Querying Dynamic and ContextSensitive Metadata in Semantic Web

Querying Dynamic and Context-Sensitive Metadata in Semantic Web Sergiy Nikitin Industrial Ontologies Group 1 University of Jyväskylä Finland Article Authors: Sergiy Nikitin Vagan Terziyan Yaroslav Tsaruk Andriy Zharko 1 – Industrial Ontologies Group web-site: http: //www. cs. jyu. fi/ai/Onto. Group

What lies beneath abstract models? How Intelligent Agent manages data?

Contents • • • Story of contextual data querying problem Contextual Data in Semantic Web RDQL patterns Use cases for pattern application in Agent Systems Conclusions Further Work

Introduction • Dynamic, semantically rich data usually contains contextual elements describing conditions under which the data is relevant, useful and up-to-date • The problem of querying contextual data appeared as a first-year challenge of Smart. Resource 1 project • Project wider objective is: – To combine the emerging Semantic Web, Web Services, Peer-to-Peer, Machine Learning and Agent technologies for the development of a global and smart maintenance management environment, to provide Web-based support for the predictive maintenance of industrial devices by utilizing heterogeneous and interoperable Web resources, services and human experts 1 - Smart. Resource project web-site: http: //www. cs. jyu. fi/ai/Onto. Group/projects. htm

Smart Resource 2005 Scenario (3 scenes) “Knowledge Transfer form Expert to Service” “Service” Labelled data d e l l Labe r ta fo a d g g ryin Que learnin Diagnostic model History data nd lts a ple resu m sa stic g n ni iagno r a Le ng d i ery u Q Qu ery ing res diag ult nos s tic Labelled data tch in dia g and gno stic query dat ing a “Device” Wa La be lle d da ta “Expert”

Smart. Resource project • The objective of project stage 1 (year 2004): – Define Semantic Web-based framework for unification of maintenance data and interoperability in maintenance system – R&D tasks included: • Development of generic semantic adapter mechanism (General Adaptation Framework) • Supporting Ontology (Resource State/Condition Description Framework) for different types of industrial resources: devices, software components (services) and humans (operators or experts).

Contextual Data • Rsc. DF (Resource State/Condition Description Framework) provides additional constructions on top of RDF-Schema • Rsc. DF is fully compliant with RDF • Contextual construction for Statement rscdfs: true. In. Context rscdfs: Context_SR_Container Statement rdf: object rdf: subject rscdfs: predicate SSS PPP OOO

Use Case Example • Query: “Select Statements corresponding to state of some device” Device 1 Sensors State Time Property Value T 1 temperature 70 rounds. Per. Minute 1500 temperature 80 rounds. Per. Minute 1700 temperature 83 rounds. Per. Minute 1750 T 2 T 3

Contextual Data Example rscdfs: true. In. Context rscdfs: Context_SR_Container Temperature Statement 1 Statement rdf: subject rscdfs: predicate rdf: object World has. Time rscdfs: predicate 07. 06. 05 T 11: 33: 12 Device 1 Both containers refer to the same time statement rdf: object rdf: subject temperature. Celsius Value: 70 Unit: Celsius rscdfs: true. In. Context rscdfs: Context_SR_Container Rotation Statement 1 Statement rdf: subject rscdfs: predicate rdf: object World has. Time rdf: object rdf: subject rscdfs: predicate 07. 06. 05 T 11: 33: 12 Device 1 rounds. Per. Minute Value: 1500 Unit: rpm

Statement Example rscdfs: true. In. Context rscdfs: Context_SR_Container Statement rdf: subject rscdfs: predicate rdf: object World has. Time rdf: subject rscdfs: predicate rdf: object Device 1 cont. Ont: resource. State 07. 06. 05 T 11: 33: 12 rscdfs: SR_Container rscdfs: true. In. Context Template Statement rdf: subject World rscdfs: predicate meas. Ont: resource. Measurement Temperature Statement 1 Rotation Statement 1

RDQL-patterns SELECT ? Value. Statements, * ? Num. Units, * ? Num. Values* WHERE (<State. Stmt. ID>, <rdf: object>, ? State. Container), * (? State. Container, <rscdfs: member>, ? Value. Statements), (? Value. Statements, <rdf: object>, ? Num. Value. Instances), (? Num. Value. Instances, <rscdfs: value>, ? Num. Values), * (? Num. Value. Instances, <rscdfs: unit>, ? Num. Units) * Statement ID * Unit * Value* Temperature Statement 1 Temperature 70 Rotation Statement 1 rounds. Per. Minute 1500

RDQL-patterns: Modularity Input Pattern Output Composed Pattern Input Pattern Output

Use cases for pattern application in Agent Systems rscdfs: Context_SR_Container rscdfs: true. In. Context rscdfs: SR_Statement rdf: object rdf: subject rscdfs: predicate rscdfs: SR_Container Agent has. Goals Goal Statement 1 Goal Statement 2 …

Use cases for pattern application in Agent Systems rscdfs: Context_SR_Container Statement rdf: subject rscdfs: predicate rdf: object Agent has rscdfs: true. In. Context Money Behaviour_Statement rdf: object rdf: subject rscdfs: predicate Behaviour_Container Agent has. Behaviour Buy Tickets

Agent Architecture Ontology Roles Goals Resource Agent Templates Behaviour rules Resource History Templates Behaviour description Executable modules or Web Services

Conclusions • Storing and managing context-enabled data via RDF storages is complicated and routine task • Repeating querying procedures can be organized into reusable querying patterns • Patterns can consist of other patterns, thus pattern ontology can be developed to represent these relationships • Patterns correspond to Properties. Property by its range value defines classes of objects which can be referred, hence these objects correspond to certain common structure

Further work • Further development of Resource Goal/Behaviour Description Framework (RGBDF) • Querying patterns for RGBDF • Deeper analysis of Pattern Ontology (how to describe relationships between patterns, how they correlate with Properties)

Welcome to IASW-2005 conference http: //www. cs. jyu. fi/ai/Onto. Group/IASW-2005/