Wikidata:WikidataCon 2017/Notes/Keynote 1

Title: Keynote #1: Fabian M. Suchanek Note-taker(s): VIGNERON

Speaker(s) edit

Useful links

Collaborative notes of the session edit

Talk about the YAGO knowledge base (extracted from Wikipedia), back in 2007. YAGO = Yet Another Great Ontology ( https://www.wikidata.org/wiki/Q8045810 )

Wikipedia category system, quite a messy thing (singer -> people by occupation -> etc. etc. ), doesn't always make sense (Bethlehem is a subcategory of Jesus...).

Use of Wordnet, problems of alignment...

Extraction of multiples Wikipedia (en, de, es, it, fr, fa, ro, etc.).

What the difference with DBpedia, YAGO was here first ;) DBpedia uses ontology of YAGO, etc.

Assumptions about completness (problem for machine learning), Elvis is married to Priscilla but was he married to others ?

Prediction: If you are a pope, you'll die in Rome (and a smart ass wikimedian ask about old popes in Avignon :D )

Incompleteness can be predirect quite accurately (especially for some cases like : if you have a death date, you should have a death place)

How can YAGO and WD can help each others? is it possible to exchange methods, data, etc.

Questions / Answers edit

What happens when you think you got a rules but then the rules change? Like parenting is 2 people and then you have surogacy.

Yes... Not always easy to follow the infoboxes...

Thiemo: you have awesome numbers of info but you have 5 millions wrong facts ?!?

We have a very high presence of wrong facts (maybe Elvis death date is one of them :P ). The only thing we can do is quantify the errors as precisely as possible. There is an probability for each relations.

Lily : Question about licensing. How can it work to import in WD?

The license is the same as WP become the knowledge from there. But the links are our own work.

Andy: do you teach your software for each infoboxes? and what do you do that now that there is Wikidata?

We are moving forward. Our task is to open new doors. Do new things : being multilingual, asses incompletness, etc. Our goal is not to be the biggest but try new things.

Shoud we import data and should there be an identifer pointing to Wikidata ?

This is a good question that should be explore, no definitive answer for the moment.

Overview of the session edit

Others projects
  • GeoNames
  • WordNet