I was perhaps particularly responsible for attacking "the era of global place or feature type taxonomies" at the AAG meeting Sean mentions, and NB I am doubtful about the whole idea of feature type taxonomies, global or otherwise. They work OK in a GIS or on a traditional topographic map which is an inventory of physical objects which you can touch, and which some kind of classification can be applied to, but in a gazetteer context you start to find that what most people mean by "place" is not really about physical features:
-- The AAG sessions were partly about specifically historical gazetteers, and over time you tend to find that "places", and place-names, are more enduring than geographical features, especially human-made features. There is a substantial tradition of linguistic place-name research, which traces names back to their earliest forms and then interprets their meaning. For example, the English county contains many names ending in "-den", which in Old English meant a woodland pasture for pigs, or more generally a clearing but go to those places and you won't find many pigs grazing. In general, large cities have overwhelmed the features they were named after: places outlive features, and with most important "places" there is a whole bunch of features associated with the name ("cities"="bunches of features"). -- It may well be that the quasi-judicial approach of the US Board on Geographic Names, focused on what things should be called rather than with the diversity of naming that actually happens, enables systematic feature typing to just about work. However, once we start crowd-sourcing gazetteers we rapidly get a total mess. For example, Geonames does include just about every sizeable settlement in Britain, but they appear variously as populated places (feature class='PPL'), third-level administrative units ('ADM3') and fourth level ('ADM4'). The uncertainty about administrative level is because parishes end up as either ADM3 or ADM4 depending on whether the next unit up is a district or a unitary authority, but the problem with the populated places is that if you follow the feature typing rules then just about every one of the administrative units SHOULD also be in there as a populated place, but in practice few are. The "feature typing" in Wikipedia is even more of a mess, although this is of course only a problem if you treat Wikipedia/Dbpedia as a large systematic gazetteer. Personally, I think "placenames" are a kind of social tagging for locations, and at any given place you may find any number of associated features, or none. Localities within suburban areas are especially interesting, as they are full of buildings but it is often VERY hard to identify particular buildings that define the locality. "Nag's Head" in north London is another of my favourite examples: the pub called that closed year's ago, but there is plenty of evidence the place still exists: http://en.wikipedia.org/wiki/Nag%27s_Head,_London From: Sean Gillies <sean.gill...@gmail.com> Date: Monday, 25 June 2012 16:58 To: Joshua Lieberman <j...@oklieb.net> Cc: Lin Clark <lin.w.cl...@gmail.com>, "geowanking@geowanking.org" <Geowanking@geowanking.org> Subject: [Geowanking] Place taxonomies (was Re: Improving http://schema.org/GeoShape) Josh (and all), IIRC, in our gazetteering session at the 2011 AAG a panel concluded that the era of global place or feature type taxonomies was over. Do you think that's still a valid conclusion? Should schema.org <http://schema.org> be advised to stay away from this area? On Sat, Jun 23, 2012 at 11:20 AM, Joshua Lieberman <j...@oklieb.net> wrote: > > > >> >> >> >> ... so what might we say of it in Schema.org <http://Schema.org> ? >> >> Well, maybe we would say it's an Island? Pretty tough, >> http://schema.org/Landform has Volcano and Continent but no island. >> Are there standard medium-sized lists of expected values we should be >> using here? > > From Geonames.org <http://Geonames.org> : > ISLisland a tract of land, smaller than a continent, surrounded by water at > high water > ISLETislet small island, bigger than rock, smaller than island. > ISLFartificial island an island created by landfill or diking and filling in a > wetland, bay, or lagoon > ISLMmangrove island a mangrove swamp surrounded by a waterbody > ISLSislands tracts of land, smaller than a continent, surrounded by water at > high water > ISLTland-tied island a coastal island connected to the mainland by barrier > beaches, levees or dikes > ISLXsection of island > ISTH isthmusa narrow strip of land connecting two larger land masses and > bordered by water > >> >> Ok so picking from http://schema.org/Place ... let's go with >> http://schema.org/TouristAttraction. Maybe we'll describe the island, >> and then use Place's containedIn relation to describe the city (also >> called Korcula) that's on the island. >> >> So, it's a thing of type http://schema.org/TouristAttraction ... it >> has a 'geo' property pointing to a thing that is of type >> http://schema.org/GeoShape ... which in turn has a 'polygon' property >> whose value is the Text, >> >> "17.052154541015, 42.984695434571 17.108459472656, 42.969589233399 >> 17.140045166015, 42.962722778321 17.166137695312, 42.942123413087 >> 17.181243896484, 42.929763793946 17.212829589844, 42.922897338868 >> 17.208709716797, 42.898178100587 17.177124023437, 42.898178100587 >> 17.090606689453, 42.895431518555 16.964263916015, 42.911911010743 >> 16.839294433594, 42.87208557129 16.833801269531, 42.888565063477 >> 16.732177734375, 42.895431518555 16.725311279297, 42.892684936524 >> 16.670379638672, 42.910537719727 16.656646728515, 42.896804809571 >> 16.618194580078, 42.927017211915 16.658020019531, 42.95997619629 >> 16.658020019531, 42.962722778321 16.603088378906, 43.001174926758 >> 16.658020019531, 43.008041381837 16.857147216797, 42.979202270508 >> 17.052154541015, 42.984695434571". >> >> As discussed earlier, the first and last pairs are identical. >> >> I could write this out long-hand in Microdata or RDFa Lite, but the >> basic model is the key concern. I think we're getting somewhere, >> thanks for your help! >> >> Nearby in the Web, Wikipedia has pages for both island and city, >> http://en.wikipedia.org/wiki/Kor%C4%8Dula and >> http://en.wikipedia.org/wiki/Kor%C4%8Dula_(town) >> <http://en.wikipedia.org/wiki/Kor%C4%8Dula_%28town%29> >> >> It's also btw the alleged-birthplace-of >> http://en.wikipedia.org/wiki/Marco_Polo but let's get the basics >> working first :) >> >> cheers, >> >> Dan > > _______________________________________________ > Geowanking mailing list > Geowanking@geowanking.org > http://geowanking.org/mailman/listinfo/geowanking_geowanking.org > > > > > -- > Sean Gillies > _______________________________________________ Geowanking mailing list > Geowanking@geowanking.org > http://geowanking.org/mailman/listinfo/geowanking_geowanking.org
_______________________________________________ Geowanking mailing list Geowanking@geowanking.org http://geowanking.org/mailman/listinfo/geowanking_geowanking.org