The artof of knowledge engineering, or: knowledge engineering of art (original) (raw)

• 1. The Art of Knowledge Engineering or: knowledge engineering of art Guus Schreiber

• 2. Knowledge engineering: From art to discipline ■ Objective of knowledge engineering (KE) in 90’s ■ Use of knowledge patterns for “expert” tasks

• 3. The popularity of “ontology” ■ One of the first ontologies: traffic ontology – “On representational promiscuity” – “Vacuous paper with no content” (Brian Gaines, KAW 93) ■ Now seen as panacea for the Holy Grail of information integration ■ There is even a Web language for it ■ But will it stand the test of time? – are our current conceptions of formal classes and properties sufficient to grasp the complexity of the Web?

• 4. The Web: knowledge engineering for the masses ■ KE is outside the former small research community ■ Everyone is building hierarchies and describes classes

• 5. Intermezzo: Knowledge democracy ■ From a human-rights point of view the Web is a leap forward ■ Possibility for “everyone” to access knowledge – the empowered citizen ■ But history will tell whether this remains true

• 6. Web KE builds on a long tradition of vocabulary construction

• 7. Semantic annotation is now feasible

• 8. Associative reasoning: following Semantic Web links

• 9. The popularity of “ontology” alignment ■ Creating the (missing) links the the Linked Open Data Cloud ■ Multitude of alignment techniques available ■ Large evaluation initiative – OAEI ■ But will our alignment methods stand the test of time? Are the results good enough?

• 10. Semantic search types ■ DISAMBIGUATE: Can you give me alternative interpretations of term T? ■ DESCRIBE: Can you give me more information about concept C or Individual I? ■ QUESTION ANSWERING: does property P hold for object O? ■ ANSWER QUESTIONING: Jeopardy! ■ RELATION SEARCH: in what way(s) are object O1 and O2 related?

• 11. Use cases for semantic search

• 12. Picasso and Moulin de la Galette Location-based: relatively easy

• 13. How are Picasso and Matisse connected: Georges Braque Style- and time-based not trivial

• 14. How are Picasso and Matisse connected: 1907 the changes in the art world in Paris anno 1907 difficult

• 15. Matisse and “les fauves” Where does this term come from? 1905: the story of the critic difficult

• 16. The importance of narratives ■ Users like to get the “stories” behind the navigation paths in the graph ■ For this we need to have some minimal “understanding” of the meaning of the paths ■ Well-constructed minimal ontologies can provide such interpretations – graph patterns ■ Example: an event ontology

• 17. Web mining of piracy events from piracy reports & Web sources van Hage, Malaisé & van Erp, 2011

• 18. Problems in ontology alignment ■ We have not agreed on an adequate alignment vocabulary – misuse of owl:sameAs ■ We have no adequate methodology for evaluating alignments (Tordai et al., 2011) ■ In particular, people do not agree on how different classes align – and this is not because they don’t do it “right”

• 19. Beyond categories ■ The set theory on which ontology languages are built is inadequate for modelling how people think about categories (Lakoff) – Category boundaries are not hard: cf. art styles – People think of prototypes; some examples are very prototypical, others less ■ We also need to make meta-distinctions explicit – organizing class: “furniture” – base-level class: “chair” – domain-specific: “Windsor chair”

• 20. KE for the Web: the way forward?! ■ We are only scratching the surface in semantic search – large-scale experimentation needed – small minimal ontologies acting as search patterns ■ We need a revised alignment vocabulary – taking Lakoff’s notions into account ■ Attention for semantic detail matters – in search, in aligment, for story telling – and lay knowledge engineers are providing it! ■ Combining this with statistical technqiues is a powerful combination

• 21. Are Picasso and Herengracht 196 related?