Dimitris Kontokostas> we made some different design choices and map wikidata 
data directly into the DBpedia ontology.

I’m very interested in this.
A simple example: bgwiki started keeping Place Hierarchy in Wikidata because 
it’s much less efficient to keep it in deeply nested subtemplates.
This made it very hard for bgdbpedia to extract this info, because how do you 
mix e.g. dbo:partOf and wd:Pnnn?
So this is a logical continuation of the first step, which was for DBpedia to 
source inter-language links (owl:sameAs) from WD.
(I haven’t tracked the list in a while, could someone give me a link to such 
dump? Sorry)

Tom Morris> abandoned critical thinking and assumed DBpedia was dead now that 
we had WikiData

That’s quite false. Both have their strengths and weaknesses.

- DBpedia has much more info than Wikidata. For Chrissake, Wikidata doesn’t 
even have category>article assignments!

- Wikidata has more entities (true, "stubs"), 
  a lot of them created for coreferencing (authority control) purposes.
  IMHO there’s a bit of a revolution in this domain, check out 
  VIAF is moving to Wikidata coreferencing, which will get them double the name 
forms, 300k orgs and 700k persons.
  This is Big Deal to any  library of museum hack.

- Wikidata has easier access to labels. In DBpedia you have to do a 
wikiPageRedirects dance,
  and if you’re naïve you’ll assume “God Does Not Play Dice” is another name 
for Einstein

- Right now IMHO Wikidata has better direct types for persons. This is a shame 
and we need to fix it in DBpedia

> without thinking about how the two could evolve / compete / cooperate / 
> thrive.

We exactly have to think about this.
Last few months I've been worrying that the two communities don't much talk to 
each other.
We as humanity should leverage the strengths of both, to gain maximum benefits.

I've become active in both communities, and I feel no shame in such split 
loyalies. :-)
I went to DBpedia Dublin, now'll go to GlamWiki Hague... 
It's structured data each way!

Some little incoherencies:
- The DBpedia Extraction framework is a very ingenious thing, and the devs 
working on it are very smart.
  But can they compete with a thousand Wikidata bot writers? (plus Magnus who 
in my mind holds semi-God status)
- Meanwhile, DBpedia can't muster a willing Wikipedia hack to service 
mappings.dbpedia.org, which is stuck in the stone age.
- Wikidatians move around tons of data every day, but their understanding of 
RDF *as a community* is still a bit naïve.
- DBpedia holds tons of structured data, but Wikidata seems to plan to source 
it by individual bot contributions... maybe in full in 5 years time?
- DBpedia has grokked the black magic of dealing with hand-written 
*multilingual* units and conversions.
  Of a gazillion units http://mappings.dbpedia.org/index.php/DBpedia_Datatypes
  Last I looked, Wikidata folks shrugged this off with "too different from our 
data types"
- https://tools.wmflabs.org/mix-n-match/ is a crowdsourcing Wonder upon God's 
Earth,  but nary a DBpedian has heard of it IMHO

A Little Cooperation Goes a Long Way.


Wikidata-l mailing list

Reply via email to