High Quality Linked Data Generation - swib.orgswib.org/swib17/slides/dimou_high-quality.pdf · 2...

High Quality Linked Data Generation

Dr Anastasia DimouPost-Doc researcher

imec.be - [email protected]@natadimou


from Heterogeneous Data

best practices for semantically annotating & connecting structured data on the (Semantic) Web



Linked Data is (still) not intuitively available on the Web



Linked Data is derived from heterogeneous data sources, structures & formats



Linked Data needs to be consistent

Human agents do not want to put in effort to provide Linked Data

until there are software agents that use it

Why is Linked Data Generation Such a difficult task?

data owner

custom implementation

Linked Data

my data

data owner

Linked Data

Linked Data

Linked Data

Linked Data

XMLCSV JSONDB

format-specific implementation

data owner

Linked Data

Linked Data

Linked Data

Linked Data

XMLCSV JSONDB

Linked Data

facilitate Linked Data generation byreducing data & semantic heterogeneity

& increasing Linked Data quality

data owner

Linked Data

Linked Data

Linked Data

Linked Data

DB CSV XML JSON

multiple data sources

data owner

DB CSV XML JSON

uniform execution

Linked Data

data owner

DB CSV XML JSON

?

uniform declaration

Linked Data

declarationexecutionassessment

data owner

DB CSV XML JSON

Linked Data

?

data owner

DB CSV XML JSON

R2RML

Linked Data

Linked Data

data owner

DB CSV XML JSON

R2RML

Linked Data

Linked Data

Extending R2RML to a source-independent mapping language for RDFA. Dimou et al.

data owner

DB CSV XML JSON

R2RML

Linked Data

http://RML.ioRML: A Generic Language for Integrated RDF Mappings

of Heterogeneous Data. A. Dimou et al.

bibo:Document

bibo:presentedAt<http://ex.com/paper/{id}>

id title venue

1 Assessing & Refining Mappings to RDF to improve Dataset Quality ISWC 2015

2 RMLEditor : a graph-based editor for Linked Data mappings ESWC 2016

3 An ontology to semantically declare and describe functions ESWC 2016

4 Modeling, Generating, Publishing Knowledge as Linked Data EKAW 2017

5 Semi-automatic example-driven linked data mapping creation ISWC 2017

<http://ex.com/conf/{venue}>

bibo:Event

bibo:Document

bibo:presentedAt<http://ex.com/paper/1>

id title venue






<http://ex.com/conf/ISWC2015>

bibo:Event

Who defines those rules?

<datafield tag="001#"><subfield code="0">2,192</subfield></datafield><datafield tag="001A"><subfield code="0">1961:22-06-88</subfield></datafield><datafield tag="001U"><subfield code="0">utf8</subfield></datafield><datafield tag="002C"><subfield code="a">tekst</subfield><subfield code="b">txt</subfield><subfield code="2">rdacontent/dut</subfield></datafield>

<datafield tag="002D"><subfield code="a">zonder medium</subfield><subfield code="b">n</subfield><subfield code="2">rdamedia/dut</subfield></datafield><datafield tag="003#"><subfield code="0">047268573</subfield></datafield><datafield tag="003O"><subfield code="a">OCoLC</subfield><subfield code="0">64827916</subfield></datafield>

Sem Web dev

rr:subjectMap [ rr:template "http://data.kb.nl/ppn/{repox:metadata/record/datafield[@tag=\"003@\"]/subfield[@code=\"0\"]}"; rr:class ex:Record ];

rr:predicateObjectMap [ rr:predicate rdfs:label; rr:objectMap [ rml:reference "repox:metadata/record/datafield[@tag=\"021A\"]/subfield[@code=\"a\"]"; rr:language "nl" ] ];

rr:predicateObjectMap [ rr:predicate dcterms:type; rr:objectMap [ rml:reference "repox:metadata/record/datafield[@tag=\"002@\"]/subfield[@code=\"0\"]" ] ];

rr:predicateObjectMap [ rr:predicate dcterms:publisher; rr:objectMap [ rml:reference "repox:metadata/record/datafield[@tag=\"033A\"]/subfield[@code=\"n\"]" ] ];

rr:predicateObjectMap [ rr:predicate rdfs:seeAlso;rr:objectMap [ rr:template "http://www.worldcat.org/oclc/{metadata/record/datafield[@tag=\"003O\"]/subfield[@code=\"0\"]}"] ].

data owner

RMLEditor: A Graph-Based Mapping Editor for Linked Data MappingsP. Heyvaert, et al.


declare mapping rules

RMLLanguage

data owner

DB CSV XML JSON

RML

Linked Data

data owner

DB CSV XML JSON

RMLRML handler

RML Processor

Linked Data

Machine-interpretable dataset & service descriptions for heterogeneous data access and retrieval. A. Dimou et al.

data owner

DB CSV XML JSON

RMLRML handler

RML Processor

Linked Data

data retrievalhandler

Machine-interpretable dataset & service descriptions for heterogeneous data access and retrieval. A. Dimou et al.

data owner

table CSV XML JSON

RMLRML handler

RML Processor

source desc

Linked Data


“15th October 2017”

“2017-10-15”^^xsd:date

data owner

table CSV XML JSON

RMLRML handler

RML Processor

source desc

Linked Data


data owner

table CSV XML JSON

RMLRML handler

RML Processor

source desc

Linked Data


functionhandler

An Ontology to Semantically Declare & Describe FunctionsB.De Meester, A. Dimou, R. Verborgh, E. Mannens & R. Van De Walle

data owner

table CSV XML JSON

RMLFnO

RML handler

RML Processor

source desc

Linked Data


functionhandler GREL

DBpediafunctions

Automated Metadata Generation for Linked Data Generation and Publishing Workflows. A. Dimou et al.

data owner

table CSV XML JSON

RMLFnO

RML handler

RML Processor

source desc

metadatahandler

Linked Data

meta data


functionhandler GREL

DBpediafunctions

data owner

table CSV XML JSON

RMLFnO

RML handler

RML Processor

source desc

metadatahandler

Linked Data

meta data

RML Mapper


functionhandler

RML Mapper: a tool for uniform Linked Data generation from heterogeneous data. A. Dimou et al.



RMLMapper

execute RML rulesRML rules

Linked Data

data

declare RML rules

RMLLanguage

bibo:Document


id title venue







bibo:Event

foaf:Person


id title venue







bibo:Proceedings

foaf:Person

bibo:presentedAt<http://ex.com/paper/1>

id title venue






<http://ex.com/conf/ISWC2015>

bibo:Proceedings

What happens then?

foaf:Person


id title venue







bibo:Proceedings

100 triples2 violations/triple200 violations!

foaf:Person


id title venue







bibo:Proceedings

1,000,000 triples2 violations/triple2,000,000 violations!

You think this doesn’t happen?!

Conference

Year 2014 2015

Solution 1.1 swc:OrganizedEvent swc:OrganizedEvent

Solution 1.2 swc:Event bibo:Conference

Solution 1.3 swrc:Event swrc:Event

Solution 1.4 swrc:Event

Challenges as enablers for high quality Linked Data: insights from the Semantic Publishing Challenge A. Dimou, et al.

Semantic Publishing

Challenge 2014 - 2016

statistics

Workshop

Year 2014 2015

Solution 1.1 bibo:Workshop bibo:Workshop

Solution 1.2 swc:Event bibo:Workshop

Solution 1.3 swrc:Event swrc:Workshop

Solution 1.4 swrc:Section


Semantic Publishing


statistics

Paper

Year 2014 2015

Solution 1.1 swrc:InProceedings foaf:Document

Solution 1.2 bibo:Article swrc:InProceedings

Solution 1.3 swrc:Publication swrc:Publication

Solution 1.4 swc:Paper


Semantic Publishing


statistics

Can we prevent violations?

RML Mapper

DQALinked Data

data

RML

violations

RML Mapper

violationsMDQA

Linked Data

data

RML

Assessing and Refining Mappings to RDF to Improve Dataset Quality. A. Dimou et al.

new RML

RML Mapper

Mapping Refinements

violationsMDQA

(optional)

Linked Data

data

RML

DBpediaUse Case

Sustainable Linked Data Generation: The case of DBpedia: W. Maroy et al.

Certain test cases require a complete Linked Data set

DBpedia Quality AssessmentLinked Data: 16hRML rules: 32s

new RML

RML Mapper

Mapping Refinements

violationsMDQA

(optional)

Linked Data

data

RML


RMLMapper

execute RML rules

RMLValidator

validate RML rules RML rules

validatedRML rules

Linked Data

data

declare RML rules

RMLLanguage


RMLMapper

execute RML rules

administrate Linked Data generation workflow RMLWorkbench

RMLValidator


validatedRML rules

Linked Data

data

declare RML rules

RMLLanguage

Human agents do not need to put in much effort to provide Linked Data

Human agents do not need to put in much effort to provide Linked Data

Intelligent software agents which function with Semantic Web technologies

will have enough Linked Data to work with


Dr Anastasia DimouPost-Doc researcher

imec.be - [email protected]@natadimou

High Quality Linked Data Generation - swib.orgswib.org/swib17/slides/dimou_high-quality.pdf · 2...

Documents

Transcript of High Quality Linked Data Generation - swib.orgswib.org/swib17/slides/dimou_high-quality.pdf · 2...