LOD Laundromat
for Data Quality

September 12th, 2016

Wouter Beek (w.g.j.beek@vu.nl)

Metcalfe's Law

The value of a network is proportional to the square of the number of connected nodes

So... how many connected nodes does the SW have?

Data growth is exponential

SW growth is linear

Part I

What is the problem?

After 15 years most data cannot be automatically:

  • found
  • read
  • queried
  • reasoned over

Many PhD students' worse nightmare…

Problem 1: Most data cannot be found

SotA findability comparable to 1995 Yahoo! index

Problem 2: Most data cannot be read

E.g., Freebase <10% syntactically correct.

Current approaches are inherently slow: standards, guidelines, best practices, tools, education.

This takes decades!

Why is data dirty?

  • Character encoding issues
  • Socket errors
  • Protocol errors
  • Corrupted archives
  • Authentication problems
  • Syntax errors
  • Wrong metadata
  • Lexical form ↛ value
  • Non-canonical lexical form
  • Logically inconsistent

Problem 3: Most data cannot be queried

Problem 4: Most data cannot be reasoned over

  • Web-scale reasoning is only performed in the lab
  • Federation does not scale to thousands of endpoints
  • Real-world reasoning immediately goes ex falso quodlibet

Part II

How to solve this?


Beek & Rietveld & Bazoobandi & Wielemaker & Schlobach “LOD laundromat: A Uniform Way of Publishing Other People’s Dirty Data” ISWC, 2014.

LOD Laundromat uses the ClioPatria triple store, written in SWI-Prolog



Wielemaker & Beek & Hildebrand & Van Ossenbruggen, ‘ClioPatria: A SWI-Prolog Infrastructure for the Semantic Web’ in Semantic Web Journal, 2016.

How to query >30B statements (1/2)

How to query >30B statements (2/2)

Rietveld & Verborgh & Beek & Vander Sande & Schlobach, “Linked Data-as-a-Service: The Semantic Web Redeployed” ESWC 2015.

Part III: Why bother?

How generalizable is SW research?

ISWC 2014 Research Track:

  • 17 datasets overall; avg. 2 per per paper (avg. 2)
  • Data was cleaned locally and deleted afterwards

Reproducing “Linked Data Best Practices” (Schmachtenberg 2014)

Original LOD Lab
Prefix #datasets %datasets Prefix #documents %documents
rdf 996 98.22% rdf 639,575 98.40%
rdfs 736 72.58% time 443,222 68.19%
foaf 701 69.13% cube 155,460 23.92%
dcterm 568 56.01% sdmxdim 154,940 23.84%
owl 370 36.49% worldbank 147,362 22.67%

L. Rietveld & W. Beek & S. Schlobach, “LOD Lab: Experiments at LOD Scale”, International Semantic Web Conference, 2015 (Best Paper Award).

Calculate a metric over all WoD documents:

              frank documents --downloadUri |


Beek & Rietveld, ‘Frank: The LOD Cloud at your Fingertips’ in ESWC Developers Workshop, 2015.

Large-scale Data Quality Improvement (1/2): Datatypes

Large-scale Data Quality Improvement (2/2): Language tags

Beek & Ilievski & Debattista & Sclobach & Wielemaker, ‘Literally Better: Analyzing and Improving the Quality of Literals’ under submission.

Evaluation results for ±600,000 datasets

De Rooij & Beek & Bloem & Schlobach & Van Harmelen, ‘Are Names Meaningful? Quantifying Social Meaning on the Semantic Web’ in ISWC, 2016.

Semantic Search Engine

Ilievski & Beek & Van Erp & Rietveld & Schlobach, ‘LOTUS: Adaptive Text Search for Big Linked Data’, ESWC 2016.


Part IV: Is semantic data quality
what we think it is?

owl:sameAs has 2 meanings

Formal meaning

$$a = b \,\longleftrightarrow\, (\forall P)(Pa = Pb)$$

Social meaning

“Include links to other URIs, to discover more things.”

Contextual semantics for owl:sameAs

Beek & Schlobach & Van Harmelen, ‘A Contextualised Semantics for owl:sameAs’ in International Semantic Web Conference, p. 405--419, 2016.

Thank you!

Mail: w.g.j.beek@vu.nl

WWW: wouterbeek.com

Triply: triply.cc