Weaving the Pedantic Web (LD
Download as PPT, PDF9 likes2,275 views
This document summarizes a talk on common errors found in Linked Data. It discusses several types of errors discovered through analyzing over 150,000 RDF documents, including HTTP-level issues like URIs that don't return RDF descriptions, inaccurate content-type reporting, and duplicate content served at different URIs. It also describes reasoning issues such as undefined classes and properties, non-unique values for inverse-functional properties, malformed datatypes, and instances of disjoint classes. The document provides solutions like application workarounds, publishing validators, and the Pedantic Web Group for improving Linked Data quality.
Weaving the Pedantic Web (LD
- 1. Digital Enterprise Research Institute www.deri.ie Weaving the Pedantic Web LDOW 2010 Aidan Hogan, Andreas Harth, Alexandre Passant, Stefan Decker, Axel Polleres 0:39:00 Copyright 2009 Digital Enterprise Research Institute. All rights reserved. 1
- 2. Linked Data… Digital Enterprise Research Institute www.deri.ie 2
- 3. Purpose of talk: Application developers… how to not sink… Digital Enterprise Research Institute www.deri.ie 3
- 4. Purpose of talk: RDF Publishers… how to avoid common mistakes… Digital Enterprise Research Institute www.deri.ie 4
- 5. Talking about errors in Linked Data… Digital Enterprise Research Institute www.deri.ie We’ll try not to ruin the party …statistics based on crawl: April 2009 5k domain limit 150k URIS, 55k RDF docs 12.5m triples (quads) Mentioning 1.6m URIs 5,850 classes/9,507 props Accept: application/rdf+xml …okay… so no RDFa Statistics are *illustrative* not exhaustive! 5
- 6. Digital Enterprise Research Institute www.deri.ie Chapter 1: HTTP-level issues… …a good RDF description these days is hard to find 6
- 7. Waldo URIs: URIs with no dereferencable RDF Digital Enterprise Research Institute www.deri.ie Not a crawler’s idea of fun… 7
- 8. Hmm not *so* many… Digital Enterprise Research Institute www.deri.ie 5.3% of HTTP URIs return 40x/50x Excluding redirects… 92.8% return 200 OK In return, only 45.4% of 200 Okay return report application/rdf+xml 34.8% return HTML… probably just HTML docs… okay… maybe a *few* contain RDFa 8
- 9. Lies… Damned Lies… & Content-Type Reporting Digital Enterprise Research Institute www.deri.ie “Trust me, it’s RDF/XML” 9
- 10. Okay… So he’s actually pretty honest Digital Enterprise Research Institute www.deri.ie 16.9% of valid RDF/XML documents returned with an invalid/more generic Content-type: text/xml (9.5%) application/xml (5.9%) text/plain (1%) text/html (0.4%) Of those returning Content-type:application/rdf+xml 98.8% were valid RDF/XML 10
- 11. Same triples, different document Digital Enterprise Research Institute www.deri.ie I wish they’d used a redirect… 11
- 12. E.g., the Miracle at Calais: turning 1,778 triples into ~∞ quads Digital Enterprise Research Institute www.deri.ie http://d.opencalais.com/1/type/em/r/SameTriplesDifferentDocument (apologies to OpenCalais guys – it’s just a convenient example) 12
- 13. Digital Enterprise Research Institute www.deri.ie Chapter 2: Reasoning issues… …or, how I learned to start worrying and stop loving OWL 13
- 14. Undefined classes and properties… Digital Enterprise Research Institute www.deri.ie It looks important, but I’m afraid I don’t fully follow 14
- 15. Quite common… Digital Enterprise Research Institute www.deri.ie 14.3% of triples use undeclared property 8.1% of triples use undeclared class Three cases: Case 1: Namespace has no vocabulary/ is not deferencable (e.g., rss:item) Case 2: Term invented in related namespace (e.g., foaf:tagLine invented by LiveJournal) Case 3: Term is misspelt version of term defined in namespace (e.g., foaf:image vs. foaf:img) 15
- 16. Not-so-unique values for Inverse-Functional Properties Digital Enterprise Research Institute www.deri.ie Despite what you claim, not all of you can *actually be* Spartacus 16
- 17. Spartacus relived… Digital Enterprise Research Institute www.deri.ie 08445a31a78661b5c746feff39a9db6e4e2cc5cf sha1-sum of „mailto:‟ common value for foaf:mbox_sha1sum An inverse-functional (uniquely identifying) property!!! Any person who shares the same value will be considered the same *I’m Spartacus!* …and so’s my wife 17
- 18. …unattended, can be pretty serious… Digital Enterprise Research Institute www.deri.ie foaf:mbox_sha1sum a owl:InverseFunctionalProperty . ?x foaf:mbox_sha1sum 08445a31a78661b5c746feff39a9db6e4e2cc5cf . OWL 2 RL rule prp-ifp: ?p a owl:InverseFunctionalProperty . ?x1 ?p ?z . ?x2 ?p ?z . ⇒ ?x1 owl:sameAs ?x2 . 106 ?x1/?x2bindings in body 1012 inferred pair-wise and reflexive owl:sameAs statements …or in simpler terms: pow! 18
- 19. Malformed/incompatible datatypes Digital Enterprise Research Institute www.deri.ie As he would undoubtedly be able to tell you, “true” is not a valid xsd:int 19
- 20. Not *too* bad… Digital Enterprise Research Institute www.deri.ie 4.7% of typed literals were “ill-typed” (lexically invalid)… mostly xsd:dateTimes (26.4% of all date-time literals were invalid; e.g., omitted the seconds field) Also, literals are sometimes incompatible with the datatype-range of a property: E.g., 21.8% of ical:description triples used language tags incompatible with the defined range of xsd:string E.g., 100% of sl:creationDate triples use plain literal values incompatible with defined range of xsd:date 20
- 21. Mystical beings… Members of disjoint classes Digital Enterprise Research Institute www.deri.ie Despite what FOAF says, it seems that Persons can also be Documents 21
- 22. Again, not *too* bad… Digital Enterprise Research Institute www.deri.ie 1,329 members of disjoint classes found Generally caused by naïve URI naming: Use of information resource URIs to name entities (particularly foaf:Persons) E.g., <me> foaf:knows <jim/foaf.rdf> . 22
- 23. Ontology hijacking… Digital Enterprise Research Institute www.deri.ie Anybody can say anything, anywhere, and unfortunately for everyone else, have a good chance of being taken seriously 23
- 24. Redefining Everything… …and home in time for tea Digital Enterprise Research Institute www.deri.ie From http://www.eiao.net/rdf/1.0 <owl:Property rdf:about="http://www.w3.org/1999/02/22-rdf-syntax-ns#type"> <rdfs:label xml:lang="en">type</rdfs:label> <rdfs:comment xml:lang="en">Type of resource</rdfs:comment> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#testRun"/> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#pageSurvey"/> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#siteSurvey"/> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#scenario"/> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#rangeLocation"/> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#startPointer"/> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#endPointer"/> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#header"/> <rdfs:domain rdf:resource="http://www.eiao.net/rdf/1.0#runs"/> </owl:Property> Ontology hijacking!! (apologies to EIAO guys – it’s just a convenient example) 24
- 25. Solutions? Digital Enterprise Research Institute www.deri.ie 25
- 26. Application side: workarounds Digital Enterprise Research Institute www.deri.ie All presented issues have a suitable antidote, once you know about them See paper for discussion… 26
- 27. Publishing side: Validators! Digital Enterprise Research Institute www.deri.ie Syntax errors quite rare, partly due to popularity of W3C RDF/XML syntax validator Need an all-in-one validation service Should not only validate strict errors, but give feedback on suspected issues We offer a prototypical service at: http://swse.deri.org/RDFAlerts/ 27
- 28. Publishing side: Pedantic Web Group Digital Enterprise Research Institute www.deri.ie Get the community to contact publishers about errors/issues as they arise Get involved: http://pedantic-web.org/ 137 members! Acknowledgements to: Aidan Hogan, Alex Passant, Me, Antoine Zimmermann, Axel Polleres, Michael Hausenblas, Richard Cyganiak, Stéphane Corlosquet 28