An Axiomatic Semantics for RDF RDFSchema and DAMLOIL

  • Slides: 14
Download presentation
An Axiomatic Semantics for RDF, RDF-Schema, and DAML+OIL Richard Fikes Deborah Mc. Guinness Knowledge

An Axiomatic Semantics for RDF, RDF-Schema, and DAML+OIL Richard Fikes Deborah Mc. Guinness Knowledge Systems Laboratory Stanford University www. ksl. stanford. edu 2/12/01 1 Knowledge Systems Laboratory, Stanford University

What’s Inferable From Semantic Markup u Formal specification of intended meaning of semantic markup

What’s Inferable From Semantic Markup u Formal specification of intended meaning of semantic markup 4 RDF, RDF Schema, and DAML+OIL u Translation of semantic markup into first order logic (FOL) 4 Produces a logically equivalent ontology in FOL 4 FOL language is KIF (Knowledge Interchange Format) u Provides axioms that specify legal inferences and constraints [137] u Facilitates query answering and constraint checking 4 By traditional theorem provers and problem solvers 4 Provides specs for special purpose reasoners u 2 Provides basis for discussing language changes and extensions Knowledge Systems Laboratory, Stanford University

Translation Method 4 Translate markup into RDF statements I. e. , into “Property P

Translation Method 4 Translate markup into RDF statements I. e. , into “Property P of resource R has value V” E. g. , “Property parent of resource Joe has value John” 4 Translate each RDF statement into a FOL sentence I. e. , into “(Property. Value P R V)” E. g. , “(Property. Value parent Joe John)” 4 Simplify property typing sentences using relation type I. e. , simplify “(Property. Value type R V)” to “(type R V)” E. g. , simplify “(Property. Value type Joe Person)” to “(type Joe Person)” 4 3 Add axioms from the semantics document Knowledge Systems Laboratory, Stanford University

RDF Classes and Properties Classes [10] (18 axioms) Properties [6] (12 axioms) Resource type

RDF Classes and Properties Classes [10] (18 axioms) Properties [6] (12 axioms) Resource type Property subject Class predicate Literal object Statement value Container _1, _2, _3, … Bag Seq Alt Container. Membership. Property 4 Knowledge Systems Laboratory, Stanford University

RDF Schema Classes and Properties Classes [2] (2 axioms) Constraint. Resource Constraint. Property Properties

RDF Schema Classes and Properties Classes [2] (2 axioms) Constraint. Resource Constraint. Property Properties [8] (18 axioms) sub. Class. Of sub. Property. Of see. Also is. Defined. By comment label range domain 5 Knowledge Systems Laboratory, Stanford University

DAML+OIL Classes Thing Nothing Disjoint Restriction Non. Negative. Integer Transitive. Property Unique. Property Unambiguous.

DAML+OIL Classes Thing Nothing Disjoint Restriction Non. Negative. Integer Transitive. Property Unique. Property Unambiguous. Property List Ontology (10 classes; 14 axioms) 6 Knowledge Systems Laboratory, Stanford University

DAML+OIL Properties equivalent. To on. Property max. Cardinality. Q same. Class. As to. Class

DAML+OIL Properties equivalent. To on. Property max. Cardinality. Q same. Class. As to. Class cardinality. Q same. Property. As has. Value inverse. Of disjoint. With has. Class first union. Of min. Cardinality rest disjoint. Union. Of max. Cardinality item intersection. Of cardinality version. Info complement. Of has. Class. Q imports one. Of min. Cardinality. Q (26 properties; 69 axioms) 7 Knowledge Systems Laboratory, Stanford University

Class Person and Person Joe u Persons are animals and a person’s parents are

Class Person and Person Joe u Persons are animals and a person’s parents are persons <rdfs: Class rdf: ID = "Person"> <rdfs: sub. Class. Of rdf: resource = "#Animal” /> <restricted. By> <Restriction> <on. Property resource = "#parent” /> <to. Class resource = "#Person” /> </Restriction> </restricted. By> </Class> u Joe is a person one of whose parents is John <Person ID = "Joe"> <parent resource = "#John” /> </Person> 8 Knowledge Systems Laboratory, Stanford University

Translation Into RDF Statements u Persons are animals (type Person Class) (sub. Class. Of

Translation Into RDF Statements u Persons are animals (type Person Class) (sub. Class. Of Person Animal) u A person’s parents are persons (type R Restriction) (restricted. By Person R) (on. Property R parent) (to. Class R Person) u Joe is a person one of whose parents is John (type Joe Person) (parent Joe John) 9 Knowledge Systems Laboratory, Stanford University

Translation Into First Order Logic (type Person Class) (Type Person Class) (sub. Class. Of

Translation Into First Order Logic (type Person Class) (Type Person Class) (sub. Class. Of Person Animal) (Property. Value sub. Class. Of Person Animal) (type R Restriction) (Type R Restriction) (restricted. By Person R) (Property. Value restricted. By Person R) (on. Property R parent) (Property. Value on. Property R parent) (to. Class R Person) (Property. Value to. Class R Person) (type Joe Person) (Type Joe Person) (parent Joe John) (Property. Value Parent Joe John) 10 Knowledge Systems Laboratory, Stanford University

Primary Axiom For to. Class u If object R is a value of restricted.

Primary Axiom For to. Class u If object R is a value of restricted. By for object C 1, and object P is a value of on. Property for R, and object C 2 is a value of to. Class for R, then for all objects I and V, if I is of type C 1 and V is a value of P for I, then V is type C 2. ” u (=> (and (Property. Value restricted. By ? c 1 ? r) (Property. Value on. Property ? r ? p) (Property. Value to. Class ? r ? c 2)) (forall (? i ? v) (=> (and (Type ? i ? c 1) (Property. Value ? p ? i ? v)) (Type ? v ? c 2)))) 11 Knowledge Systems Laboratory, Stanford University

Is John a Person? u From – 4 (Property. Value restricted. By Person R)

Is John a Person? u From – 4 (Property. Value restricted. By Person R) 4 (Property. Value on. Property R parent) 4 (Property. Value to. Class R Person) 4 (=> (and (Property. Value restricted. By ? c 1 ? r) (Property. Value on. Property ? r ? p) (Property. Value to. Class ? r ? c 2)) (forall (? i ? v) (=> (and (Type ? i ? c 1) (Property. Value ? p ? i ? v)) (Type ? v ? c 2)))) u Infer – 4 (forall (? i ? v) (=> (and (Type ? i Person) (Property. Value parent ? i ? v)) (Type ? v Person))) 12 Knowledge Systems Laboratory, Stanford University

Is John a Person? u From – 4 (Type Joe Person) 4 (Property. Value

Is John a Person? u From – 4 (Type Joe Person) 4 (Property. Value Parent Joe John) 4 (forall (? i ? v) (=> (and (Type ? i Person) (Property. Value parent ? i ? v)) (Type ? v Person))) u Infer – 4 (Type John Person) u 13 “Yes”, John is a Person. Knowledge Systems Laboratory, Stanford University

Summary u Formal specification of intended meaning of semantic markup 4 RDF, RDF Schema,

Summary u Formal specification of intended meaning of semantic markup 4 RDF, RDF Schema, and DAML+OIL u Translation of semantic markup into first order logic (FOL) 4 Produces a logically equivalent ontology in KIF u Provides axioms that specify legal inferences and constraints u Facilitates query answering and constraint checking 4 By traditional theorem provers and problem solvers 4 Provides specs for special purpose reasoners u 14 Provides basis for discussing language changes and extensions Knowledge Systems Laboratory, Stanford University