March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

24 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic DL Architecture Limited to two-variable structures. Handles acyclic, as well as cyclic statements. Acyclic: LHS has no terms which appear in RHS Mos Ξ Man Π hasChild.Man Cyclic: LHS has a term which appears in RHS Momo Ξ Man Π hasChild.Momo March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

25 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic DL Architecture DL defines statements by stating assertions about a subject using its vocabulary. Type of statements which make up a DL Ontology are the ABox and the TBox formalisms. TBox = terminology used in a domain. ABox = assertions which make statements about particular object instances in that domain, using the terminology in TBox. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

26 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic DL Architecture Knowledgebase TBox Description Language Reasoning ABox Application Programs Rules March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

27 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic DL Architecture TBox Father Ξ Man Π hasChild.Person Mother Woman Π hasChild.Person MotherWithoutDaughter Mother Π hasChild.¬Woman ABox MotherWithoutDaughter(MARY) Father(PETER) hasChild(MARY, PETER) hasChild(PETER, HARRY) hasChild(MARY, PAUL) March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

28 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Reasoning Structures DL Knowledgebase stores more then definitions and assertions. DL contains implicit knowledge which can be made explicit through inferences. Inferences are made through a set of rules of inference. TBox contains these statement. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

29 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Reasoning Structures Modeling a domain requires defining a terminology, say T, which satisfies all previously defined statements about that domain. Any new interpretation must satisfy the axioms of T, and result in a non-empty set of concepts. Otherwise it is unsatisfiable. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

30 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Reasoning Structures Satisfiability: A concept C is satisfiable with respect to T if there exists a model I of T such that CI is nonempty: I |= C Subsumption: A concept C is subsumed by a concept D with respect to T if CI  DI for every model I of T: C T D T |= (C D) March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

31 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Reasoning Structures Subsumption con’t LHS is subsumed by the more general terms in RHS: Mother Woman Father Man Brunch Meal Sport Car Vehicle March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

32 Description Logic Reasoning Structures
Equivalence: Two concepts C and D are equivalent if C subsumes D and D subsumes C, with respect to T if CI = DI for every model I of T: [ C ΞT D and T |= C Ξ D ] [ C D and D C ] Disjointness: Two concepts C and D are disjoint with respect to T if: CI  DI =  March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

33 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL Different extensions to AL-languages. ALU Union (C  D) I ALε Full existential quantification R.C ALN Number restrictions ≥ n R and ≤ n R ALFL- Non-atomic negation A AL in combination with U, ε, N, and FL- form 8 languages which are pairwise non-equivalent. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

34 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL The special extension ALUε , represented by ALC, describe a commonly used language with axioms for union and existential quantification. The ALC extension is the basis for SHOIN, which in turn is the basis for the Web Ontology Language OWL OWL is formally recommended by the World Wide Web Consortium, W3C. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

35 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL OWL is a markup language given the tasks of providing syntax used to model ontologies on the web. The Semantic Web is an attempt to organize the plethora of information available on the internet March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

36 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL On the web, instead of showing all available links and infinite levels of sub-topics, a high level “description” is given. Description identifies the subject of the content on the web. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

37 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL Cyclic self-referenced nodes are contained within a higher level concept. Bound by fixpoints. This abstraction inherited from DL ensures satisfiability in the system. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

38 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL OWL Example: “Daugther” (misspelling preserved) <owl:Class rdf:ID="Daugther"> <owl:equivalentClass> <owl:Class> <owl:intersectionOf rdf:parseType="Collection"> <owl:Class rdf:ID="Child"/> <owl:Class rdf:ID="Woman"/> </owl:intersectionOf> </owl:Class> </owl:equivalentClass> <owl:disjointWith> <owl:Class rdf:ID="Son"/> </owl:disjointWith> - March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

39 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL OWL Example: “Daugther” (misspelling preserved) <owl:Class rdf:ID="Daugther"> <owl:equivalentClass> <owl:Class> <owl:intersectionOf rdf:parseType="Collection"> <owl:Class rdf:ID="Child"/> <owl:Class rdf:ID="Woman"/> </owl:intersectionOf> </owl:Class> </owl:equivalentClass> <owl:disjointWith> <owl:Class rdf:ID="Son"/> </owl:disjointWith> Daugther Ξ Child Π Woman Daugther Π Son - March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

40 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL Description Logic can be analyzed for satisfiability with Theorem Provers (Prover9 / Vampire / Otter) Tableau Calculus Tableaux Calculus is a decision procedure solving the problem of satisafiability. If a formula is satisfiable, the procedure will constructively exhibit a model of the formula. March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

41 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL Tableau Calculus con’t The basic idea is to incrementally build the model by looking at the formula, by decomposing it in a top/down fashion. The procedure exhaustively looks at all the possibilities, so that it can eventually prove that no model could be found for unsatisfiable formulas. Unsatifiable formula contains a contradiction such as: p(a) Π p(a) March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

42 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic Applications of DL Closed vs Open World Semantics Connections to external databases require database schema mappings. Communicating TBoxes need to be mapped in order for the data in ABoxes to be meaningful. Closed systems represent internal databases where no mappings are required. A complete theory in a closed system may not be complete in an open world semantic March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

43 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic References Enrico Franconi, DESCRIPTION LOGICS - Tutorial Course Information, Faculty of Computer Science, Free University of Bozen-Bolzano, Italy; Franz Baader, Diego Calvanese, Deborah L. McGuinness, Daniele Nardi, Peter F. Patel-Schneider (Eds.): The Description Logic Handbook: Theory, Implementation, and Applications. Cambridge University Press 2003 Christine Golbreich, A SWRL/OWL Demo Ontology About Family Relationships, OWL Ontologies, ProtegeOntologiesLibrary; March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science

44 Bart Gajderowicz, Ryerson University, Dept. of Computer Science
Description Logic The End Thank you Q & A Bart Gajderowicz March-03-09 Bart Gajderowicz, Ryerson University, Dept. of Computer Science


Download ppt "Bart Gajderowicz, Ryerson University, Dept. of Computer Science"

Similar presentations


Ads by Google