Simplified Approach for Representing Part-Whole Relations in OWL-DL - - PowerPoint PPT Presentation

Simplified Approach for Representing Part-Whole Relations in OWL-DL Ontologies

A.Aziz Altowayan and Lixin Tao

Pace University IEEE BigDataSecurity, 2015

• Aug. 24, 2015

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion

Outline

1

Ontology and Knowledge Representation Ontology for representing knowledge Ontology Relations OWL: Representing and Reasoning

2

Representation Problem in OWL Scenario Current Approach (W3.org Best Practice)

3

Proposed Alternative Proposed Approach Comparison and Demo

4

Conclusion Limitations and Future Work

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Ontology for representing knowledge Ontology Relations OWL: Representing and Reasoning

Ontology and Knowledge Representation

Why we need to represent knowledge?

Definition: an ontology formally represents knowledge as a set of concepts

within a domain, using a shared vocabulary to denote the types, properties and interrelationships of those concepts. (Wikipedia)

Goals for having ontology (AI systems): What is this ? (Representation). How this relates to that ? (Inference).

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Ontology for representing knowledge Ontology Relations OWL: Representing and Reasoning

Ontology and Knowledge Representation

Example of an ontology model

Sample ontology for a human body anatomy

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Ontology for representing knowledge Ontology Relations OWL: Representing and Reasoning

IS-A and Part-Whole Relations

Class hierarchy

We are focusing on two types of ontology taxonomy: Generalization-based (e.g. Heart is an Organ) Partonomic-based (e.g. Retina is part of the Eye)

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Ontology for representing knowledge Ontology Relations OWL: Representing and Reasoning

TBox vs. ABox Statements

Definitions

Terminological Box and Assertional Box Statements

TBox: Relations Between Concepts, e.g. Every Car is a Vehicle. ∀Car ⊑ Vehicle An Engine is part of a Vehicle. Engine ⊑ CarPart OR Engine ⊑ ∃ partOf Vehicle ABox: Assertion Sentences (between an instance and concept), e.g. BMW is a Car. Car(BMW) TBox + ABox = Knowledge Base

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Ontology for representing knowledge Ontology Relations OWL: Representing and Reasoning

OWL

Building ontology

Web Ontology Language: OWL is a family of knowledge representation languages for authoring ontologies (Semantic Web). Description Logics: DL (variant of FOL) used in OWL to specify the semantics of a relation. “Reasoning in ontologies and knowledge bases is one of the reasons why a specification needs to be formal one.”

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Scenario Current Approach (W3.org Best Practice)

What is the Problem ?

Let’s introduce the problem through a simple Use Case.

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Scenario Current Approach (W3.org Best Practice)

Scenario

For representing a simple IS-A and Part-Of relations in OWL

Example: How to assert the following model? In words, Car class as a generalization of (is a) Vehicle class, and Engine class as a component of (part of) Vehicle class.

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Scenario Current Approach (W3.org Best Practice)

Scenario

For representing a simple IS-A and Part-Of relations in OWL . . .

IS-A relation (very simple way) :

Example <owl:Class rdf:about="ns#Car"> <rdfs: subClassOf rdf:resource="ns#Vehicle"/> </owl:Class>

Part-Whole relation: . . . OWL does not provide simple (and straightforward) primitives for part-whole relations.

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Scenario Current Approach (W3.org Best Practice)

Why is that?

Difficulty in representing part-whole relations

Because,

Part-Whole relations are variants (many different forms). Thus Some constraints (e.g. transitivity, cardinality) need to be imposed to specify the semantic of the relation.

• C. M. Keet and A. Artale, Representing and reasoning over a taxonomy of part-whole relations, Applied Ontology,
• vol. 3, Jan. 2008

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Scenario Current Approach (W3.org Best Practice)

Current Approach

For representing Part-Whole in OWL

The Current Approach provides a manual implementation of the underlying steps (thus a long and error-prone process) for achieving a single task.

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Scenario Current Approach (W3.org Best Practice)

Current Approach

Three steps implemented separately and manually

1

Create an objectProperty (i.e. partOf) with specifying the characteristics it holds.

2

Create a new part-aggregating class (i.e. VehiclePart) to represent the parts type. And (using OWL-DL) make it:

equivalentClass to the restriction partOf some Vehicle. e.g. carPart ≡ ∃ partOf Vehicle

3

Extend the part Class Engine to cope with the constraints. By using OWL-DL to:

make it a subClassOf of the restriction partOf some

• Vehicle. e.g.

Engine ⊑ ∃ partOf Vehicle

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Scenario Current Approach (W3.org Best Practice)

Current Approach

The three steps serialized in RDF/XML syntax

Example <!-- Step 1 --> <owl: ObjectProperty rdf:about="&part;partOf"> <rdf:type rdf:resource="&owl; TransitiveProperty "/> </owl: ObjectProperty > <!-- Step 2 --> <owl:Class rdf:about="ns# VehiclePart "> <owl: equivalentClass > <owl: Restriction > <owl:onProperty rdf:resource="ns#partOf"/> <owl: someValuesFrom rdf:resource="ns#Vehicle"/> </owl: Restriction > </owl: equivalentClass > </owl:Class> <!-- Step 3 --> <owl:Class rdf:about="ns#Engine"> <rdfs: subClassOf> <owl: Restriction > <owl: onProperty rdf:resource="ns#partOf"/> <owl: someValuesFrom rdf:resource="ns#Vehicle"/> </owl: Restriction > </rdfs: subClassOf > </owl:Class> A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Proposed Approach

Simplifying Part-Whole Relations in OWL

Integrate the three steps into a single line

i.e. part-whole relations in a similar manner to subClassOf, as follows:

Example <owl:Class rdf:about="ns#Engine"> <relation:partOf transitive ="yes" rdf:resource="ns#Vehicle"/> </owl:Class>

Then, automatically1extract and build the relation’s constraints.

1Using pyowl module. A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Proposed Approach

Work flow of our method Simplified Relations Using OWL’s Annotations ↓ Extraction and Transformation Relation ↔ Elements Mapping ↓ Generate Final layout Standardized RDF/XML

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Proposed Approach

Simple Conceptual Model

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Transformation Method

Mapping stage 1

Extend current class to be a subClassOf its declared parent

Example <owl:Class rdf:about="onto#Engine"> <!-- Auto generated mapping

• ->

<rdfs: subClassOf> <owl: Restriction > <owl: onProperty rdf:resource="onto#partOf"/> <owl: someValuesFrom rdf:resource="onto#Vehicle"/> </owl: Restriction > </rdfs: subClassOf > <!-- End mapping

• ->

</owl:Class>

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Transformation Method

Mapping stage 2

Generate an objectProperty for the relation

Example <!-- Property

• ->

<!-- Auto generated mapping

• ->

<owl: ObjectProperty rdf:about="onto#partOf"> <rdf:type rdf:resource="&owl; TransitiveProperty "/> </owl: ObjectProperty > <!-- End mapping

• ->

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Transformation Method

Mapping stage 3

Generate the parts-aggregating Class as an equivalentClass of parent

Example <!-- Auxiliary Class

• ->

<!-- Auto generated mapping

• ->

<owl:Class rdf:about="onto# partOf_Vehicle "> <owl: equivalentClass > <owl: Restriction > <owl: onProperty rdf:resource="onto#partOf"/> <owl: someValuesFrom rdf:resource="onto#Vehicle"/> </owl: Restriction > </owl: equivalentClass > </owl:Class> <!-- End mapping

• ->

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Comparison

Ontology Metrics of Both Approaches

Sample Ontology Metrics

The proposed method reduces the axiom count by 40% to 50% simplify the expressivity level (of DL) from ALE+ to AL

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Comparison

Evaluation: classification results are the same (safe representation)

Inferred model2 from each approach3:

(a) Our Transformed Approach (b) Current Approach

2OWL-Viz and HermiT 1.3.8 reasoner plugins (built-in in Protege) 3Ontology example3.owl from Rector, Alan, et al. A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Proposed Approach Comparison and Demo

Demo of the approach

A complete working example: see

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Limitations and Future Work

Limitations

Of the proposed approach

Currently, this approach supports ALE+ expressivity level

three constraints Transitivity, Cardinality, and Inverse.

applies to relations of TBox statements only.

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Limitations and Future Work

Future Work and Contributions

Future Work

Support different DL levels (ALEHI+ e.g. subProperties). Applicability to other OWL syntaxes (e.g. Manchester and functional). Interfacing the module with Protege.

Contribution

The proposed approach contributes to: simplify the complexity involved in representing part-whole relations. reduce work overhead and amount of work needed, thus minimizing the possibility of error making. provide support to Semi-Auto Ontology Building. proved the feasiability of the approach.

A.Aziz Altowayan and Lixin Tao

Ontology and Knowledge Representation Representation Problem in OWL Proposed Alternative Conclusion Limitations and Future Work

Conclusion

In summary, Representing the semantics of part-whole relations is a fairly complex

• process. Why?

Because each relation may hold different (inference) characteristics depending on the context. Description Logics introduced in OWL to capture such varieties. So what’s wrong? OWL-DL is applied through a manual (and separately implemented steps

• f the same) process. Therefore, we introduced

Simplified approach for applying OWL-DL for representing part-whole relations.

A.Aziz Altowayan and Lixin Tao

Questions ?

Thank You For Your Attention