Download to read offline
Uploaded bySemantic Technology Institute International
STI Summit 2011 - Mlr-sm
The document discusses using usage analysis to improve ontology engineering. It describes analyzing query logs over datasets like DBpedia to identify frequently queried triples and patterns. This can reveal missing or inconsistent data and suggest new links between entities. The analysis helps increase data quality and acquire new knowledge that benefits both the dataset and Web of Data as a whole. While complete automation may not be needed, supporting usage analysis and endpoint access allows publishers to play a role in maintaining datasets and the Web of Data.
STI Summit 2011 - Mlr-sm
- 2. My Seman)cs of “Train” <owl:Thing rdf:about="#LevisTrain"> <rdf:type rdf:resource="#Train"/> <rdfs:label>Levi’s Train</rdfs:label> <madeOf rdf:resource="#Plas&c"/> </owl:Thing> <owl:Thing rdf:about="#LevisTrain"> <rdf:type rdf:resource="#Train"/> <rdfs:label>Levi’s Train</rdfs:label> <madeOf rdf:resource="#Wood"/> </owl:Thing>
- 3. A Usage-‐dependent Life Cycle Request to put • toy train away the “train” • toy train • made of plas)c • SELECT * WHERE ?t • made of wood a:madeOf a:Plas)c • SELECT * WHERE ?t b:madeOf b:Wood Nego)ate Enter the room understanding USAGE
- 4. Yet another… OTK … The NeOn Maintenance METHONTOLOGY Black Box DILIGENT Make it less a methodology but support the people to get their “Things” done!
- 5. Who is hurt by that? • rather small/simple ontologies – min. effort for OE – “under-‐engineered” • unknown user requirements
- 6. Hey “LOD people”, do you think that ontology engineering maaers? Usage-‐based ontology engineering
- 7. Survey covering approx. 25% of all cloud datasets • size • complexity • engineering methodology • … Publishers of 75% of the dataset do not feel 99% responsible for their data? Survey ran in October 2010
- 8. Concrete Example of Usage-‐based Approach digging in log files
- 9. Usage? Request to put away the “train” • SELECT * WHERE ?t a:madeOf a:Plas)c Yes*! But beyond? • SELECT * WHERE ?t b:madeOf b:Wood USAGE • What about the future of SPARQL endpoints on the WoD? * W.r.t. an architecture proposed by a famous “Web-‐Extremist”
- 10. You should have a query endpoint! Effort Distribution between Publisher and Consumer • You get Pays-As-You-Go something valuable Consumer generates/ data mines links en the data publisher,t data integration effort is out of i the publisher which helps mer and third parties. !"# Effort $%&'()) *(+( Distribution ,-+&.'(+"/- lisher s data as RDFyou to play 011/'+ your role erms from common vocabularies s and publishes mappings Publisher provides links Links as hints ties on the p WoD! pointing at y g your data mappings to the Web 567)"83&'98 011/'+ 23"'4 5('+: 011/'+ Christian Bizer: Pay-as-you-go Data Integration (21/9/2010) sumer ;/-86<&'98 o the rest 011/'+ ta mining techniques for
- 11. Usage Analysis • queries • paaerns • triples • primi)ves visualize heat maps zoom in and see details
- 12. Some Results (DBpedia Analysis) • ns:Band ns:instrument ?x inconsistent • ns:Band ns:genre ?y data • ns:Band ns:associatedBand ?z • ns:Band ns:knownFor ?x • ns:Band ns:na)onality ?y missing facts Complete analysis can be found at hap://page.mi.fu-‐berlin.de/mluczak/pub/visual-‐analysis-‐of-‐web-‐of-‐data-‐usage-‐dbpedia33/
- 13. Some Thoughts about Benefit • usage analysis helps to acquire new knowledge – links between data helps to increase • lightweight approach the quality of data on helps to bootstrap the Web linked data – external schema It is not necessary to automate everything if the result has enough (business) value in a problem domain anyway.
- 14. make it less a methodology • LOD vocabularies are specific ontologies • and need specific life cycle our data provide (query) access to y support T • endpoint and pcan your to mon the WoD usage analysis lay help role aintain them a • (and the mplicitly necessary to automate it is not i data) k • things ahat enable automa)on publisher this is t benefit for the dataset e and the Web of data as a whole A Hey “LOD people”, do you think that w dataset maintenance maaers? a y Markus Luczak-‐Rösch ([email protected]‐berlin.de) Freie Universität Berlin, Networked Informa)on Systems (www.ag-‐nbi.de)
- 15. Actual Addi)on • “15.500.000 people in Germany are not willing to use the internet” – emphasis on the ESWC discussion: bridging the gap (directly or indirectly) between these people and the internet/Web has a high poten)al to influence societal transforma)on (they are not going to use a browser or an iPhone and they do not care for seman)cs) Source: ARD-‐Morgenmagazin, 08-‐07-‐2011