Experiences of linking ontologies to a geospatial database Research & Innovation

advertisement
Experiences of linking ontologies to a
geospatial database
Cathy Dolbear, John Goodwin, Katalin Kovacs, Glen Hart
Research & Innovation
Overview
Introduction to Ordnance Survey’s data
Authoring topographic ontologies
 Open world modelling, closed world thinking
Linking ontologies to our database
 Immature technology
Introduction to Ordnance Survey
 National Mapping Agency of Great Britain
 Digital data accounts for 80% of business
activity
 3.5 terabytes of contiguous feature based
data
 5,000 updates a day, change captured
within 6 months
 2000+ Concepts
OS MasterMap – integrated layers
Overview
Introduction to Ordnance Survey’s data
Authoring topographic ontologies
 Open world modelling, closed world thinking
Linking ontologies to our database
 Immature technology
Our approach to ontologies
Capturing organisational knowledge about topography
Driving database modelling vs using legacy systems
Up to 80% of applications have a geospatial component
 Topography not a stand-alone application
 Our data nearly always integrated with other datasets
Authoring topographic ontologies
Domain Ontology
Conceptual
Aspect
Built by
domain
expert
A River
Stretch is
part of a
River
 River_
Stretch
direct_part_of
∃ River
Computational
Aspect
Ontology
engineer
translates
into OWL
<owl:ObjectProperty rdf:ID="directPartOf">
<rdf:type rdf:resource="&owl;FunctionalProperty"/>
</owl:ObjectProperty>
<owl:Class rdf:ID="River"/>
<owl:Class rdf:ID="RiverStretch">
<rdfs:subClassOf>
<owl:Restriction>
<owl:onProperty rdf:resource="#directPartOf"/>
<owl:someValuesFrom rdf:resource="#River"/>
</owl:Restriction>
</rdfs:subClassOf>
</owl:Class>
Domain experts think in a closed world
 Structured sentences default to disjoint, XOR etc.
“River can overlap with Stream”
 Closure axioms are needed
“A River flows into the Sea or a Lake or a River”
(it has to flow into something, and can only
flow into these things, and nothing else)
 Difficulties in modelling “sometimes” or “may” in OWL
“A Channel may contain Water”
(All Channels enable Some ContainmentOfWater
All TypicalChannels contain Some Water)
Ontologies vs Databases: Issues
Is it an instance or a concept?
 Water
 Winter
Should it be stored in the ontology or database?
 East Anglia, Midlands, the Fens
 “A Burn is a Stream in Scotland”
Identity
Identity of Place - Topographic Identifiers (TOIDs)
Woodland
Naming
Land
Networks
cover
geography
Itchen Valley
Southampton, Basset
A Road
Populated Land
Roundabout
Overview
Introduction to Ordnance Survey’s data
Authoring topographic ontologies
 Open world modelling, closed world thinking
Linking ontologies to our database
 Immature technology
Data repurposing using ontologies
has Form
Sluice Gate
Ordnance
Survey
ontology
has Form
Weir
Sluice
Weir
Is a
Flood
defence
Is a
Environment
Agency
ontology
OS
Database
Database - ontology mapping
SPARQL
query
Query
result
D2RQ Server
Joseki SPARQL service
JDBC
connection
Oracle spatial
database
D2RQ
RDF virtual
graph
D2RQ Map
Jena
OWL ontology
DIG reasoner
(e.g. Pellet)
Database & ontology structural differences
Ontologies designed to work with RDF triple data,
not relational databases
Don’t bury semantics in mapping file
Efficiency is a big issue
hasFootprint
hasTOID
Building
Footprint
has Use
TOID
Use
is a
University
Building
is a
Database Feature Table
TOID
Type
Form
1234
Building
Church Place of Worship SDO_GEOMETRY
Shape
5678 Inland Water
River
Inland Navigation SDO_GEOMETRY
9101
House
Residence
Residence
…
House
Function
Building
….
…
….
SDO_GEOMETRY
SDO_GEOMETRY
How can we do this…
Ontology: River -> flowsInto some Sea
Data:
 Forth rdf:type River
 Tay rdf:type River
 Clyde rdf:type River
 Forth flowsInto North_Sea
 Tay flowsInto North_Sea
 North_Sea rdf:type Sea
Query:
select ?a
where {?b rdf:type Sea
?a flowsInto ?b}
Closed World result: Forth, Tay
Open World result: Clyde as well.
Summary
Domain experts find open-world thinking difficult
Software tools still have some way to go
Ontologies will help improve database design
Open world vs closed world isn’t our biggest problem
– content is.
Download