Saturday, June 17, 2017

#Wikidata vs #GeoNames - the first to throw a stone

Wikidata has some vocal people vilifying GeoNames. They insist that no data from GeoNames is included in Wikidata because "the quality is so bad". In my last post I wrote down assertions about Wikidata. One of them is that "Never mind how "bad" an external data source is, when they are willing to cooperate on the identification and curation of mutual differences, they are worthy of collaboration".

I wrote an email to Markc Wick, the founder of GeoNames and with his permission I can publish our mail exchange.

Hoi,The import of data from GeonNames into Wikipedia has been controversial. People say that the quality of the GeoNames data is not "good enough". It resulted in the deletion of thousands of articles from the Swedish Wikipedia. I am not Swedish, I did not follow their discussions but the problem is it sours collaboration with other parties because "their data might not be 100%".
This happened in the past, I care for the future. In Wikidata we do link to GeoNames (example Almere [1]).
There are several ways in which we can help each other and potentially even benefit from a collaboration. Wikidata is licensed with a CC-0 license and therefore GeoNames can have all our data and do with it as they please.
My initial proposal is for a comparison of the shared data. The data where GeoNames differs from Wikidata is potentially problematic. Concentrating on these differences together will improve both our and your data.
Would you be interested?
Thanks,
       GerardM
       Gerard Meijssen
His answer is everything I could hope for:
Hi Gerard
Thanks a lot for your email. A couple of weeks ago I have started to parse the wikidata extract and look for the matching attributes. Unfortunately I got interrupted and have not yet looked at the result of the parsing. I will continue as soon as I find the time.
The goal is to add the wikidata identifier to the alternatenames table with pseudos language code 'wkdt'. What I have noted so far is that sometimes the geonameids in wikidata go the wrong concept. For instance going to the city feature when the article is speaking about the administrative division or vice versa. This is one of the things I would like to check before adding the wikidataid as alternatename. GeoNames also has links to wikipedia.
I don't think wikipedia should import all geonames features, not all of them are relevant enough to justify a wikipedia article.
Best Regards 
Not only is there an interest to collaborate; Marc is checking the links in Wikidata referring to GeoNames and as can be expected he finds issues. As I asserted, this is to be expected and collaboration is the only way forward for optimal results.
Thanks,
      GerardM
Post a Comment