15396 Reporter: mraymond Type: Improvement Summary: Data published by a country about a continent giving too few results Resolution: WontFix Status: Resolved Created: 2014-03-18 10:39:16.394 Updated: 2014-04-30 14:51:55.112 Resolved: 2014-04-30 14:51:55.089
Description: When I apply a filter of Publishing Country= France, and then add another filter of Continent=Asia, I only get one result (in Ukraine). When I look at this map http://www.gbif.org/country/FR/publishing it looks like there should be quite a lot of data published by France about Asia. Is this the correct interpretation of how the filters should work? *Reporter*: Melianie Raymond *E-mail*: [mailto:mraymond]]]>
Author: firstname.lastname@example.org Comment: The continent filter only works on continent values supplied by the publisher - we do not derive the continent from coordinates or countries (yet) Created: 2014-03-18 10:44:31.894 Updated: 2014-03-18 10:44:31.894
Author: email@example.com Created: 2014-03-18 11:02:10.527 Updated: 2014-03-18 11:02:10.527 It would be good to have a clear definition of the continent extents, there are various views on this from the extreme of Britain not being Europe to the inclusion of all islands and even larger water bodies on Earth: http://en.wikipedia.org/wiki/Continent#Extent_of_continents
Author: firstname.lastname@example.org Created: 2014-03-18 12:12:37.188 Updated: 2014-03-18 12:12:37.188 I propose that unless we start interpreting continent correctly, we disable (e.g. hide) that filter. It needs to do a decent interpretation or else causes confusion and makes us look daft. Naively you could think of using a country->continent mapping, but that will fail for overseas territories. I don't believe there is a sensible way of doing this without using a ruleset that makes use of verbatim continent and latitude, similarly to how we interpret countries. E.g. something along the lines of Use case 1: verbatim record has continent and latitude and longitude - assert that the stated continent looks sensible for the lat / lng provided or else flag as suspicious and set to null. Use case 2: verbatim record has continent, country and latitude and longitude - assert that the stated continent and country look sensible for the lat / lng provided or else flag as suspicious and set both country and continent to null. Use case 3: verbatim record has continent but no latitude and longitude - interpret continent Thoughts?
Author: email@example.com Comment: i agree with Tim - a straight string parse is not going to be what people expect Created: 2014-03-18 12:14:12.568 Updated: 2014-03-18 12:14:12.568
Author: firstname.lastname@example.org Comment: fully agree too - a new jira for the improved interpretation or reuse this one? Created: 2014-03-18 12:16:31.898 Updated: 2014-03-18 12:16:31.898
Author: email@example.com Created: 2014-03-18 12:27:47.025 Updated: 2014-03-18 12:27:47.025 I think it would be more intuitive for visitors if the points shown corresponds to the 'country' view of data publishing. I guess this is where MR get confused when she could see lots of points published by France in Asia. Since verbatim continent is not always available from data publishers and lat/lng are more likely to present, I therefore think another use case could be considered: Use case 4: verbatim record has no continent but lat/lng exists, or interpreted lat/lng exists: - define continent - interpret according to lat/lng To my knowledge there is a GBIF-defined "regions" in adminstration which could be used to define or sync with our definition about continent. This way also helps communicate how our continent is defined.
Author: mraymond Comment: I think this will be a useful filter to have in future but agree that this should not be included in its current form as it is too confusing. Created: 2014-03-18 13:58:22.635 Updated: 2014-03-18 13:58:22.635
Author: firstname.lastname@example.org Comment: Replaced by new, proper continent filter as described in POR-2055 Created: 2014-04-30 14:51:55.11 Updated: 2014-04-30 14:51:55.11