Uplift – Generating RDF datasets from non-RDF data with R2RML

Upli% – Genera-ng RDF datasets
from non-RDF data with R2RML

h>p://bit.ly/upli%-r2rml
Dr. Christophe Debruyne

BACKGROUND
•  Resource Descrip-on Framework (RDF)
•  RDF Schema
•  SPARQL
•  Linked Data
04/11/17 christophe.debruyne@adaptcentre.ie 2

Context
•  The Open Data Engagement Fund – developed by both the Department of Public Expenditure
and the Open Data Governance Board – aims to provide support towards improving the
availability and usage of data on data.gov.ie.

•  We observe that the vast majority of the data on the portal provide CSV (or XML), which are
structured, but not necessarily (as) meaningful (as RDF datasets using recognized
vocabularies)
•  The goal of this seminar is to inform how one can generate RDF datasets from the non-RDF
datasets – a process called upli% – that are available on the open data portal using a W3C
Recommenda-on called R2RML – which stands for RDB to RDF Mapping Language.
•  The genera-on of RDF and the adop-on of appropriate vocabularies not only improve the
availability, but also its use and even accessibility both at a technical level (querying), and at
the level of understanding (seman-cs).
3 04/11/17 christophe.debruyne@adaptcentre.ie

Background – RDF
•  Stands for Resource Descrip4on Framework
•  RDF is not a language but a model (!!!)
•  RDF is a W3C recommenda-on
•  RDF is designed to be read by computers
•  RDF is for describing resources on the Web
in terms of triples (subject – predicate – object) [in graphs]
•  RDF uses URIs to iden-fy and refer to resources on the Web
•  RDF/XML is just one way of serializing RDF
–  We will use the Terse RDF Triple Language – TURTLE, which does not
provide support for named graphs
–  A serializa-on format with support for named graphs is N-QUADS

Background – RDF
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix voc: <http://example.org/example#> .
<http://christophedebruyne.be> voc:title "Christophe’s Page" .
<http://christophedebruyne.be> voc:topic "Cats" .
<http://christophedebruyne.be> voc:owned <http://example.org/#xof> .
@prefix voc: <http://example.org/example#> .
<http://christophedebruyne.be>
voc:title "Christophe’s Page" ;
voc:topic "Cats" ;
voc:owned <http://example.org/#xof> ;
.

Background – RDFS (and OWL)
•  RDF provides li>le support to describe a domain; we can declare proper-es,
state that things are of a certain type, declare collec-ons and containers, but
that’s about it…
•  RDF Schema (RDF) is an extension of RDF – also a W3C Recommenda-on
•  RDFS provides a framework for describing vocabularies
•  RDFS describe resources with classes, proper-es and values
•  Allows us to do some “lightweight” reasoning; infer implicit informa-on from
explicit informa-on
•  The Web Ontology Language (OWL) – not covered here – is an even more
expressive ontology language suitable for more complex reasoning tasks

Background – RDFS
@base <http://foo.bar/> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix foaf: <http://xmlns.com/foaf/0.1/> .
<#Cat> a rdfs:Class ; rdfs:label "Cat" .
<#owns> a rdf:Property ; rdfs:label "owns" ;
rdfs:domain foaf:Person ; rdfs:range <#Cat> .
<#victor> a <#Cat> ; foaf:name "Victor" .
<#louis> a <#Cat> ; foaf:name "Louis" .
<#bettina> a <#Cat> ; foaf:name "Bettina" .
<#chrdebru> a foaf:Person ; foaf:name "Christophe Debruyne" ;
<#owns> <#victor> ; <#owns> <#louis> ;
<#owns> <#bettina> .
7
Exercise: What could we infer from <#kevin> <#owns> <#gomez> ?
04/11/17 christophe.debruyne@adaptcentre.ie

Background – SPARQL
•  Stands for SPARQL Protocol and RDF Query Language
–  Recursive acronym
•  SPARQL 1.1 is a W3C Recommenda-on since March 2013
(h>p://www.w3.org/TR/sparql11-overview/)
•  SPARQL allows us to:
–  Pull values from structured and semi-structured data
–  Explore data by querying unknown rela-onships
–  Perform complex joins of disparate databases in a single, simple query
–  Transform RDF data from one vocabulary to another
–  …

Background – SPARQL
Structure of a SPARQL SELECT Query.
A SPARQL query comprises, in order:
•  Prefix declara4ons for URI shorthands
•  A result clause, iden-fying what
informa-on to return from the query
•  Dataset defini4on, sta-ng what RDF
graph(s) are being queried
•  The query paVern, specifying what to
query for in the underlying dataset
•  Query modifiers, slicing, ordering, and
otherwise rearranging query results
# prefix declarations
PREFIX foaf: <http://.../0.1/>
...
# result clause
SELECT ...
# dataset definition
FROM ...
# query pattern
WHERE {
...
}
# query modifiers
ORDER BY ...
9
Namespace shortened for brevity
Adapted from “SPARQL by Example” by Feigenbaum and Prud’hommeaux 04/11/17 christophe.debruyne@adaptcentre.ie

Background – Simple SPARQL Queries
•  Variables start with a ques-on
mark and can match any node
(resource or literal);
•  Triple pa>erns are just like triples,
except that any of the parts of a
triple can be replaced with a
variable;
•  The SELECT result clause returns a
table of variables and values that
sa-sfy the query.
•  “a” is syntac-c sugar for “rdf:type”
Finding all cats and their names
PREFIX ex: <http://foo.bar/#>
PREFIX foaf: <http://.../0.1/>
SELECT ?cat ?name
WHERE {
?cat a ex:Cat.
?cat foaf:name ?name.
}


Background – Simple SPARQL Queries
11
PREFIX ex: <http://foo.bar/#>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
SELECT ?cat ?name
WHERE { ?cat a ex:Cat.
?cat foaf:name ?name. }
-----------------------------------------
| cat | name |
=========================================
| <http://foo.bar/#bettina> | "Bettina" |
| <http://foo.bar/#louis> | "Louis" |
| <http://foo.bar/#victor> | "Victor" |
-----------------------------------------

Background – GeoSPARQL
Give a me a list of coun-es whose
borders touch County Dublin
12
SELECT ?county ?w2 WHERE {
?dublin a osi:County .
?dublin rdfs:label "DUBLIN" .
?county a osi:County .
?dublin geo:hasGeometry ?g1 .
?g1 geo:asWKT ?w1 .
?county geo:hasGeometry ?g2 .
?g2 geo:asWKT ?w2
FILTER (geof:sfTouches(?w1, ?w2)) .
}
RDF allows one to engage with data in a declara-ve
manner, not bound to par-cular structure.

Linked Data
•  Linked Data started oﬀ as a ini-a-ve called the Linking Open Data (LOD)
project.

•  Linked Data is a global ini-a-ve to publish and interlink structured data on
the Web using a clever combina-on of simple, standardized technologies.
–  Uniform Resource Iden-ﬁers – to name things;
–  Resource Descrip-on Framework – to represent things;
–  HTTP infrastructure – to obtain those representa-ons.
04/11/17 13 christophe.debruyne@adaptcentre.ie

(Non-)Informa-on Resources
•  Informa-on resources are documents – referred to by a URI – that
describe non-informa-on resources – named with a URI – that represent
things such as cars, people, etc.
•  The NIR http://dbpedia.org/resource/James_Joyce is described by
the following IRs:
•  The web page http://dbpedia.org/page/James_Joyce
•  The RDF doc http://dbpedia.org/data/James_Joyce
•  Either is returned depending on what you need. How?
04/11/17 14 christophe.debruyne@adaptcentre.ie

Content Nego-a-on
04/11/17 15
Image from h>p://www.w3.org/TR/swbp-vocab-pub/
christophe.debruyne@adaptcentre.ie

Content nego-a-on in data.geohive.ie
Proxy Server
TPF
Server
TPF
Web
Client
Linked
Data
Frontend
Website
(dumps)
Ontologies
SPARQL Endpoint
Triplestore
TPF Client

UPLIFT WITH R2RML

Diﬀerent types of mappings…
•  Correspondences vs. Mappings
•  Correspondences capture how en--es are related
•  Mappings describe how to relate en--es
•  Correspondences are 'symmetrical', mappings have a direc-on
•  Matching vs. Mapping
•  Diﬀerent types of mappings
•  Between ontologies
•  Between datasets or instances
•  Upli%: From non-RDF to RDF
•  Downli%: from RDF to non-RDF

RDB2RDF
foaf:based_near
<person/1> <person/2>
foaf:Person
<city/1> <city/2>
ChristopheDebruyne Kevin
GhentDublin
rdf:type rdf:type
foaf:firstNamefoaf:lastName
foaf:namefoaf:name
foaf:based_near
foaf:firstName
Person
ID First Last CityID
1 Christophe Debruyne 1
2 Kevin NULL 2
City
ID Name
1 Dublin
2 Ghent

RDB2RDF: W3C Recommenda-ons
•  It all started with Tim Berners-Lee proposing a direct mapping from
rela-onal databases to RDF.
•  Over the years, two W3C Recommenda-ons (standards) to map rela-onal
data to RDF emerged.
–  A Direct Mapping of rela-onal data to RDF
–  R2RML: an RDB to RDF Mapping Language that is highly customizable
to annotate rela-onal data with ontologies to generate RDF.

RDB2RDF
Tim Berners-Lee described a mapping between rela-onal databases and RDF that
can be automated, see “Rela-onal Databases on the Seman-c Web” via
h>p://www.w3.org/DesignIssues/RDB-RDF.html
“The seman-c web data model is very directly connected with the model of
rela-onal databases. A rela-onal database consists of tables, which consists of
rows, or records. Each record consists of a set of fields. The record is nothing but
the content of its fields, just as an RDF node is nothing but the connec-ons: the
property values. The mapping is very direct:
•  a record is an RDF node;
•  the field (column) name is RDF propertyType; and
•  the record field (table cell) is a value.”
But can it be that simple? Can you give examples?

Direct Mappings
TBL proposed a direct mapping. Direct mappings immediately reflect the
structure of the database à The target RDF vocabulary directly reflects the
names of database schema elements, and neither structure nor target vocabulary
can be changed. [R2RML]
Process:
•  Exis4ng table and column names are encoded into URIs.
•  Data is (i) extracted, (ii) transformed into RDF and then (iii) loaded into a
triplestore. This is thus an ETL process.
This proposal – over -me – was refined into a W3C recommenda-on, published in
fall 2012, called “A Direct Mapping of Rela-onal Data to RDF”
h>p://www.w3.org/TR/rdb-direct-mapping/

Direct Mappings
•  The database (both schema and data), primary keys and foreign keys are
given to a direct mapping engine to produce an RDF graph.
–  Fields are mapping to literals;
–  Primary keys are used to construct URIs for resources;
–  And foreign keys are used to construct proper-es and relate
resources.
•  Example…

Direct Mapping Example (from W3C)
@base <http://foo.example/DB/> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
<Person/ID=1> rdf:type <Person> .
<Person/ID=1> <Person#ID> 1 .
<Person/ID=1> <Person#First> "Christophe" .
<Person/ID=1> <Person#Last> "Debruyne" .
<Person/ID=1> <Person#CityID> 1.
<Person/ID=1> <Person#ref-CityID> <City/ID=1> .
<Person/ID=2> rdf:type <Person> .
<Person/ID=2> <Person#ID> 2 .
<Person/ID=2> <Person#First> "Kevin" .
<Person/ID=2> <Person#CityID> 2.
<Person/ID=2> <Person#ref-CityID> <City/ID=2> .
<City/ID=1> rdf:type <City> .
<City/ID=1> <City#ID> 1.
<City/ID=1> <City#Name> "Dublin" .
<City/ID=2> rdf:type <City> .
<City/ID=2> <City#ID> 2.
<City/ID=2> <City#Name> "Ghent" .
Given a base URI
h>p://foo.example/DB/
Discuss: why would a
base URI be important?
City
ID Name
1 Dublin
2 Ghent
Person
2 Kevin NULL 2

Direct mappings for tabular data
•  Genera-ng RDF from Tabular Data on the Web
•  A W3C Recommenda-on since 2015 - h>ps://www.w3.org/TR/csv2rdf/
•  A direct mapping approach for CSV and tabular data
•  Note: The Recommenda-on provides a sec-on on comparing this
approach to the direct mappings for rela-onal databases; “seman-cally
equivalent” when dealing with one table, not guaranteed to be
“seman-cally equivalent” when mul-ple interrelated tables are involved.

Direct Mappings
Discussion:
Are direct mappings meaningful?
Can you iden-fy poten-al problems?

R2RML
•  R2RML: RDB to RDF Mapping Language
–  A W3C Recommenda-on since fall 2012
–  h>p://www.w3.org/TR/r2rml/
•  Crea-ng an R2RML file that annotates a rela-onal database with exis-ng
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.

Example
@prefix rr: <h>p://www.w3.org/ns/r2rml#> .
@prefix foaf: <h>p://xmlns.com/foaf/0.1/> .
@prefix dbpedia: <h>p://dbpedia.org/ontology/> .

<#CityTriplesMap>
a rr:TriplesMap ;

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

rr:subjectMap [
rr:template "h>p://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

Example
@prefix rr: <h>p://www.w3.org/ns/r2rml#> .
@prefix foaf: <h>p://xmlns.com/foaf/0.1/> .
@prefix dbpedia: <h>p://dbpedia.org/ontology/> .

<#CityTriplesMap>
a rr:TriplesMap ;

rr:logicalTable [
rr:sqlQuery """SELECT ID, Name FROM City WHERE 1""" ;
] ;

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

rr:objectMap [ rr:column "Name" ] ;
] ;
.
City
ID Name
1 Dublin
2 Ghent

Example
<#PersonTriplesMap>
a rr:TriplesMap;
rr:logicalTable [ rr:tableName "Person" ] ;
rr:subjectMap [
rr:template "h>p://foo.example/Person/{ID}" ;
rr:class foaf:Person ;
];
rr:objectMap [ rr:column "First" ] ;
];
rr:predicate foaf:based_near ;
rr:objectMap [
rr:parentTriplesMap <#CityTriplesMap> ;
rr:joinCondi4on [
rr:child "CityID" ;
rr:parent "ID" ;
]
]
] ;
.
Rela-ng People to
Addresses.
City
ID Name
1 Dublin
2 Ghent
Person
2 Kevin NULL 2

In short…
Direct vs. Declara4ve
•  A Direct Mapping of Rela-onal Data to RDF - W3C
•  R2RML: RDB to RDF Mapping Language - W3C
Direct Mappings
•  Reflect the structure of the data source, including vocabulary
•  Mapping implemented in algorithm
•  Easy to generate, but not (necessary) meaningful
•  Use rules (SWRL or SPARQL CONSTRUCT) to use different vocabularies
Declara4ve Mappings
•  Relate your source dataset with a target RDF dataset with a vocabulary
•  Requires more effort, but more meaningful RDF

R2RML
(From h>ps://www.w3.org/TR/r2rml/)

R2RML: Running Example
Using our implementa-on of R2RML
h>ps://opengogs.adaptcentre.ie/debruync/r2rml

1.  The implementa-on is available in the folder
2.  Go inside example-1 and execute
$ java -jar ../r2rml.jar conﬁg.proper-es
Note: CSV ﬁles are turned into rela-onal tables.
Column names that contain no spaces are
CAPITALIZED.


R2RML: Running Example

R2RML Example 1
conﬁg.proper4es
CSVFiles = person.csv;city.csv
mappingFile = ./mapping.ttl
outputFile = ./output.ttl
format = TURTLE
Directory lis4ng of example-1
•  city.csv
•  conﬁg.proper-es
•  mapping.>l
•  output.>l [will be generated]
•  person.csv

R2RML
•  R2RML supports mapping values with constants, column values or column
values applied to a template.
•  If an Object Map does not refer to a column or has no language tag, then
term types default to IRIs unless you explicitly specify it to be a Literal.
…
rr:predicateObjectMap[
rr:predicateMap [ rr:constant foaf:name ] ;
rr:objectMap[
rr:template "{first} {last}" ;
rr:termType rr:Literal ;
]
]
…

R2RML
•  Term maps with a TermType of rr:Literal can have a language tag.
…
rr:predicateObjectMap[
rr:predicateMap [rr:constant rdfs:label];
rr:objectMap[
rr:column "{title}";
rr:language "en";
]
]
…
Why is the above TermType rr:Literal?
It uses a column.

R2RML: Mul-ple Languages
•  What if you have a table with mul-ple languages and these need to be separated
by langue?
•  Assuming you have a discriminator, you can create a logical table with an SQL
query for each language. How?
•  Create one logical table and use a language column to create a single mapping for
all languages. Unfortunately, this is not part of the recommenda-on and support
depends on the implementa-on…
rr:objectMap [ 
rr:column "TITLE"; 
rrx:languageColumn "TITLE_LANG"; 
].

R2RML: Datatypes
Only on TermMaps that are of type rr:Literal and without a language.

rr:objectMap [  
rr:column "EMPNO" ;  
rr:datatype xsd:positiveInteger ; 
].

R2RML
•  R2RML provides a highly customizable language for mapping rela-onal
databases to triples.
•  Unlike direct mapping that reﬂects the database’s structure, the author of
the mapping decides on the structure and ontology.
–  Mapping projec-ons, e.g., a Triples Map
for π{gender}(Person)
–  Mapping selec-ons, e.g., a Triples Map
for σ{gender=‘F’}(Person) to create instances
of ex:Woman
–  …
Person
ID name gender
1 Christophe M
2 Kevin M
3 Aoife F

RDB2RDF
•  How to annotate the exis-ng database?
–  Translate data to RDF: generates a RDF dump for immediate
consump-on, but will be harder to maintain (A)
•  Loaded into triplestores such as Jena TDB (with Fuseki for a SPARQL
endpoint), or Virtuoso.
–  A mapping from the RDB to RDF, genera-ng SPARQL queries into
(intermediate) SQL queries, but will have longer query -mes (B)

Implementa-ons
•  Stardog -- h>p://stardog.com/
–  implements both R2RML and their own mapping language.
•  Virtuoso – h>p://virtuoso.openlinksw.com/
•  Oracle Spa-al and Graph –h>ps://www.oracle.com/database/spa-al/index.html
–  Support for both transforming rela-onal databases into RDF using R2RML as well
as crea-ng R2RML views using those mappings
•  OnTop -- h>p://ontop.inf.unibz.it/
–  Access rela-onal databases as virtual graphs using mappings
•  D2RQ -- h>p://d2rq.org/
–  Direct mapping and D2RQ mapping language. No complete support for R2RML yet.
–  Comes with a SPARQL Endpoint and Linked Data Frontend based on Pubby.

Other beneﬁts
R2RML is a vocabulary and mappings are stored as RDF. This means:
•  One can add provenance informa-on to the mapping, which can be
queried
•  One can query over mappings, which can facilitate knowledge discovery,
reuse, etc.
–  Give me all mappings that use these predicates, these namespaces, etc.
–  Give me all mappings created by a par-cular person
–  ...
•  Provides a basis for four star and even ﬁve star Linked Data if appropriate
choices (e.g., in terms of URIs and how they resolve) are taken

OTHER INITIATIVES
•  R2RML-F – extending R2RML with func-ons
•  RML – an R2RML superset for mul-ple datasources

Extending R2RML – with Func-ons
•  R2RML allows one to describe mappings and assumes the database to
conform to the Core SQL 2008 speciﬁca-on.
•  But what if the underlying technology does not support certain data
manipula-ons?
–  Underlying technology not expressive enough.
–  Procedural domain knowledge part of applica-on.
•  More complex data processing “pipelines” have a nega-ve impact on
transparency and traceability of the RDF dataset genera-on process.
•  Inclusion of func-ons in ECMAScript (JavaScript) in mapping languages,
provided tractability is not a problem.

R2RML-F – R2RML with func-ons
•  Namespace rrf: h>p://kdeg.scss.tcd.ie/ns/rrf#
•  FuncBons have a func-on name and body.
•  Func-ons are wri>en in ECMAScript.
<#Multiply>
rrf:functionName "multiply" ;
rrf:functionBody """
function multiply(var1, var2) {
return var1 * var2 ;
}
""" ;
.

R2RML-F – R2RML with func-ons
A “func-on valued” term map calls a funcBon and the parameters are
themselves term maps.
<#TriplesMap1>
rr:logicalTable [ rr:tableName "Employee"; ];
rr:subjectMap [ rr:template "http://org.com/employee/{ID}"; ] ;
rr:predicate ex:salary ;
rr:objectMap [
rr:datatype xsd:double ;
rrf:functionCall [
rrf:function <#Multiply> ;
rrf:parameterBindings (
[ rr:constant "12"^^xsd:integer ]
[ rr:column "monthly_salary" ]
) ;
] ;
] ;
] ;
.
Parameter bindings as
an RDF Collec4on.
Parameter bindings can
be empty.
Term Maps as parameters.

R2RML-F
•  Introduc-on of a FuncBon Valued Term Map that allow for user deﬁned
func-ons at the cost of tractability.
•  h>ps://opengogs.adaptcentre.ie/debruync/r2rml
•  C. Debruyne and D. O'Sullivan. R2RML-F: Towards Sharing and Execu-ng
Domain Logic in R2RML Mappings. In Proceedings of the Workshop on
Linked Data on the Web, LDOW 2016, co-located with the 25th
Interna-onal World Wide Web Conference (WWW 2016), Montreal,
Canada, April 12th, 2016, 2016.

RML
•  Developed by iMinds (now imec) at Ugent
–  h>p://semweb.mmlab.be/rml/spec.html
–  h>ps://github.com/mmlab/RMLProcessor
•  An extension of R2RML to support XML, HTML, CSV, JSON…
–  Basically a superset of R2RML.
•  See also: A. Dimou, M. Vander Sande, P. Colpaert, R. Verborgh, E.
Mannens, R. Van De Walle: “RML: A Generic Language for Integrated RDF
Mappings of Heterogeneous Data”, in Proceedings of the 7th Workshop
on Linked Data on the Web, WWW14, 2014 Seoul, Korea.

TARQL
•  SPARQL for Tables
h>ps://tarql.github.io/
•  Use SPARQL CONSTRUCT
queries to transform data in
CSV/TSV into RDF
•  Construct queries can be
regarded as rules declaring how
to transform data
•  It adopts a W3C standard
(SPARQL), through relies on
some bespoke data
transforma-on func-ons (e.g.,
string explosion)
•  But not captured as RDF
CONSTRUCT {
?URI a ex:Organization;
ex:name ?NameWithLang;
ex:CIK ?CIK;
ex:LEI ?LEI;
ex:ticker ?Stock_ticker;
}
FROM <file:companies.csv>
WHERE {
BIND(URI(CONCAT('org/', ?Stock_ticker)) AS ?URI)
BIND(STRLANG(?Name, "en") AS ?NameWithLang)
}
Example from h>ps://tarql.github.io/

References
•  Satya S. Sahoo et al. A Survey of Current Approaches for Mapping of Rela-onal Databases to RDF.
W3C RDB2RDF XG Report, W3C, 2009.
h>p://www.w3.org/2005/Incubator/rdb2rdf/RDB2RDF_SurveyReport.pdf
•  Berners-Lee, T. Rela-onal Databases on the Seman-c Web, 1998, via
h>p://www.w3.org/DesignIssues/RDB-RDF.html
•  R. Cyganiak, C. Bizer, J. Garbers, O. Maresch, and C. Becker. The D2RQ mapping language. h>p://
d2rq.org/d2rq-language, March 2012.

TUTORIAL 1 – WEATHER STATION DATA

Fingal County Council Weather Sta-ons
•  Example from the h>p://data.gov.ie/ portal
•  Create instances of geo:Feature and geo:Geometry from GeoSPARQL
•  Use RDFS for labels
•  Assume h>p://data.fingalcoco.ie/ as the base URI for the resources and
h>p://www.fingalcoco.ie/ont# as the namespace for predicates specific
for this dataset
Records in fccweathersta-onsp20110829-2221.csv
Name Weather Reading Agency LAT LONG
M50 Blanchardstown h>p://… Na-onal Roads Authority 53.37046603 -6.380851447
M50 Dublin Airport h>p://… Na-onal Roads Authority 53.40964111 -6.227597428
Dublin Airport h>p://… Met Éireann 53.42150608 -6.29784754

Crea-ng a conﬁg.proper-es ﬁle
CSVFiles = weatherstations.csv
mappingFile = ./mapping.ttl
outputFile = ./output.ttl
format = TURTLE
We will later see extend
the example with named
graphs and change the
format accordingly

Tutorial 1
•  Add the following namespaces
–  rr: h>p://www.w3.org/ns/r2rml#
–  fcc: h>p://www.ﬁngalcoco.ie/ont#
–  geo: h>p://www.opengis.net/ont/geosparql#
•  Create a Triples Map for weather sta-ons, and make the subjects and
instance of geo:Feature and fcc:WeatherSta-on

Tutorial 1
@prefix rr: <http://www.w3.org/ns/r2rml#> .
@prefix fcc: <http://www.fingalcoco.ie/ont#> .
@prefix geo: <http://www.opengis.net/ont/geosparql#> .
<#WeatherStation>
a rr:TriplesMap ;
rr:logicalTable [ rr:tableName "WEATHERSTATIONS" ] ;
rr:subjectMap [
rr:template "http://data.fingalcoco.ie/ws/{NAME}" ;
rr:class geo:Feature ;
rr:class fcc:WeatherStation ;
] ;
.
We generate URIs with the name. Discuss.

Tutorial 1

Tutorial 1
•  Give the features both a default and an English label (use rdfs:label)
•  Provide informa-on on where the weather readings can be found (use
both fcc:withWeatherReading and rdfs:seeAlso)
–  Make sure you add the rdfs namespace
•  Relate to the agencies with fcc:withAgency

Tutorial 1
rr:predicate rdfs:label ;
rr:objectMap [
rr:column "NAME" ;
rr:language "en" ;
] ;
];
rr:objectMap [
rr:column "NAME" ;
] ;
]
rr:objectMap [
rr:column "NAME" ;
rr:language "en" ;
] ;
rr:objectMap [
rr:column "NAME" ;
] ;
]
More terse…

Tutorial 1
rr:predicate rdfs:seeAlso ;
rr:predicate fcc:withWeatherReading ;
rr:objectMap [
rr:column "WEATHER_READING" ;
] ;
] ;
rr:predicate rdfs:seeAlso ;
rr:predicate fcc:withWeatherReading ;
rr:objectMap [
rr:column "WEATHER_READING" ;
rr:termType rr:IRI ;
] ;
] ;
Or... Given that these are URIs

Tutorial 1

Tutorial 1
•  Create instances of geo:Geometry for the loca-on of these weather sta-ons.
We will use Well Known Text (WKT) to create points.
–  h>ps://en.wikipedia.org/wiki/GeoSPARQL (for an example with polygons)
–  h>ps://en.wikipedia.org/wiki/Well-known_text (for examples of points)
–  Hint: ﬁrst longitude, then la-tude. You can test whether points “make
sense” with Wicket. Just plot the point without the data type:
h>ps://arthur-e.github.io/Wicket/sandbox-gmaps3.html
•  Some considera-ons:
–  How could the URIs look like? Discuss.
–  Could we use blank nodes? Discuss.

Tutorial 1

Tutorial 1
<#Geometries>
a rr:TriplesMap ;
rr:logicalTable [ rr:tableName "WEATHERSTATIONS" ] ;
rr:subjectMap [
rr:template "http://data.fingalcoco.ie/geom/{LONG}/{LAT}" ;
rr:class geo:Geometry ;
] ;
rr:predicate geo:asWKT ;
rr:objectMap [
rr:template "POINT({LONG} {LAT})" ;
rr:datatype geo:wktLiteral ;
] ;
] ;
.

Tutorial 1

Tutorial 1
rr:subjectMap [
rr:template "http://data.fingalcoco.ie/geom/{LONG}/{LAT}" ;
rr:termType rr:BlankNode ;
] ;
# OR...
rr:subjectMap [
rr:template "{LONG}/{LAT}" ;
rr:termType rr:BlankNode ;
] ;
•  Values (template, column or constant) are
needed to diﬀeren-ate between diﬀerent
blank nodes.
•  Template is not used to create a IRI, hence
we can eliminate some redundant
informa-on

A note on blank nodes…
•  Blank nodes are resources without an iden-fier. They are generally to be
avoided, unless you know what you are doing.
•  Blank nodes are given a blank node iden-fier by triplestores and SPARQL
engines. They are not guaranteed to be uniquely iden-fying! In fact, the
specifica-ons even state that one 1) should not rely on consistent blank
node iden-fiers, and 2) the same blank node iden-fier in different
(named) graphs do not refer to the same thing.

Blank nodes at the OSi
data.geohive.ie is an ongoing collabora-on
between ADAPT and the Ordnance Survey
Ireland to publish OSi’s authorita-ve
geospa-al informa-on as Linked Data.

Star-ng from publicly available boundary
data, suppor-ng two use cases: provision
of diﬀerent geometries for features, and
provenance and evolu-on of features and
their geometries

Blank nodes at the OSi
@prefix geo: <http://www.opengis.net/ont/geosparql#> .
@prefix osi: <http://ontologies.geohive.ie/osi#> .
<http://data.geohive.ie/resource/county/2AE19629144F13A3E055000000000001>
a osi:County ;
a geo:Feature ;
rdfs:label "Baile Átha Cliath"@ga ;
rdfs:label "DUBLIN"@en ;
rdfs:label "DUBLIN" ;
geo:hasGeometry [
a geo:Geometry ;
geo:asWKT "MULTIPOLYGON (((-6.1 53.4, ..."^^geo:wktLiteral ;
] ;
.
04/11/17 71
Encourage one to link to features
rather than their geometries, which
are “merely” an a>ribute.
christophe.debruyne@adaptcentre.ie

Tutorial 1
•  Relate the geometries with their feature (hint parent triples map)
–  Who’s the parent, who’s the child? Or what is the direc-on of this
rela-on?

Tutorial 1
rr:predicate geo:hasGeometry ;
rr:objectMap [
rr:parentTriplesMap <#Geometries>;
rr:joinCondition [
rr:child "NAME" ;
rr:parent "NAME" ;
] ;
] ;
] ;

Tutorial 1 – (Named) Graphs
•  Named graphs are a convenient way to organize informa-on.
•  Unless otherwise speciﬁed, triples are stored in the default (or unnamed)
graph, which is also referred to with rr:defaultGraph.
•  Turtle does not support named graphs, but n-quads, for instance, does

TRIPLESTORE
Named graphs in data.geohive.ie
Prime2 Database
Ontologies
R2RML
Mapping
R2RML
Processor
Graphs Use Case 1
default •  Types
•  Labels
•  Links
•  100m resolu-on
50 meters •  50m resolu-on
20 meters •  20m resolu-on
•  Default geometry
links •  With LOD cloud
Graphs Use Case 2
default •  Ac-vi-es [PROV-O]
•  En--es [PROV-O]
•  History of
100m resolu-on
50 meters •  History of
50m resolu-on
20 meters •  History of
20m resolu-on

•  Change config.proper-es to generate an RDF dataset
–  format = NQUADS
–  outputFile = ./output.nq
•  Using h>p://www.w3.org/2003/01/geo/wgs84_pos#
–  Give the features a geo2:lat and a geo2:long in a different graph, you
may use h>p://data.fingalcoco.ie/graph/geo

# TO BE ADDED TO THE WEATHER STATION TRIPLES MAP
rr:graph <http://data.fingalcoco.ie/graph/geo> ;
rr:predicate geo2:lat ;
rr:objectMap [ rr:column "LAT" ; ] ;
] ;
rr:graph <http://data.fingalcoco.ie/graph/geo> ;
rr:predicate geo2:long ;
rr:objectMap [ rr:column "LONG" ; ] ;
] ;

Uplift – Generating RDF datasets from non-RDF data with R2RML

Recommandé

Recommandé

Contenu connexe

Tendances

Tendances (20)

Similaire à Uplift – Generating RDF datasets from non-RDF data with R2RML

Similaire à Uplift – Generating RDF datasets from non-RDF data with R2RML (20)

Plus de Christophe Debruyne

Plus de Christophe Debruyne (20)

Dernier

Dernier (20)

Uplift – Generating RDF datasets from non-RDF data with R2RML