modular ontology architecture for data integration in the geolink …€¦ · 25-02-2016 ·...

Modular Ontology Architecture for Data Integration inthe GeoLink Project

Adila Krisnadhi

Wright State University

Ontology Summit 2016

Krisnadhi GeoLink Data Integration Ontology Summit 2016 1 17

Motivation

Needed

Data integration providing unified view over data at different sources


DI Challenges and GeoLink Project

Challenges (regardless of architecture)

Syntactic heterogeneity different data formats serializationsSemantic heterogeneity different vocabulary different level ofgranularity in data different conceptualizationSocialnon-technical inabilityunwillingness to participate fear ofunanticipated cost worry with major changes in their local systemskeptic with scalability

GeoLink Project (wwwgeolinkorg)

Part of NSFrsquos EarthCube Program ndash one among dozens of buildingblock projectsLinked Data + Ontology design pattern-based integration


Why Ontology Patterns

Upper-level and many domain ontologies are

Hard to understand mdash too many terms too abstract too complicatedaxioms too far from real dataImpose ontological commitments that may not be acceptable by allpartiesBrittle mdash costlyhard to extend carelessly extending may cause thewhole thing breaks

Ontology design pattern (ODP) a (ldquoreusablerdquo) solution of afrequently occurring modeling problem in the domain and can act asa building block of a more complex ontology

Content pattern (CP) an ODP that models a particular genericnotion in a particular domain

Community engagement via collaborative modeling


GeoLink Integration Architecture


Modeled Notions

Content patterns corresponding to concrete domain notions

Cruise Vessel Person Organization Funding Award ProgramPhysical Sample Dataset Digital Object Publication Platform PlaceTime

Content patterns from abstraction in modeling

Agent Agent Role Event Information Object Identifier Personal InfoItem Person Name Property Value


Ontology Overview

Each node represents a content pattern


CP Example Cruise

Generate competency questionsldquoFind all cruises passing through Gulf of Maine in August 2013rdquoldquoShow the tracks of cruises in operation in September 2013rdquoldquoList all cruise vessels that departed from Woods Hole in 2012rdquoldquoFind the chief scientists of any cruise that collected samples ofcarbon-isotope data in Lake SuperiorrdquoldquoWhat datasets were produced by the cruise AE0901rdquoldquoWhich cruises are funded by the NSF award DBI-0424599rdquo

Understand the nature of things we model

Cruise is an EventTrack maybe complex reuse Trajectory pattern1

Vessel maybe complexChief scientist a role of an agentDataset maybe complexFunding award maybe complex

1Hu et al ldquoA geo-ontology design pattern for semantic trajectoriesrdquo COSIT 2013


Cruise CP


Axiomatization

Use informal natural language to model axioms together with domainexpertsdata providers

Cruise v Event

Cruise has exactly 1 trajectory and is undertaken by exactly 1 vesselCruise v (=1 hasTrajectoryTrajectory) u (=1 isUndertakenByVessel)

Cruise is described by exactly 1 information objectCruise v (=1 isDescribedByInformationObject)

Trajectory of a cruise must be traveled by the vessel by which thecruise is undertakenhasTrajectoryminus isUndertakenBy v isTraveledBy


Cruise Trajectory


Bridging Semantic Heterogeneity

Since patterns represent key notions as understood by domain expertsand data providers intuitively an appropriate mappingalignmentexists between ldquolocalrdquo vocabulary and the patterns

A (local) pattern view between a data source and the patterns makessuch a mapping explicit

View is a very minimalistic schema (class names property namessimple domain and range axioms)Separating ldquocore conceptualizationrdquo and ldquonomenclaturerdquo issuesvocabulary terms in a local view may be repository-specific and neednot be the same as the patternsMapping can be expressed in rules that help populating the patternsData providers can populate the global schema (pattern collection) bysimply populating a local viewExisting controlled vocabulary can also be accommodated as a patternview


Pattern View Example

Producer populates view

to populate Cruise Agent Role and Person patterns


Mapping Rule

vCruise(X) and vhasChiefScientist(XY ) and vPerson(Y )

minusrarr existZ(cCruise(X) and arprovidesAgentRole(XZ)

and cChiefScientistRole(Z)

and arisPerformedBy(Z Y ) and pPerson(Y ))


Mapping Rule

CONSTRUCT

X a cCruise

arprovidesAgentRole [ a cChiefScientistRole

arisPerformedBy Y ]

Y a pPerson

WHERE

X a vCruise vhasChiefScientist Y

Y a vPerson


Conclusion and Outlook

The GeoLink modular oceanography ontology = collection of contentpatterns in oceanography

Collaborative modeling approach

Two-layered ontology architecture with patterns and local views helpssemantic interoperability across different data sources while allowingdata providers to retain their own local vocabulary and schema

See also httpwwwgeolinkorg httpschemageolinkorg


Questions

Acknowledgement

NSF for GeoLink funding


Motivation

Needed

Data integration providing unified view over data at different sources

















Modeled Notions






Ontology Overview



CP Example Cruise







Cruise CP


Axiomatization


Cruise v Event





Cruise Trajectory











Mapping Rule






Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement


















Modeled Notions






Ontology Overview



CP Example Cruise







Cruise CP


Axiomatization


Cruise v Event





Cruise Trajectory











Mapping Rule






Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement



Ontology Overview



CP Example Cruise







Cruise CP


Axiomatization


Cruise v Event





Cruise Trajectory











Mapping Rule






Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement



CP Example Cruise







Cruise CP


Axiomatization


Cruise v Event





Cruise Trajectory











Mapping Rule






Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement



Cruise CP


Axiomatization


Cruise v Event





Cruise Trajectory











Mapping Rule






Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement



Axiomatization


Cruise v Event





Cruise Trajectory











Mapping Rule






Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement



Cruise Trajectory











Mapping Rule






Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement












Mapping Rule






Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement



Mapping Rule

CONSTRUCT

X a cCruise


arisPerformedBy Y ]

Y a pPerson

WHERE


Y a vPerson








Questions

Acknowledgement









Questions

Acknowledgement



modular ontology architecture for data integration in the geolink …€¦ · 25-02-2016 ·...

Documents