A Lightweight Approach to Explore, Enrich and Use Data with a Geospatial Dimension with Semantic Web Technologies

A Lightweight Approach to Explore, Enrich
and Use Data with a Geospa9al Dimension
using Seman9c Web Technologies
Christophe Debruyne (TCD), Kris McGlinn (TCD),
Lorraine McNerney (OSi), and Declan O’Sullivan (TCD)
2017-05-14 @ GeoRICH 2017
The ADAPT Centre is funded under the SFI Research Centres Programme (Grant 13/RC/2106) and is co-funded under the European Regional Development Fund.

www.adaptcentre.ieIntroduc9on
•  Linked Data is a set of best pracXces and guidelines to publish and
interlink data on the Web by cleverly combining several standardized
technologies.
•  GeospaXal informaXon is an important part of the LD Web. In fact, most
datasets have a geospaXal dimension and loca%on provides a convenient
way to explore, analyze and align datasets.
•  Its importance is evidenced by the vast amount of geographic data on the
Web and bespoke tools that are available that support the consumpXon of
such data (e.g., GeoSPARQL-enabled triplestores).
•  How can we limit relying on such bespoke tools, and leverage engagement
with geospaXal Linked Data to analyze, explore and enrich non-RDF data?

www.adaptcentre.ieContext – Open Data Engagement Fund
•  Organized by the Department of Public Expenditure and Reform in
conjuncXon with the Open Data Governance Board
•  The goal of this iniXaXve is to improve the availability and usage of data
on the hap://data.gov.ie/ portal
•  We observed that not many datasets are available as RDF and thus also
not available as Linked Data. They are available as CSV, TSV, etc. But,
many datasets have a geospaXal dimension, which ideally should be linked
to authoritaXve geospaXal Linked Data (as provided by the Ordnance
Survey Ireland).

www.adaptcentre.ieContext – data.geohive.ie
data.geohive.ie is an ongoing
collaboraXon between ADAPT and
the Ordnance Survey Ireland to
publish OSi’s authoritaXve
geospaXal informaXon as Linked
Data.

StarXng from publicly available
boundary data, supporXng two
use cases: provision of diﬀerent
geometries for features, and
provenance and evoluXon of
features and their geometries

www.adaptcentre.ieApproach
We propose a method and a set of tools for:
•  Uplif: transforming non-RDF into RDF
•  Interlinking
•  Downlif: transforming back to non-RDF, and
•  Engagement
CSV
File
Uplift
Data
in RDF
Links
in RDF
Merge /
Combine
Enriched
descriptions
in RDF
Engage
Enriched CSV
File
Downlift
!
Create Links
Link
Discovery
Tools
SPARQL
CONSTRUCT
queries
...

www.adaptcentre.ieBackground
Triple PaJern Fragments (Verborgh et al. 2016)
•  SPARQL endpoints require a lot of resources on one’s server. And one
ofen provides data dumps and resolvable URIs as a “good enough”
pracXce to avoid this problem (Verborgh et al. 2016)
•  Distribute load between a TPF client and server
•  Less load on the server at the cost of increased bandwidth
•  Easy to setup locally and to “simulate” federated queries

GeoSPARQL
•  An OGC standard for represenXng and querying geospaXal data on the
SemanXc Web. Deﬁnes both a vocabulary for represenXng geospaXal
informaXon and an extension to the SPARQL query language.
R. Verborgh, M. Vander Sande, O. HarXg, J. Van Herwegen, L. De Vocht, B. De Meester, G. Haesendonck, and P.
Colpaert. 2016. Triple Paaern Fragments: A low-cost knowledge graph interface for the Web. J. Web Sem. 37-38
(2016), 184–206

www.adaptcentre.ieR2RML
R2RML: RDB to RDF Mapping Language
•  A W3C RecommendaXon since fall 2012
•  CreaXng an R2RML file that annotates a relaXonal database with exisXng
vocabularies and/or ontologies (RDFS or OWL).
•  That R2RML file goes through an R2RML Mapping Engine to produce RDF.
•  R2RML specified
•  An ontology to specify those mappings;
•  How those mappings should be interpreted to produce RDF.
•  R2RML files are thus stored as RDF.

www.adaptcentre.ie
@prefix rr: <hap://www.w3.org/ns/r2rml#> .
@prefix foaf: <hap://xmlns.com/foaf/0.1/> .
@prefix dbpedia: <hap://dbpedia.org/ontology/> .

<#CityTriplesMap>
a rr:TriplesMap ;

rr:logicalTable [ rr:tableName "City" ] ;

rr:subjectMap [
rr:template "hap://foo.example/City/{ID}" ;
rr:class dbpedia:Place ;
] ;

rr:predicateObjectMap [
rr:predicate foaf:name ;
rr:objectMap [ rr:column "Name" ] ;
] ;
.
What is being
mapped? A logical
table/view or an SQL
query.
How to generate and
state something about
the subject of those
triples.
How to generate
predicates and objects.
City
ID Name
1 Dublin
2 Ghent
Background – R2RML Example

www.adaptcentre.ieApproach – Running Example
•  Fingal County Council Weather StaXons
•  Example from the hap://data.gov.ie/ portal
Name
Weather
Reading Agency LAT LONG
M50
Blanchardstown hap://…
NaXonal Roads
Authority
53.3704660
3
-6.38085144
7
M50 Dublin Airport hap://…
NaXonal Roads
Authority
53.4096411
1
-6.22759742
8
Dublin Airport hap://… Met Éireann
53.4215060
8 -6.29784754
Records in fccweatherstaXonsp20110829-2221.csv

www.adaptcentre.ieApproach – Uplift
@prefix odef: <http://adaptcentre.ie/ont/odef#> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
@prefix csv: <file:///...#> .
<#TriplesMap>
rr:logicalTable [
rr:sqlQuery """SELECT rownum() AS ROW_NUM, *
FROM fccweatherstationsp201108292221;"""
] ;
rr:subjectMap [
rr:template "http://www.example.org/record/{ROW_NUM}" ;
rr:class odef:Record ;
] ;
rr:predicate csv:LONG ;
odef:label "LONG" ;
odef:order "5"^^xsd:int ;
rr:objectMap [ rr:column "LONG" ] ;
] ;
# Other predicate-object maps omitted for brevity
rr:predicate csv:ROW_NUM ;
rr:objectMap [ rr:column "ROW_NUM" ] ;
] .
Adopt the W3C RDB to RDF
Mapping Language
(R2RML) to generate a
mapping from CSV to RDF
Insert metadata containing
informaXon about the non-
RDF resource’s structure in
the mapping

www.adaptcentre.ieApproach – UpliX
Generate RDF with R2RML-F – an implementaXon of R2RML supporXng
funcXons in JavaScript (haps://opengogs.adaptcentre.ie/debruync/r2rml)
Christophe Debruyne, Declan O'Sullivan: R2RML-F: Towards Sharing and
ExecuXng Domain Logic in R2RML Mappings. LDOW@WWW 2016
<http://www.example.org/record/2>
a <http://adaptcentre.ie/ont/odef#Record> ;
<file:///...#AGENCY> "National Roads Authority" ;
<file:///...#LAT> "53.4096411069945" ;
<file:///...#LONG> "-6.22759742761812" ;
<file:///...#NAME> "M50 Dublin Airport" ;
<file:///...#ROW_NUM> "2"^^<http://www.w3.org/2001/XMLSchema#int> ;
<file:///...#WEATHER_READING> "http://www.nratraffic.ie/..." ;
geo:asWKT "POINT(-6.22759742761812 53.4096411069945)"^^geo:wktLiteral .

www.adaptcentre.ieApproach – Interlinking
Either with SPARQL CONSTRUCT queries or use the RDF using link discovery
tools such as Silk (Jentzsch et al. 2010) or LIMES (Ngonga et al. 2011)

SPARQL CONSTRUCT queries can be formulated using TPF. For support for
GeoSPARQL, we use an extension of the TPF client (Debruyne et al. 2017)

In our running example, we can adopt GeoSPARQL to ﬁnd out where the
weather staXons are based by either:
•  ConcatenaXng the LAT and LONG to create a WKT point
•  Extending the skeleton mapping and generate WKT points
Anja Jentzsch, Robert Isele, ChrisXan Bizer: Silk - GeneraXng RDF Links while Publishing or Consuming
Linked Data. ISWC Posters&Demos 2010
Axel-Cyrille Ngonga Ngomo, Sören Auer: LIMES - A Time-Eﬃcient Approach for Large-Scale Link Discovery
on the Web of Data. IJCAI 2011: 2312-2317
Christophe Debruyne, Eamonn Clinton, Declan O'Sullivan: Client-side Processing of GeoSPARQL FuncXons
with Triple Paaern Fragments. LDOW@WWW 2017

www.adaptcentre.ieApproach – Interlinking
rr:predicate geo:asWKT ;
rr:objectMap [
rr:template "POINT({LONG} {LAT})" ;
rr:termType rr:Literal ;
rr:datatype geo:wktLiteral ;
] ;
] ;
A B
C
D
ExecuXng a CONSTRUCT query with
A.  OSi's TPF server
B.  the local TPF server
C.  the query (using our extension), and
D.  the resulXng RDF graph.

Runs in a browser, and one does not need
to install and populate a GeoSPARQL-
enabled triplestore.

Straigh~orward federated querying.
Extending the R2RML Mapping

www.adaptcentre.ieApproach
Two possible approaches to extending TPF with GeoSPARQL:

A) Extending a TPF Client
•  TPF server specificaXon intact (backwards compaXble)
•  Possibly more network overhead

B) Extending the TPF server
•  Outside server specificaXon, but proven to be viable for
substring filtering (Van Herwegen et al. 2015)
Addi9onal requirement: a pure JavaScript implementa9on
•  Allows one to run the client in a browser and hence facilitate
stakeholders in formulaXng GeoSPARQL queries
J. Van Herwegen, L. De Vocht, R. Verborgh, E. Mannens, and R. Van de Walle. 2015. Substring Filtering for Low-Cost
Linked Data Interfaces. In The SemanXc Web - ISWC 2015 - 14th InternaXonal SemanXc Web Conference,
Bethlehem, PA, USA, October 11-15, 2015, Proceedings, Part I (LNCS), Vol. 9366. Springer, 128–143.

www.adaptcentre.ieApproach – DownliX
1.  Merge generated RDF and links
2.  Provide downlif algorithm
1.  The merged RDF
2.  The R2RML mapping – to reconsXtute the original ﬁle
3.  A set of predicates (URIs) – for the addiXonal columns.
Empty values will be generated when no links exist
Name
Weather
Reading … hJp://www.opengis.net/ont/geosparql#within
M50
Blanchardstown hap://… … hap://data.geohive.ie/resource/county/2AE19629144F13A3E055000000000001
M50 Dublin Airport hap://… … hap://data.geohive.ie/resource/county/2AE19629144F13A3E055000000000001
Dublin Airport hap://… … hap://data.geohive.ie/resource/county/2AE19629144F13A3E055000000000001

www.adaptcentre.ieImplementa9on
On the tools extension for a TPF Client
•  Uplif: haps://opengogs.adaptcentre.ie/debruync/generate-mapping
•  Downlif: haps://opengogs.adaptcentre.ie/debruync/downlif
•  Our R2RML engine: haps://opengogs.adaptcentre.ie/debruync/r2rml
•  All available with accessible licenses (MIT)

On the GeoSPARQL extension for a TPF Client
•  We extended V2.0.4 of the TPF Node.js Client
(Available at haps://github.com/chrdebru/Client.js)
•  Made available a web-client using this extension
(At hap://theme-e.adaptcentre.ie/geo-tpf/)

www.adaptcentre.ieDiscussion
On the meaning of the generated RDF and enriched CSV
•  Goal was to enrich non-RDF with a geospaXal component, that also
allowed one also to engage with that data
•  The mappings that are generated are not necessarily meaningful; the
generated RDF is just as meaningful as the original ﬁle and users are not
obliged to adopt established vocabularies, but they can…
•  AdopXon of established vocabularies would allow for more meaningful
links, which are then stored back into the CSV, and the mapping would
provide a basis for creaXng meaningful RDF, and even Linked Data.
On the datasets
•  We used a simple dataset and only demonstrated SPARQL CONSTRUCT
with GeoSPARQL. More exercises are conducted in the context of the
Open Data Engagement Fund; to be reported.

www.adaptcentre.ieDiscussion (not in paper – new)
Apprecia9on during public seminars
•  About 40 registered aaendees from academia, government agencies, and
industry held on the 4th and 10th of May.
•  First seminar on Linked Data (principles and case studies)
•  Second seminar on Linked Data publishing uplif
•  A recurring concern about Linked Data publishing and interlinking is
relying on third parXes to conduct the exercise, and how one can acquire
the experXse.
•  Simulated direct mapping, and its generator, was appreciated as it
allowed to incrementally generate "meaningful" linked data. Downlif was
appreciated as it facilitated relying on their exisXng processes.

www.adaptcentre.ieWith respect to related work…
The GeoKnow EU FP7 Project (hap://www.geoknow.eu/)
•  Relies on a parXcular sofware stack
•  No explicit noXon of downlif
•  Provides support uplif of unstructured text and shapeﬁles
Our approach was aimed to be lightweight, but
we are aware that approach will likely not scale
as well as bespoke systems built for this
parXcular purpose. Our goal was to leverage
the enrichment of non-RDF open data with
a geospaXal dimension with
semanXc technologies.

www.adaptcentre.ieSummary, lessons learned, and future work
Summary
•  We presented a lightweight approach to uplif, interlinking and enriching,
and downlif, as well as engage with that informaXon.
•  Developed in the context of a iniXaXve to improve the usage and
availability of open data on the Irish open data portal
•  Prototypes have been made available with an accessible license
Lessons learned
•  We deem our approach viable. Even though we are aware our approach
will not outperform certain bespoke systems in terms of GeoSPARQL
processing and analysis, our goal was to leverage aforemenXoned
processes.
Future work
•  User studies, and inclusion of governance aspects
•  AddiXonal exercises with other datasets to be reported in a whitepaper

A Lightweight Approach to Explore, Enrich and Use Data with a Geospatial Dimension with Semantic Web Technologies

More Related Content

What's hot (6)

Similar to A Lightweight Approach to Explore, Enrich and Use Data with a Geospatial Dimension with Semantic Web Technologies (20)

More from Christophe Debruyne (20)

Recently uploaded (20)

A Lightweight Approach to Explore, Enrich and Use Data with a Geospatial Dimension with Semantic Web Technologies