The Empirical Turn in Knowledge Representation
Download as PPTX, PDF10 likes1,478 views
The document discusses advancements in knowledge representation (KR) and the observation of large-scale knowledge bases, highlighting various KR methods such as RDF, first-order logic, and Bayesian networks. It outlines an experiment that managed to analyze around 100 billion facts and how this analysis can be utilized for better querying tools and understanding patterns of knowledge. The overall aim is to leverage these insights to enhance applications in artificial intelligence.
![x T
[<x> IsOfType <T>]
different
owners & locations
< analgesic >
BTW: How did it get so big?
On the Web,
anybody can say anything about anything](https://image.slidesharecdn.com/empiricalkr-171227122316/85/The-Empirical-Turn-in-Knowledge-Representation-17-320.jpg)
![Identity clusters
LOD-a-lot File
http: //lod-a-lot.lod.labs.vu.nl
[Fernández 2017]
558 millions owl:sameAs (309 millions distinct terms)
≈ 4 hours
1. Extracting all owl:sameAs statements on the LOD
HDT File
(4.5 GB)](https://image.slidesharecdn.com/empiricalkr-171227122316/85/The-Empirical-Turn-in-Knowledge-Representation-28-320.jpg)
The Empirical Turn in Knowledge Representation
- 1. Creative Commons CC BY 3.0: allowed to share & remix (also commercial) but must attribute Frank van Harmelen The empirical turn in Knowledge Representation Contributions from many people in the KR&R group over many years. And thanks to NWO for a 750k€ TOP grant for this
- 2. KR in the pre-empirical era
- 3. Handbook of Knowledge Representation (1000 pages, ToC alone is 14 pages) • propositional logic & satisfiability solvers • first order logic & resolution • description logic • constraint (logic) programming • nonmonotonic reasoning • belief revision • qualitative reasoning • model-based diagnosis • bayesian networks • temporal logic • spatial reasoning • epistemic logic • deontic logic • situation calculus • default logic • event calculus • ……
- 4. KR metrics in the pre-empirical era KR = logic • Show small examples • Prove properties (expressivity, complexity) • Give algorithms (sound, complete) KR = engineering • Build applications • Show high performance • Show low engineering costs
- 5. BUT AN EXPERIMENT IN THE PAST 10 YEARS MADE IT POSSIBLE TO DO SOMETHING VERY DIFFERENT: OBSERVE HOW KNOWLEDGE REPRESENTATIONS BEHAVE AT VERY LARGE SCALE
- 7. Rest of the talk • Which KR’s were part of the experiment? • How much of it was there to observe? • How did we manage to observe it? • What did we learn from observing it?
- 8. Which KR’s ?
- 10. RDF (for logicians) • ground binary predicate: 𝑃(𝑂1, 𝑂2) • Limited existential variables: ∃𝑥: 𝑃 𝐶1, 𝑥 ∧ 𝑃 𝐶2, 𝑥 • Type is unary predicate: 𝑇𝑖 𝑥 • Subtypes ∀𝑥: 𝑇1 𝑥 → 𝑇2(𝑥) • Type restrictions ∀𝑥, 𝑦: 𝑃 𝑥, 𝑦 → 𝑇1 𝑥 ∧ 𝑇2(𝑦) • Equality: 𝑂1= 𝑂2 • Extensions to DL: – Distjointness of types – Cardinality restrictions (0,1) – always decidable: sub-FOL.
- 11. RDF deduction
- 12. OWL Semantics
- 13. How much is there to observe?
- 14. ± 45-100 billion facts
- 15. 1 fact How big is 100 billion
- 16. Denny Vrandečić – AIFB, Universität Karlsruhe ≈ 1 fact per web-page 100 billion golfballs ≈ Jupiter
- 17. x T [<x> IsOfType <T>] different owners & locations < analgesic > BTW: How did it get so big? On the Web, anybody can say anything about anything
- 18. BTW: How did it get so big? On the Web, anybody can say anything about anything x T R
- 19. How did you manage to observe it?
- 22. LOD Laundromat Beek & Rietveld et al. 2014, LOD laundromat: a uniform way of publishing other people's dirty data http://lodlaundromat.org/pdf/lodla undry.pdf HDT Fernández & Martínez-Prieto & Gutiérrez, 2013, Binary RDF representation for publication and exchange (HDT) LDF Verborgh & Vander Sande et al. 2014, Web-Scale Querying through Linked Data Fragments
- 24. Surprisingly efficient 1 file 28,362,198,927 unique triples >650K data documents 524 GB of disk space 16 GB of RAM Only €305,- hardware cost Meta-Data for a lot of LOD http://www.semantic-web-journal.net/content/meta-data-lot-lod-2
- 25. Statistics (boring) triples 28,362,198,927 subject 3,214,347,198 predicates 1,168,932 objects 3,178,409,386 literals 5.3B
- 26. Re-use is fairly high… or not…
- 27. Analysing Logical identity Joe Raad Wouter Beek ESWC2018, under submission
- 28. Identity clusters LOD-a-lot File http: //lod-a-lot.lod.labs.vu.nl [Fernández 2017] 558 millions owl:sameAs (309 millions distinct terms) ≈ 4 hours 1. Extracting all owl:sameAs statements on the LOD HDT File (4.5 GB)
- 29. HDT File (4.5 GB) Identity Closure 1 Identity Closure 2 Identity Closure 89 387 082… - The largest Identity Closure contains 177 794 terms (contains all the countries in the world, Albert Enstein, « empty string », etc.) - The smallest Identity Closure contains 2 terms x owl:sameAs y z owl:sameAs y Identity Closure x y z 2. Generating the Identity Closure
- 31. Identity Closure « Cities » 3. Detecting Communities (using the Louvain Algorithm) This network (i.e. identity closure) has a community structure, as it can be grouped into different sets of nodes, with each set of nodes being densely connected internally. Goal: Find (and later Evaluate) the most “suspicious” identity links (i.e. the links between different communities)
- 32. 4. Application: debugging identity statements Identity closure containing the term “dbpedia.org/page/Barack_Obama” This Identity Closure contains 388 terms (i.e. 387 distinct terms are owl:sameAs this term) 95 communities detected largest community = 99 terms
- 33. 4. Application: debugging identity statements comm 0 comm 3 2 links Community 0 1. dbpedia.org/resource/B_hussein_obama 2. dbpedia.org/resource/Barack_H_Obama,_Jr 3. dbpedia.org/resource/Barak_hussein_obama 4. dbpedia.org/resource/President_Barack 5. dbpedia.org/resource/Senator_Barack_Obama 6. dbpedia.org/resource/Obama … 99. dbpedia.org/resource/Hussein_Obama Community 3 1. dbpedia.org/resource/Presidency_of_Barack_Obama 2. dbpedia.org/resource/Barack_Obama_Administration 3. dbpedia.org/resource/Barack_Obama_Cabinet 4. dbpedia.org/resource/Obama_White_House 5. dbpedia.org/resource/Obama_regime 6. dbpedia.org/resource/America_under_Obama … 52. dbpedia.org/resource/Presidential_transition_of_Barac k_Obama
- 34. Symbols or words? Steven de Rooij Peter Bloem Wouter Beek (ISWC 2016) http://www.cs.vu.nl/~frankh/postscript/ISWC2016.pdf
- 35. Symbols or words? Symbol names are supposed to be meaningless Aspirin headache analgesic pain symptomdrug treats treats
- 36. Measure mutual information content between string and semantics of a symbol E(x) = efficient encoding of x Mutual information content M(x,y) =E(x) + E(y) – E(x,y) Take x = symbol name of x as a string Take 𝑦1 = {types of x} ≈ semantics of x Take 𝑦2 = {properties of x} ≈ semantics of x Calculate M(x, 𝑦1) and M(x, 𝑦2) for all symbols in 600k datasets
- 37. But variables do encode meaning! Fraction of datasets with redundancy for types/predicates at significance level > 0.99 BTW, this is 600.000 datapoints (RDF docs)
- 38. Very different network structures for different predicates Tobias Kuhn Wouter Beek http://ceur-ws.org/Vol-1946/paper-05.pdf
- 39. skos:exactMatch
- 40. foaf:knows
- 41. osspr:contains
- 44. • We now have larger KB’s than ever before • We now have the instruments to observe and analyse these very large KB’s • We can use these insights for better tools: – query & inference – publish & maintain – visualise & explain – …
- 45. But my secret hope is that this will help us to understand the patterns of knowledge: AI as a computational theory of knowledge