1. SISAI STATISTICAL INFORMATION SYSTEMS ARCHITECTURE AND INTEGRATION
WORKING GROUP 3rdMEETING MAY 2013 ITEM 2.4 Integrating Statistical Data with Semantic Web

2. From SDMX to RDF Data Cube Vocabulary: Integrating Statistical Data
with Semantic Web
Monica Scannapieco
Italian National Institute of Statistics (Istat)
Joint work with:
Raffaella M. Aracri, Andrea Pagano,
Laura Tosco, Luca Valentino
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg

3. Open Data & Linked Open Data
Format that allows «usage, reuse and redistribution»
Linked Open
Standard models and formats intended for data integration on the Web
Current situation:
Overcoming pattern «formerly open - then linked»
Data straightly expressed as Linked Open Data (LOD)
Example 1 - US data.gov: PA open data portal in the PA. In progress migration to LOD
Example 2: DBpedia & Wikidata
DBpedia: data extraction from Wikipedia infoboxes
Wikidata: structured database, aiming to feed Wikipedia infoboxes 1
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 3

4. Linked Open Data and Semantic Interoperability
Semantic Web Stack
Knowledge
Semantic
Format and syntax
Linked Open Data
Data represented by means of RDF (Resource Description Framework) languages
Interconnected => Semantic Interoperability 2
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 4

5. Background: RDF, RDFS, OWL
Resource Description Framework (RDF)
Language for representing information about resources in the World Wide Web
W3C Recommendation 10 February 2004
RDF Schema (RDFS)
Language for RDF vocabulary sharing
10 February 2004 (as part of a wider revision of RDF) RDFS became a W3C Recommendation
OWL
Language for publishing and sharing ontologies on the World Wide Web.
OWL is developed as a vocabulary extension of RDF and is derived from the DAML+OIL Web Ontology Language
OWL W3C Recommendation 10 February 2004
OWL 2 W3C Recommendation 11 December 2012
OWL 2 adds new functionality with respect to OWL 1. Some of the new features are syntactic sugar (e.g., disjoint union of classes) while others offer new expressivity, including:
keys;
property chains;
richer datatypes, data ranges;
qualified cardinality restrictions;
asymmetric, reflexive, and disjoint properties; and
enhanced annotation capabilities 3
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 5

6. Background: SPARQL and RDF Data Cube Vocabulary
SPARQL (Sparql Protocol And RDF Query Language) is a language with a syntax similar to SQL for querying RDF data and a communication protocol based on HTTP
A SPARQL client can query a SPARQL endpoint with queries on a RDF graph
SPARQL allows “graph pattern matching” on RDF data
W3C Recommendation 15 January 2008
RDF Data Cube (RDF QB) is a
W3C Working Draft of the 12 March 2013
RDF QB is based on SDMX
Focused only on the publication on the web of multi-dimensional data
Built on the SDMX information model 4
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 6

7. RDF Basics Resource Description Framework (RDF)
Allows to represent data/metadata trough assertions, called triples
A triple: <subject> <property> <object>
A resource is uniquely identified by a URI
A puts in relationship other two
A property can also put in relationship a resource and a «literal», i.e., a pure symbolic expression, e.g., a number, a string
In such a way we have the RDF graph
Example:
<
<
< 5
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 7

8. How can we "integrate" the Istat data with data
Problem
ISTAT
Dati sul WEB
Reference Metadata
Enhanced SDMX
BB1
BB2
BBn
SDMX
Web Service Provider
SEP
Tablets/
smartphones
Structural Metadata
WEB GUI …
Istat
Information System
Building Blocks
Excel
Plug
-in
Metadata Management System
How can we "integrate" the Istat data with data
on the Web already interconnected with each other? 6
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg

9. Translator Solution 7 SDMX <structure:Concept id="REF_AREA">
<structure:Name xml:lang="en">Geographical reference area</structure:Name>
</structure:Concept> 7
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg

10. Transformation from SDMX to RDF-QB
Code list
Concept
dimension, attribute,
measure ?
DSD
SDMX
RDF QB
Data file
SDMX
Data set
RDF
transformation 8
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 10

11. Analysis of the Technological Environments
R Environment:
Package RSDMX: not complete and not actively maintained
Java Environment:
Apache JENA: framework for reading, processing and writing data in RDF, SPARQL queries are processable and more
Input SDMX not covered
MIMAS Project ( )
Transformation of data using XSLT 9
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 11

12. Technological Choice and Design
Usage of XSLT transformations
Execution engine of transformation: Saxon (Home Edition)
Supports XSLT 2.0, XQuery 1.0, XPath 2.0
Available in both Java and .NET
(Principal) Differences with MIMAS
Level of generalization: our translator is generalized, while MIMAS provides transformations ad-hoc to datasets
Generation of separate files for SDMX Data, SDMX DSDs, and Codelists 10
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 12

13. Example of Mapping SDMX DSD RDF QB 11
<structure:Concept id="REF_AREA">
<structure:Name xml:lang="en">Geographical reference area</structure:Name>
</structure:Concept>
RDF QB
<rdf:Description rdf:nodeID="REF_AREA">
      <qb:dimension rdf:resource="
      <dc:language>en</dc:language>
      <rdf:type rdf:resource="
      <rdf:type rdf:resource="
      <sdmx:codeList rdf:resource="
      <rdfs:range rdf:resource="
      <rdfs:label xml:lang="en">Geographical reference area</rdfs:label>
   </rdf:Description> 11
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 13

14. Example of XSL File: Definition of Transformation Rules
Reading from SDMX: Concept and Codelist
<xsl:for-each
<xsl:variable name="conceptRef"
<xsl:variable name="codeList"
<xsl:variable name="codeListName"
<xsl:element name="rdf:Description" >
<xsl:attribute name="rdf:nodeID" select="$conceptRef"/>
<xsl:element name="qb:dimension">
<xsl:attribute name="rdf:resource" select="concat($IstatRoot,'/code/',$codeListName)"/>
</xsl:element>
Writing in RDF QB: Concept and Codelist 12
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 14

15. Syntactic and Semantic Validation
Syntactic Validation: syntactic validation of RDF files creating RDF triples and their graph representation
Used the free software validator (
Verified also compliance with turtle (Eurostat format for DSD)
from RDF XML to turtle format (.ttl)
Software any23 ( )
Semantic Validation: data model is a «valid» RDF QB model
Used Openlink Virtuoso ( ) 13
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 15

16. Semantic Validation
SELECT
obs_id,
ref_area,
obs_value,
time_period,
territoryLabel
WHERE
typeofWaste=9
AND
Time= 14
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 16

17. Test Case and Performance
Data Input Size (KB)
Execution Time (min,sec)
Base
1.338
6m 7,9s
2.674
12m 32,86s
4.009
18m 42,248s
5.345
23m 49,127s
Optimized_1
2m 17,71s
4m 36,933s
7m 28,945s
9m 12,627s
Optimized_2
5,6s
7,87s
11,003s
13,968s
Optimized 1: mimimize DSD access
Optimized 2: in memory access of DSD representation 15 17

18. Test Case e Performance
Milliseconds (Log Scale) KB 16
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 18

19. Prototype development SDMX data format: Compact
Conclusions and Future Developments
Prototype development
SDMX data format: Compact
Possible extensions to other SDMX formats (i.e., Generic, Cross-Sectional)
Extension of the transformation rules of the constructs
Integration with the Istat Single Exit Point (SEP) 17
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 19

20. Complexity Query Data Model Discussion: RDF Publication Architectures
(+) Linking with other sources
(-) Redundancy wrt SDMX
(-) Deployment dedicated
SPARQL EndPoint with triple
(+) Easy to be published
(-) Not queryable
High
Complexity
Query
RDF in flat file
RDF via SDMX
Low
(+) Recovery of SDMX investment (SEP)
(-) Cross-format translation during the query phase
(-) Linking with other sources
Low
High
Data Model 18
Monica Scannapieco , 3rd SISAI Meeting, May 2013, Luxembourg 20