Wikidata:Property proposal/territory overlapse

From Wikidata
Jump to navigation Jump to search

territory overlapse[edit]

Originally proposed at Wikidata:Property proposal/Place

Descriptionpart or all of the area associated with (this) entity overlaps part or all of the area associated with that entity
Data typeItem
Domaingeographical areas
Allowed valuesother geographical areas of different nature (symmetric constraint?)
ExampleDiocese of Auxerre (Q866941) --> Yonne (Q12816) (start time (P580) 1790 ; end time (P582) 1801)
instances of special-purpose district of the United States (Q610237) -> cities/counties they encompass/have a part in
Great Britain (Q23666) -> United Kingdom (Q145)
Planned useDocument the correspondance (partial or total) between areas described by different classifications.
See alsoWikidata:Property proposal/french canton
Motivation

located in the administrative territorial entity (P131) shows its limitations when it comes to representing overlapse between territorial entities arising from distinct classifications. Administrative vs religious is one (as with administrative subdivisions of countries vs dioceses), administrative vs political (as is the case with american special districts), geographical vs administrative (landforms vs countries or cities). We need a way to impart territorial correlation between these entities, and located in the administrative territorial entity (P131), which is purely administrative in nature, is not it. Alphos (talk) 12:24, 11 September 2016 (UTC)

Discussion
  • Symbol support vote.svg Support -Ash Crow (talk) 15:03, 11 September 2016 (UTC)
  • Symbol support vote.svg Support. good idea. author  TomT0m / talk page 17:00, 11 September 2016 (UTC)
  • Symbol oppose vote.svg Oppose I think this needs a better definition or at least a "subject item". I removed the label repeated in the definition. I do think we need a better solution than "has part" for this situation.
    --- Jura 11:09, 12 September 2016 (UTC)
    So in this diagram, does B has part (P527) A or A has part (P527) B ? Imagine we're in the situation where we don't want an entity for { A ∩ B }, which would mean creating entities of the form "part of A which are also in B" which aren't defined anywhere else. Alphos (talk) 14:08, 12 September 2016 (UTC)
    I added a definition the the description field. Does this fit?
    --- Jura 14:23, 12 September 2016 (UTC)
    Not quite. If that were it, has part (P527) would obviously do the trick Face-wink.svg
    It's more along the lines of "part or all of this entity overlaps part or all of that entity", which is slightly different. Glad I could convince you to at least not oppose this proposal :-)
    Alphos (talk) 15:52, 12 September 2016 (UTC)
    For lack of a simpler description - at least for now -, I added french and english descriptions ; and removed the tidbit you mistakenly added in the status line. Alphos (talk) 16:02, 12 September 2016 (UTC)
    Why are you omitting area and just using "entity"?
    --- Jura 17:25, 12 September 2016 (UTC)
    Because they're not just areas. They're attached with a meaning. Would you say France is an area ? Other than that, I wouldn't be opposed to saying "area" in the description. Alphos (talk) 18:39, 12 September 2016 (UTC)
    @Jura1: I didn't understand what you meant, but that makes the description of the property clearer. Thanks ! Alphos (talk) 11:35, 15 September 2016 (UTC)
  • Symbol support vote.svg Support - France is a complete mess... --Hsarrazin (talk) 18:45, 12 September 2016 (UTC)
  • Pictogram voting comment.svg Comment There seem to be a concensus up to now for this property, but a question arosed in the other proposal about a specialisation of this property for a french administration usecase arose a question that I think is also relevant at this level and I think should then alsobe answere here because answering yes/no could have a huge impact of the property scope : the overlapse beetween the timeline of existence of the divisions. There is no restriction currently written in this proposition, so this could lead to a huge number of statements because there is a huge number of divisions that existed other times and a nowday division could overlap beetween each of the now unused old division. I think we should then add a "contemporary" constraint. Also, this seems to be incomplete and we should have other propositions such as "division territory same as" to link two kind of administrative divisions that share exactly the same territory. note aux lecteurs francophone: je peux traduire en francais si quiconque me le demande. author  TomT0m / talk page 08:58, 14 September 2016 (UTC)
    • I think the advantage of this property is that it could be used for historic regions that can't be situated with certainty in present day boundaries, but where some overlap with present day boundaries is generally agreed. So contemporary constraints wont work. I don't think it's going to solve your canton problem, but people don't seem to want to work on that anyways.
      --- Jura 09:02, 14 September 2016 (UTC)
      • Maybe we need two versions of this property then. One alternative would be to create a "pivot item" which plays the same role as wikidata items to interwiki to avoid the combinatorial explosion of linking each administrative territory to each other over history - such as linking territory items free of any administrative reference to administrative divisions item rather than to link administrative divisions to each other I don't think that commenting on your perception of the intent of people will result to anything good. author  TomT0m / talk page 09:28, 14 September 2016 (UTC)
        • Nuvola apps error.png Not OK Creating intersection items is NOT the proper way to handle this : we would be creating the intersections, and would in some cases be the first to talk about them. Why do you think I mentionned the relationship Diocese of Auxerre (Q866941) --> Yonne (Q12816) (start time (P580) 1790 ; end time (P582) 1801) ? They overlap partially, but there is no such thing as "part of Yonne that was in the former diocese of Auxerre", creating it would be novel and impossible to source. Alphos (talk) 09:46, 14 September 2016 (UTC)
          • @Alphos: I'm not thinking of instersection items and I don't understand where you saw this. author  TomT0m / talk page 10:16, 14 September 2016 (UTC)
            • @TomT0m: It's pretty much what pivot items would entail for that property. And it's a bad idea on its face, as it would merely add complexity to something that can be represented quite simply. Alphos (talk) 10:36, 14 September 2016 (UTC)
  • Symbol support vote.svg Support Very useful property with many interesting usecases. Léna (talk) 10:25, 14 September 2016 (UTC)
  • Pictogram voting question.svg Question How do other ontologies handle this case? ChristianKl (talk) 19:55, 14 September 2016 (UTC)
    @ChristianKl: Ahem. They… don't.
    Take the example of Benton Community School District (Q4890552), which overlaps 8 towns in Iowa : its item doesn't list the towns it overlaps, nor do the items about the towns mention the special district. Granted, in this particular case has part (P527) would do the trick, as it comprises the entirety of each of the 8 towns, but not all special districts are cut to match town or county limits, if I'm not mistaken.
    Second example (yes, I did know I wasn't mistaken, please forgive my rhetorics :tongue:), Coachella Valley Water District (Q5137619). This special district is a twofer, per enwiki :
    • it covers 1000 square miles "mostly" (which means "not the whole special district") "within" (so "not the whole valley either") "Coachella Valley".
    • and it's "mostly within" (same snide remarks :tongue:) Riverside County, California ; but it also extends into "small portions of Imperial County and San Diego County, California".
    I'll admit we could merely give it located in the administrative territorial entity (P131) California, which would be undeniably true ; but then we'd miss the information that it overlaps (it doesn't comprise, and isn't comprised by) the Coachella Valley, and the three counties of Riverside, Imperial, and San Diego.
    This property, if created, would allow linking by intersection (which is relatively vague, I'll admit it) rather than by strict inclusion, thus opening new possibilities for Wikidata that no other property that I can think of can offer - without relying on original research and creating items (which would some of the time be meaningless) representing the intersection of two or more other items, thus adding complexity where none is really needed.
    Alphos (talk) 23:52, 14 September 2016 (UTC)
    @ChristianKl: I feel dumb : people pointed out that you may have referred to other ontologies than Wikidata, not other ontologies than France.
    I found two external ontologies that describe equivalent properties :
    • The UK Ordnance Survey defined an OWL for Spatial Relations, which described a property <owl:TransitiveProperty rdf:ID="spatiallyRelated">.
    • The Open Geospatial Consortium recommends in its Abstract test suite for CityGML instance documents, section B.1.5 "Spatial topology relations" (page 148 of the document, page 164 of the PDF) to "Verify that all spatial topology relations between spatial geometry objects are expressed using the XML concept of XLinks provided by GML version 3.1.1".
      • It also defines in its OpenGIS® Geography Markup Language (GML) Encoding Standard (page 147 of the document, 157 of the PDF) gml:topoComplexProperty as such : "The gml:topoComplexProperty property element encodes the relationship between a GML object and a topology complex", and gives the example "This allows a feature collection to contain or reference a topology complex that contains topology objects referenced by members of the feature collection". In effect, it allows a set of elements (for instance an object representing an area) to reference a topology complex (for instance another object representing an area) which is itself referenced by members of the initial set of elements (for instance, smaller areas or points belonging to the initial area).
    Sorry about my misunderstanding of your question !
    Alphos (talk) 08:36, 15 September 2016 (UTC)
  • Symbol support vote.svg Support, with some comments on the name and the definition (to follow shortly)... Robevans123 (talk) 17:16, 17 September 2016 (UTC)
Pictogram voting comment.svg Comment The link given to the Ordnance Survey ontology seemed to be dead, but I got to it by other means: Ordnance Survey (Q548721) Spatial Relations Ontology, where it is imported from GeoSPARQL (Q5533918), where the rdf is here. I found this a little difficult to read..., but the (eight) topographical relationships between spatial objects (three types of) are summarised on page 8, (page 26 of pdf) of this document OGC GeoSPARQL - A Geographic Query Language for RDF Data from the Open Geospatial Consortium (Q776903).
The three types of spatial objects (geometry types) mentioned (on page 7, 25 in pdf) are:
  • 0-dimensional geometries, symbol P, eg points
  • 1-dimensional geometries, symbol L, eg lines
  • 2-dimensional geometries, symbol A, eg polygons (areas)
and the eight relationships are:
  • equals
  • disjoint
  • intersects
  • touches
  • within
  • contains
  • overlaps - effectively the property in this proposal
  • crosses
Even after reading through the document I still couldn't quite get my head round, for example, the differences between intersects and overlaps... This page, Performing Relational Operations, from GitHub (Q364), goes a long way to explaining the different relationships.
So, after all that background, it seems that it might be useful to adopt these relationships.
  • the proposed property "territory overlapse" could be renamed "spatially overlaps" and could be applied to area/area overlaps, line/line overlaps, and point/point overlaps.
This could be used for any geospatial overlap that is noteworthy and is not properly covered by more hierarchical relationships, for example, a wildlife conservation area could overlap the territories of two countries
  • the other relationships could also be adopted (but this might need some more thought as some at least are subsets or supersets of existing properties, and some might be identical to existing properties). Robevans123 (talk) 18:29, 17 September 2016 (UTC)
Absolutely, with a caveat : I'd rather keep this property for territorial/topographical overlapse of areas, and create a new bit of ontology (metaproperties and derived properties) for other usecases.
For instance, the OGC also defines a Time overlapse, which would be useful to mark historically useful and well-documented relationships between time periods. For instance, the Middle Age in France is often said to be overlapsed by Italian Renaissance (more specifically, people usually say that Italy started its Renaissance when France was still in the Middle Ages).
Spatial overlapses could be used for vertical elements and horizontal ones, for instance Gulf Stream (Q130905) and Caribbean Sea (Q1247). Neither the former nor the latter are strictly two-dimensional elements, the former having a thickness which doesn't encompass the entire sea depth, and being vastly longer than it is thick ; and the latter encompassing the entire sea depth (obviously), and can be defined as a local pile of vertical divisions of the sea.
Obviously, spatial overlapse would be a metaproperty of the property I'm currently proposing, and having derived properties such as line overlapse and point overlapse (although I can't currently think of a usecase for point overlapse, I don't see why it should be dismissed).
In effect, this property and its derived property (also proposed) french canton, which carries additional meaning, are the first building blocks of a set of properties new to Wikidata, but that has already been defined by other groups.
I'd also like to apologize for not providing a proper link to the Ordnance Survey, relying on web.archive.org instead, and couldn't at the time find a documentation for it Face-wink.svg
Thanks for understanding the underlying idea !
Alphos (talk) 20:28, 17 September 2016 (UTC)
No worries on the links, and time and volume overlaps are something that needs to be covered someday. Yes - it may well be easier to understand area/line/point overlaps as three separate properties/relationships. The 0-dimensional geometries also include multi-points (for example a set of bus stops for a bus service) which could overlap with another set of points (for a different bus service). Bon voyage! Robevans123 (talk) 21:33, 17 September 2016 (UTC)
@Robevans123: Wow, I hadn't even though of that ! I have to say I love it :-) Alphos (talk) 08:39, 18 September 2016 (UTC)
  • ✓ Done @Alphons: please make good use of it.
    --- Jura 08:15, 18 September 2016 (UTC)