This paper presents an example of practical application of Linked Statistics to the problem of checking facts in news articles. We consider a use case of verifying a news article that exploits statistical facts from the Italian National Institute of Statistics (ISTAT). To realise the use case, we publish a subset of ISTAT as Linked Data1, thus, contributing to the presence of statistics on the Web of Data. We discuss how links from the news article to LinkedSTAT can enable creation of automated tools and services for fact-checking. We received a positive evaluation of our demo use case from ISTAT experts. With our work we hope to promote the value that could be obtained by publishing statistics as Linked Data by official statistical agencies and organisations.
Seismic Method Estimate velocity from seismic data.pptx
News Fact-checking: One Practical Application of Linked Statistics
1. News Fact-checking:
One Practical Application of Linked Statistics
3.3.
LinkedSTAT http://linkedstat.spaziodati.eu
ISTAT SDMX SOAP Web Service
http://sdmx.istat.it/WS_T/NsiStdV20Service.asmx
SDMX-ML SDMX-to-RDF XSL transformations
https://github.com/csarven/linked-sdmx
Virtuoso
Quad Store
http://www.ladige.it/articoli/2012/07/17/poverta-trentino-resta-isola-felice
NEWS FACT-CHECKING is the process of verifying accuracy of facts in publications
Fact-checking is a tedious, time-, resource-consuming and error-prone process
* The original text is:
“Nel 2011, secondo i dati Istat, ...
la provincia di Trento (3,4%), la Lombardia (4,2%),
la Valle d'Aosta e il Veneto (4,3%)
presentano i valori più bassi
dell'incidenza di povertà [relativa].”
“In 2011, according to Istat, ...
the province of Trento (3.4%),
Lombardy (4.2%), Valle d'Aosta and
Veneto (4.3%) have the lowest value
of the incidence of [relative] poverty.” *
RDF/XML
RDF Data Cube Vocabulary
PROV-O Ontology
SKOS and XKOS
SDMX-RDF, ...
Fact-checking: How to find a right set of dimension/value
pairs for a given fact to construct queries for it?
ISTAT: new possibilities to disseminate statistics and
facilitate data certification.
● DBpedia/Wikipedia: automatic updates of statistical data
“In 2011, according to Istat, ... the province of Trento
(3.4%) ... value of the incidence of [relative] poverty.”
Dimension Value
territory
linked-istat-property:REF_AREA
“Provincia Autonoma Trento”
http://linkedstat.spaziodati.eu/code/1.1/
CL_REFAREA/ITD2
reference time period
linked-istat-property:TIME_PERIOD
“2011”
<http://reference.data.gov.uk/id/year/2011>
statistical indicator
linked-istat-property:IND_TYPE
“incidenza di povertà relativa familiare”
<http://linkedstat.spaziodati.eu/code/1.1/
CL_AGGREG_FAMIGLIE/INCID_POVREL_
FAM>
http://linkedstat.spaziodati.eu/sparql
MANUAL FACT-CHECKING – review of the citations' content
● dedicated fact-checking departments
- only major infrequent periodicals can afford them (Der Spiegel, The Guardian, Esquire, Forbes);
no budget in small publishing organisations
- impractical for frequent publications
● nonprofit fact-checking organisations (FactCheck.org, PolitiFact.com)
● crowd-checking platforms (FactCheckEU.org)
Tatiana Tarasova tarasova@spaziodati.eu SpazioDati, Trento, Italy
“Poverty, Trentino remains a happy island” l'Adige.it, 17 Luglio 2012
What if the facts would be linked to the underlying data sources?
publishing ISTAT http://dati.istat.it/ as Linked Data
http://dati.istat.it/
All the queries and scripts produced during the LinkedSTAT project are available at
https: //www.assembla.com/spaces/linked-istat/
Fact-checking with LinkedSTAT
SELECT DISTINCT ?dataset ?title ?structure
WHERE {
?dataset a qb:DataSet .
?dataset dcterms:title ?title .
FILTER(contains(str(?title), "Incidenza di povertà relativa"))
?dataset qb:structure ?structure .}
SELECT DISTINCT ?codeList ?p ?o
WHERE {
<http://linkedstat.spaziodati.eu/property/REF_AREA> qb:codeList ?
codeList .
?codeList ?p ?o}
SELECT DISTINCT ?obs ?value
WHERE {
?obs rdf:type qb:Observation .
?obs linked-istat-property:REF_AREA
<http://linkedstat.spaziodati.eu/code/1.2/CL_REFAREA/ITD2> .
?obs linked-istat-property:TIME_PERIOD
<http://reference.data.gov.uk/id/year/2011> .
?obs linked-istat-property:IND_TYPE
<http://linkedstat.spaziodati.eu/code/1.1/CL_AGGREG_FAMIGLIE/INCID_POVREL_FAM> .
?obs linked-istat-property:OBS_VALUE ?value .}
Step1: retrieve the structure of the relevant dataset
Step2: retrieve code lists that provide values
Step3: retrieve the value of the required observation
Future