39
International Conference on computer science – research and applications IBIZA 2014 21 March 2014 UMCS Lublin

Modeling Ontologies with Natural Language

Embed Size (px)

DESCRIPTION

Dr Alessandro Seganti from Cognitum presented basics of Semantic Technologies, OntorionCNL, Ontorion Semantic Framework and Fluent Editor during International Conference on Computer Science -- Research and Applications IBIZA 2014, UMCS Lublin. To learn more visit: http://www.cognitum.eu/semantics/

Citation preview

Page 1: Modeling Ontologies with Natural Language

International Conference on computer science – research and applications IBIZA 2014

21 March 2014 UMCS Lublin

Page 2: Modeling Ontologies with Natural Language

Alessandro SegantiData Engineer @Cognitum

[email protected]

The company, product and service names used in this web site are for identification purposes only. All trademarks and registered trademarks are the property of their

respective owners.

Modeling Ontologies with Natural Language

Fluent Editor and Ontorion Server

Page 3: Modeling Ontologies with Natural Language

OWL and OCNLintroduction

Page 4: Modeling Ontologies with Natural Language

Semantic Web: Multiple ontologies written in OWL/RDF, locally or in the web

Page 5: Modeling Ontologies with Natural Language

DL

OCNL

SWRL

Horn Logic

Page 6: Modeling Ontologies with Natural Language

{New−York}⊑(city)⊓(∃(belong−to).({New−York−State}))

New-York is a city and belongs-to New-York-State.

DL

OWL

OCNL

Page 7: Modeling Ontologies with Natural Language

intermediate−customer⊑customer

Every intermediate-customer is a customer.

DL

OWL

OCNL

Page 8: Modeling Ontologies with Natural Language

industrial−customer≡(customer)⊓(∃(have−industry).(⊤))

Something is an industrial-customer if-and-only-if-it is a customer that has-industry something.

DL

OWL

OCNL

Page 9: Modeling Ontologies with Natural Language

⌂city(?city0)⋀state(?state0)⋀belong−to(?city0,?state0)→contain(?state0,?city0)

If a city belongs-to a state then the state contains the city.

SWRL

OCNL

Page 10: Modeling Ontologies with Natural Language

Ontorion CNL : Basic ConceptsSemantic technologies

Page 11: Modeling Ontologies with Natural Language

Concept/Class definition

young-male-man very-beautiful-girl

Every parent is a person.

Class identifiers start with a small letter and they use dashes between words, e.g. very-beautiful-girl, young-male-man.

Classes are defined when their names are used in at least one sentence.

OWL → FE CNL: VeryBeautifulGirl → very-beautiful-girl

Page 12: Modeling Ontologies with Natural Language

Instances

John is a person.

Instance identifier = each part starts with a capital letter and they are separated with hyphens

John-Dow Tanker-Accident-X

OWL → FE CNL: JohnDow → John-Dow

The-”K22 P2”

Page 13: Modeling Ontologies with Natural Language

Dealing with namespacesAt first line, when creating a new ontology model in FE, you can declare a default namespace for your ontology.Namespace: 'http://www.ontorion.com/ontologies/family-model-instances'.

At last lines, you can define references of other ontology models that you can use when writing the current model.References: [family] 'http://www.ontorion.com/ontologies/family' ('.\family-model.encnl').

Then, writing the model, you use the reference name with [] together with notions of referenced ontology.Mary is a woman[family]. Rose has-parent[family] Mary.

Page 14: Modeling Ontologies with Natural Language

Fluent Editor and OntorionSemantic technologies

Page 15: Modeling Ontologies with Natural Language

Fluent Editor™ 2

User is precisely guided during the entire editing process

Hints

Explanations

Predictive editor

Taxonomy preview

Page 16: Modeling Ontologies with Natural Language

Fluent Editor™ 2

build complex scenarios using referenced ontologies from Web or servers

Page 17: Modeling Ontologies with Natural Language

Architecture Overview

Executors

Any huge documentrepository

modularizator

reasoner

Ontologydatabase

reasoner

reasoner

reasoner

External data stream(social, logs, sensors)

CNL Semantic Applications (Desktop or Web)

OCNL

Page 18: Modeling Ontologies with Natural Language

Cognitum CNL : OWL equivalenceSemantic technologies

Page 19: Modeling Ontologies with Natural Language

Concept subsumption

Every boy is a young-male-man.

Saying that one concept subsumes the other we define IS-A/taxonomic relation and a concept hierarchy.

Page 20: Modeling Ontologies with Natural Language

Concept equivalence

Something is a boy if-and-only-if-it is a young-male-man.

Every boy is a young-male-man.Every young-male-man is a boy.

Saying that two concepts are equivalent we say that they also subsumes each other.

Page 21: Modeling Ontologies with Natural Language

Concept intersection

Something is a young-male-man if-and-only-if-it is a young-thing that is a male-thing and is a person.

We can define intersections and unions of any given length.

Something is a young-male-man if-and-only-if-it is a young-thing and is a male-thing and is a person.

Page 22: Modeling Ontologies with Natural Language

Concept union

Something is a child if-and-only-if-itis a boy and-or is a girl.

We can define unions of concepts with the use of and-or keyword.

Page 23: Modeling Ontologies with Natural Language

Negations / Complement

Something is a young-thing if-and-only-if-it is not an adult-thing.

No young-thing is an adult-thing.

Every-single-thing that is a young-thingis not an adult-thing.

Everything in the world is an adult-thing or a young-thing.

Every young-thing is not an adult-thing.

The classes of young-thing and adult-thing are disjoint.

Page 24: Modeling Ontologies with Natural Language

Value partition / Disjoint union

Something is a person if-and-only-if-it-eitheris a child, is a young-thing, is a middle-age-thing or is an old-thing.

A disjoint union axiom states that a class C is a disjoint unionof the class expressions CEi , 1 ≤ i ≤ n, all of which are pairwise disjoint.

Page 25: Modeling Ontologies with Natural Language

• Existential role restrictions

• Universal role restrictions

Every person is-a-child-of a parent.

Every person is-a-child-of nothing-but parents.

These restrictions are complementary to each other. However, they do not imply each other.

Something is a happy-person if-and-only-if-it has-child a happy-person and has-child nothing-but happy-persons.

Page 26: Modeling Ontologies with Natural Language

Restrictions

Something is a johns-children if-and-only-if-it has-parent John.

Something is a my-birthday-guests if-and-only-if-it is either John, Mary or Bill.

Property value

Enumeration of individuals

Page 27: Modeling Ontologies with Natural Language

Property restrictionsEvery person is-a-child-of at-most two parents.

Every person is-a-child-of at-least two parents.

Every person is-a-child-of two parents.

keyword less-than more-than

≤ at-most

≥ at-least

≠ different-than

= -

Page 28: Modeling Ontologies with Natural Language

Restrictions & inversion of roles

Every child is loved by parents.

We can use inverse of properties in ontology axioms using „be” and „by” keywords in FE.

Page 29: Modeling Ontologies with Natural Language

Keys in ontology

Keys are for uniquely identifying an individual.

Every X that is a person is-unique-if X has-id something and X has-name something.

Page 30: Modeling Ontologies with Natural Language

Cognitum CNL : Property CharacteristicsSemantic technologies

Page 31: Modeling Ontologies with Natural Language

Property subsumption

If X loves Y then X likes Y.

Role inclusion

Page 32: Modeling Ontologies with Natural Language

Axioms on properties (1)

If X has-ancestor something that has-ancestor Y then X has-ancestor Y. has-ancestor is

transitive

X has-sibling Y if-and-only-if Y has-sibling X. has-sibling is

symmetric

has-child is an inverse of has-parent

X has-child Y if-and-only-if Y has-parent X.

If X has-child Y then X not has-spouse Y. has-spouse and has-child are disjoint

Page 33: Modeling Ontologies with Natural Language

Axioms on properties (2)

Domain of is-a-wife-of property is a woman class

Every-single-thing that is-a-wife-of is a woman.

Range of is-a-wife-of property is a person class

Every-single-thing is-a-wife-of nothing-but persons.

Every-single-thing is not a thing that has-sibling itself.

Every person likes itself.

has-sibling is irreflexive

likes is reflexive

Page 34: Modeling Ontologies with Natural Language

Complex role/property chainshas-parent has-brother

has-uncleIf X has-parent something that has-brother Y then X has-uncle Y.

Object property chains provide a means to define properties as a composition of other properties.

SubObjectPropertyOf ( ObjectPropertyChain( OPE1 ... OPEn ) OPE) states that any individual x connected with an individual y by a chain of object properties expressions OPE1, ..., OPEn is necessary connected with y by the object property OPE.

Page 35: Modeling Ontologies with Natural Language

Data property assertions

John has-name equal-to 'John'.

Lenka borns-on-date equal-to '1975-11-10'.

Tanker-Accident has-time equal-to '2013-07-08T9:30:40.40'.

Page 36: Modeling Ontologies with Natural Language

Data property domain & range

Every-single-thing that has-name (some-value) is a person.

Every-single-thing has-name nothing-but (some-string-value).

Keywords used to define date property values are:some-value, some-string-value, some-integer-value, some-boolean-value, some-real-value, and some-datetime-value.

Page 37: Modeling Ontologies with Natural Language

Data property axioms

Every adult-person has-age greater-or-equal-to 18.

= equal-to greater-than lower-than

≥ lower-or-equal-to

≤ greater-or-equal-to

≠ different-from

Page 38: Modeling Ontologies with Natural Language

Regular expressionsEvery-single-thing that has-name that-matches-pattern '.*a' is a female-person.

Regular expressions

[] alternative sign in square brackets, e.g. [A-Z], [tT]he

[^ ] negation, e.g. [^0-9]

| disjunction, e.g. the|The

? optional previous char, e.g. colou?r

* 0 or more of previous char, e.g. [A-Z][a-z]*

+ 1 or more of previous char

. any char

^ beginning anchor

$ end anchor

Page 39: Modeling Ontologies with Natural Language

to learn more visit our website:

www.cognitum.eu/semantics/