Creating translation rules for integration of geo-data sets using OWL (Web Ontology Language)

jumentousmanlyInternet και Εφαρμογές Web

21 Οκτ 2013 (πριν από 3 χρόνια και 8 μήνες)

63 εμφανίσεις

May 26, 2004
1
Creating
translation
rules
for
integration
of
geo-data
sets
using
OWL
(Web
Ontology
Language)
Marne-la-
Vallée
/EuroSDR
workshop, 15-16 April 2004
Marian de
Vries
OTB, Section GIS Technology
May 26, 2004
2
Geo-information
integration

Combine geo-data from different data sources in one
(Web) client

Issues

consistent
map
visualization
(classification,
legenda)

integrated querying over these different data
sources (transparent to the user):
“select
all
18
th

century
buildings
on
the
topographic
map
that
are
also
official
national
monuments

May 26, 2004
3
Geo-information
integration

Ambition
levels
->
integration
levels

2003:
EuroSDR workshop + EC-GIS workshop
(
Woodsford,
Luzet
)

Hybrid mapping

Consistent mapping
May 26, 2004
4
Requirements
->
translation
strategies
At
data
source
(views
on
tables,
automated
replication)
Pre-defined
High
Time-critical
applications
By
application
software
(Styled
Layer
Descriptor,
XQuery

filters,
ontology
reasoners)
On-the-fly,
during
data
retrieval
Medium
or
high
Projects
or
structural
cooperation
-
-
Best guess, fuzzy
Exploratory,
ad
hoc
Where
and
How
When
Quality of
result
Type
of
use
May 26, 2004
5
Adapting
existing
data
models?

‘Pre-defined’, e.g. federated database

within enterprise or across organizations on a
structural basis

local database schemas
can
be adapted

On-the-fly combining of data (Web mapping)

users not known, user queries not known

local database schemas can
not
be adapted
to all
these different (unknown) information needs
May 26, 2004
6
Relevant
research
fields

Database research: federated databases, schema
integration

Digital library science: information search strategies,
taxonomies,
false
negatives,
false
positives

Semantic Web: rich meta-data with help of
ontologies

Knowledge engineering: Description Logics

Software engineering: intelligent agents, reasoners
May 26, 2004
7
Example:
IMKICH
=
cultural
heritage

4 Dutch organizations involved in cultural history

Partially overlapping data sets about:

(national) monuments

archeological
sites

historical landscape elements

Goal of project:

offering integrated access to these data sets via
standards based Web services
May 26, 2004
8
Example:
IMKICH
=
cultural
heritage

Intended
users
are:

civil
servants
involved
in
urban
planning
and
rural
planning (= main purpose of project)

expert users from within the domain (archaeologists,
historians,
landscape
architects)

other
professional
GI
users
->
questions
are
unknown, combination with other data sets is
unknown
May 26, 2004
9
Step
1:
describe
existing
data
sets

Look
at
implemented
(technical)
database
schemas

Look at data set
content
(the actual attribute values of
the object instances)

Reverse-engineer
into
conceptual/logical
models
(UML
class diagrams)

Talk to the experts (database designers, information
modellers)
May 26, 2004
10
Demo
data
content
May 26, 2004
11
Classify
content
(attribute
values)
May 26, 2004
12
Step
2:
design
‘global

data
model
May 26, 2004
13
Step
3:
from
UML
to
OWL
ontologies

Semantically
identical
instances
->
owl:Class

look
for
class-subclass
relations
(taxonomy-like):
rdfs:subClassOf

define restrictions/constraints: owl:
hasValue,
owl:
allValuesFrom, owl:
intersectionOf

(re)group
classes
with
owl:
unionOf


May 26, 2004
14
Example
1
<owl:Class
rdf
:ID="Church">
<
rdfs
:
subClassOf rdf
:resource="#Building"/>
<
rdfs
:
subClassOf
>
<owl:Restriction>
<owl:
onProperty rdf
:resource="#
hasFunction
"/>
<owl:
hasValue rdf
:resource="#Religion"/>
</owl:Restriction>
</
rdfs
:
subClassOf
>
</owl:Class>
<owl:Class
rdf
:ID=

Cathedral">
<
rdfs
:subClassOf rdf
:resource="#Church"/>
</owl:Class>
May 26, 2004
15
Example
2
<owl:Class
rdf
:ID="
ArcheologicalMonument">
<
rdfs
:
subClassOf rdf
:resource="#
CultHistObject
"/>
</owl:Class>
<owl:Class
rdf
:ID="
ProtectedArcheologicalMonument
">
<owl:
intersectionOf rdf
:
parseType
="Collection">
<owl:Class
rdf
:about="#
ArcheologicalMonument
"/>
<owl:Restriction>
<owl:
onProperty rdf
:resource="#
legalStatus
"/>
<owl:
hasValue rdf
:resource=

#
officialMonument

/>
</owl:Restriction>
</owl:
intersectionOf
>
</owl:Class>
May 26, 2004
16
Step
4:
match
ontologies

Comparable attributes -> owl:
EquivalentProperty

Comparable
classes
->
owl:
EquivalentClass

Options:

Only match local and global models

Also construct relations between local and local
models (peer-to-peer)
May 26, 2004
17
Mapping
ontologies
:
issues

Less possibilities to specify formal constraints on
individual values than e.g. OCL (Object Constraint
Language)

c.f. OCL invariant: Road.
speedLimit >= 120

Within an ontology property names must be unique
(concatenate
with
class
name
e.g.)
->
difference
between the OWL property name and the actual
database attribute name
May 26, 2004
18
C.f.

create
view
’ (SQL)
CREATE VIEW
CultHistObject
AS
SELECT
concat
(
'archis
.',id#) AS fid,
decode(WAARDE,
'archeologische betekenis'
, 'not assessed'
...

'hoge archeologische waarde'
, 'high'

'zeer hoge archeologische waarde'
, 'very high'

'zeer hoge archeologische waarde
,
beschermd'
, 'very high'
'unknown') AS
cultHistValue
,
decode(WAARDE,
'archeologische betekenis'
, 'not assessed yet'

'zeer hoge archeologische waarde
,
beschermd'
, 'monument'
'none') AS
juridStatus
,
to_char(null) AS
partOfCultHistObj
#,
...
FROM
archis
UNION
SELECT
concat
(
'monreg
.',id#) AS fid,
to_char(null) AS
cultHistValue
,
'national monument' AS
juridStatus
,

concat
(
'monreg
.',Complex-id#) AS
partOfCultHistObj
#,
...
FROM
monreg
UNION ...
May 26, 2004
19
(Re)structure:
test
the
possibilities
<owl:Class
rdf
:ID="Highway">
<owl:
equivalentClass
>
<owl:Restriction>
<owl:
onProperty rdf
:resource="#
hasRoadType
" />
<owl:
hasValue
>highway</owl:
hasValue
>
<!-- is this permitted?
<owl:
hasValue rdf
:resource="#Highway" /-->
</owl:Restriction>
</owl:
equivalentClass
>
</owl:Class>
May 26, 2004
20
First
findings

Concentrate on content of

core’ selection attributes:
legalStatus
, cultHistValue (attribute values)

Create taxonomies with class-subclass relations

for constructing map classification

maybe: as input for search agents (broadening,
narrowing selection criteria)
May 26, 2004
21
Conclusion
1:
use
of
OWL

For
conceptual
modeling
+/-

Highway:
Road.
speedLimit
>=
100
not
possible

For
ontology
matching
++

EquivalentClass
,
EquivalentProperty
,
unionOf

For
conversion
scripts
(local
->
global
database
scheme)
+/-

must
have
unique
names
for
properties,
loose
1
to
1
mapping
to
database
schemes

For
helping
user
with
search/key
words
construction
(thesaurus)
++

For
on-the-fly
mediation:
depends
on
capabilities
of
reasoner
software
->
future
research
May 26, 2004
22
Conclusion
2:
creating
domain
model

Get the requirements right:

Mission
critical?
(are
correct
and
complete
answers
of critical importance?)

Intended users

Links
with
a
thesaurus?

So: how important is harmonization of vocabulary?

Look at actual content (attribute values), hidden
semantics
can
be
discovered
this
way
May 26, 2004
23
Quest
for
interoperability

Syntactic interoperability: almost achieved?

OpenGIS
interface specifications for WMS, WFS, WCS,
WTS …

OGC Web service

meta data
’ operations:

GetCapabilities
request

DescribeFeatureType

request
May 26, 2004
24
Quest
for
interoperability

Next step: semantic interoperability

How?

OpenGIS DescribeFeatureSemantics
request ?

Reference to a ‘Semantic Reference System

(W.
Kuhn) for each feature type (c.f. EPSG:... for the
Spatial Ref. System) ?