[Commons-l] Friendliness & Lack of User Recognition
Paul Houle
paul at ontology2.com
Thu Feb 24 15:22:38 UTC 2011
On 2/23/2011 9:20 PM, geni wrote:
> Problem is that this is in practice a far better fit for wikipedia
> where such lists are generated in passing than commons.
And that's a critical insight.
From my perspective, wikipedia is a skeleton and commons is the
flesh that's hanging on it. If you want to improve the organization of
Commons, you've got the most incredible resource in the world to do
that... Wikipedia.
Today, Freebase and Dbpedia can be used together to form a rich
and powerful database and ontology that describes the contents of
Wikipedia. A "top 100" list doesn't need to be compiled by experts or
even by humans, but can be produced by a largely automated process.
For instance, you could look for things that are typed '/people/person'
in Freebase and then sort them in the order of how many Wikipedia
articles and produce a list of the "top 100" people that is pretty good
(except for the minor embarrassment that U.S. President #43 is the most
linked person in my sample.)
With a little work, it should be possible to build something that
makes lists like "Train stations in Poland that don't have pictures in
en.wikipedia", though it's a query that's not on my fingertips because
my system was built to pay attention to things that have photographs
and ignore things that don't.
Anyhow, Freebase is CC-BY so there's no problem feeding data from
it back into Wikipedia/commons. I'll offer that data quality is a lot
better in Freebase than dbpedia, so anyone going down this route will
save a lot of time by relying primarily on Freebase and using Dbpedia to
fill gaps.
More information about the Commons-l
mailing list