Misinterpretation of country - GBIF datasets

[ManonGros]

Hi, I am trying to sort a bit some of the issues on the GBIF feedback repository and I thought I would forward you the following issues:

• Processing Country from Plazi is erroneous. The country is Nicaragua as is evident and the title of the paper and as communicated by the collector only noticed this because he has never collected in Grenada & yet it appears in bionomia.net gbif/portal-feedback#2866: concerns this dataset: https://doi.org/10.11646/zootaxa.4161.2.3
• Neotropical does not usually include Iceland. gbif/portal-feedback#2420: concerns this dataset: https://doi.org/10.11646/zootaxa.4427.1.1
• Georgian records in the US gbif/portal-feedback#2380 which you are already aware of but since it is still on going, I thought I would put it here to keep in mind.

Let me know if you need anything!

[mguidoti]

Hi @ManonGros,

Thanks for this list. I'll check with @myrmoteras how he want to handle them, and go from there.

But just so I can understand: I think gbif/portal-feedback#2420 and gbif/portal-feedback#2380 are related as the first cites the latter?

All the best,

[ManonGros]

yes I think so (but I am not sure)

[ManonGros]

It seems like this new issue could also be related: gbif/portal-feedback#3030

[mguidoti]

I'm checking with @myrmoteras and Guido whether I'll act on these issues, or Guido will do it.

Thanks for reporting them!

[gsautter]

It's a similar problem as (gbif/portal-feedback#2380), yes ... with "Wales" as in "New South Wales" being the culprit this time ... that together with an erroneous materials citation split between "N.S." and "Wales" (in what is supposed to read "N. S. Wales") elevates "Wales" to a standalone country level, which ISO normalizes to the UK.

Geographical homonyms (if infix based ones in this instance) ... we're doing a lot more to sort these out than we did back in 2016, when this article was processed, but these legacy error remain.
The best way of going about this might well be using the stats to get all the materials citations containing "Island" (German name of Iceland, and thus all the trouble), "Wales", and "Georgia" (plain top level ambiguity in English) to get an overview, and then devise a viable cleanup strategy ... I manually corrected a lot of the "Georgia" mistakes earlier this year, but the other two might be concerning far more treatments, so some kind of automated first approach might help to at least reduce cleanup effort.

[gsautter]

That said, maybe we should add a "problematic/ambiguous country name" check in the QC?

[mguidoti]

Hi @gsautter ,

I think an additional QC rule to cover this might be very interesting. Better false positives than data issues, in my opinion.

And let us cover this, then. I'll query TB Stats and divide the task among people. Sooner than later we will have this covered.

What do you think?

[gsautter]

Let's get the stats and analyze this first ... maybe we can devise some automated filter to deal with the lion's share of the obvious cases ... no need to have the office do more tedious and repetitive work than necessary ...

[mguidoti]

@gsautter,

Found 175 articles that might contain the Wales/New South Wales issue... but you might be able to filter them a bit further using your own API.

Please, tell me what you think.

Thanks!

[gsautter]

Well, I don't really have an "own API" or anything ... just using the stats line anyone else ... but sure, there might be ways of filtering further, if somewhat downstream using the whole treatments, e.g. looking for other UK constituents like "Scotland" and "England", or for the presence or absence of Australia or other Australian territories ... we'll see, the data will tell.