Re: [OSM-talk] 112k Wikidata tags to add to OSM
I adjusted my criteria for islands, villages, towns and cities. There are now 102,691 matches and 230 mismatches. http://edwardbetts.com/osm-wikidata/ http://edwardbetts.com/osm-wikidata/mismatches.html -- Edward. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
Edward Betts edward at 4angle.com writes: I adjusted my criteria for islands, villages, towns and cities. There are now 102,691 matches and 230 mismatches. http://edwardbetts.com/osm-wikidata/ http://edwardbetts.com/osm-wikidata/mismatches.html How old are the datasets you are working from? -- Andrew ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
On 06/09/2014 09:15, Edward Betts wrote: I adjusted my criteria for islands, villages, towns and cities. There are now 102,691 matches and 230 mismatches. http://edwardbetts.com/osm-wikidata/ http://edwardbetts.com/osm-wikidata/mismatches.html One question that immediately springs to mind - is wikidata in any way geographically searchable or extractable, in the same way that OSM is via Overpass (like e.g. http://overpass-turbo.eu/s/4VH ) ? A list of mismatches worldwide (while useful in terms of total numbers) isn't useful in terms of what's near me that I might want to have a look at. Cheers, Andy ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
Dear Edward, first: great! And I really support the idea of importing wikidata tags. One thing I still don't get is why you are relying on wikipedia categories instead of on the instance of property in wikidata. In my view the correct way would be to use these and if they are not present at the moment make them become present. Since you already have good guesses about what an entity is an isntance of, maybe your guesses could be integrated in a game like: https://tools.wmflabs.org/wikidata-game/# . Afterwards you could rely on these and both communities would have gained. Cheers Morray ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
Morray os...@go4more.de wrote: One thing I still don't get is why you are relying on wikipedia categories instead of on the instance of property in wikidata. In my view the correct way would be to use these and if they are not present at the moment make them become present. Since you already have good guesses about what an entity is an isntance of, maybe your guesses could be integrated in a game like: https://tools.wmflabs.org/wikidata-game/# . Afterwards you could rely on these and both communities would have gained. I used the Wikipedia categories because the Wikidata 'instance of' property is often missing. I'm struggling to figure out how to use the Wikidata API to search for items by 'instance of' property. You're right I should add 'instance of' claims to Wikidata items. -- Edward. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
https://wdq.wmflabs.org/ is a powerful Tool for Wikidata I used the Wikipedia categories because the Wikidata 'instance of' property is often missing. I'm struggling to figure out how to use the Wikidata API to search for items by 'instance of' property. You're right I should add 'instance of' claims to Wikidata items. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
On 06/09/2014 22:15, Archer wrote: https://wdq.wmflabs.org/ is a powerful Tool for Wikidata Thanks, but the query builder doesn't appear to actually work, though (in either Chrome or SeaMonkey on Windows) without cutting and pasting the resulting magic codeinto a separate search form, and the results seem a bit confused (I searched at a location in Derbyshire for wikidata around here and while everything else was nearby, top of the list was a town 30 miles to the east). But at least it's an API - it should be possible to do something with it. Cheers, Andy ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
[OSM-talk] 112k Wikidata tags to add to OSM
I modified my code, adding more categories and extending the matching distance for some objects. I started checking addr:housename, some buildings have this tag but are missing the name tag. http://edwardbetts.com/osm-wikidata/ There are now 112,278 matches found. I thought the extended range would help reduce the number of mismatches, but I now have 2,393 mismatches. http://edwardbetts.com/osm-wikidata/mismatches.html I've got some ideas about how to fix some of the mismatches. Many of the mismatches are villages represented by both a node and a relation, but the relation isn't tagged with place=village, so my code can't tell it represents the same thing. Maybe relations that represent villages should be tagged with place=village. I'll could modify my code so it rejects nodes inside a way or relation with the same name. Example: Sachsendorf, Germany http://www.openstreetmap.org/node/240130457 http://www.openstreetmap.org/relation/2253462 -- Edward. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
Am 05.09.2014 11:06, schrieb Edward Betts: . I've got some ideas about how to fix some of the mismatches. Many of the mismatches are villages represented by both a node and a relation, but the relation isn't tagged with place=village, so my code can't tell it represents the same thing. Maybe relations that represent villages should be tagged with place=village. I'll could modify my code so it rejects nodes inside a way or relation with the same name. Example: Sachsendorf, Germany http://www.openstreetmap.org/node/240130457 http://www.openstreetmap.org/relation/2253462 This is rather off-topic in the sense that it has nearly nothing to do with wikidata, but in general using place on admin boundaries is considered bad practice (not the least because the administrative entities very often do not correspond to just one place or to a recognizable place at all). Opinions tend to vary on the best fix, one is to add the place node as admin_centre to the boundary relation, but that is not undisputed. Getting back to wikidata, IMHO there should be likely be separate wikidata entries allowing you to match the admin entity with the admin entry and the place with the place ... Simon signature.asc Description: OpenPGP digital signature ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
There is a difference between municipality and settlements like villages. Wikipedia articles are often about the whole municipality and not about a single village. So the wikidata-tag should only be tagged onto the administrative relation for the municipality Please do not add any Wikidata-Tags to German villages/municipality. This would cause a big mess as the mismatch list shows. I'd oppose this import at all at the moment. There are currently 21533 objects with wikidata tags in the OSM database at the moment. Your algorithm produces 2393 mismatches. This is a error rate of 10 % !! I think it's better to have not so much wikidata-tags in our database than about 11000 wrong tags. 2014-09-05 11:06 GMT+02:00 Edward Betts edw...@4angle.com: I modified my code, adding more categories and extending the matching distance for some objects. I started checking addr:housename, some buildings have this tag but are missing the name tag. http://edwardbetts.com/osm-wikidata/ There are now 112,278 matches found. I thought the extended range would help reduce the number of mismatches, but I now have 2,393 mismatches. http://edwardbetts.com/osm-wikidata/mismatches.html I've got some ideas about how to fix some of the mismatches. Many of the mismatches are villages represented by both a node and a relation, but the relation isn't tagged with place=village, so my code can't tell it represents the same thing. Maybe relations that represent villages should be tagged with place=village. I'll could modify my code so it rejects nodes inside a way or relation with the same name. Example: Sachsendorf, Germany http://www.openstreetmap.org/node/240130457 http://www.openstreetmap.org/relation/2253462 -- Edward. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
Archer arc...@gulli.com wrote: There is a difference between municipality and settlements like villages. Wikipedia articles are often about the whole municipality and not about a single village. So the wikidata-tag should only be tagged onto the administrative relation for the municipality Please do not add any Wikidata-Tags to German villages/municipality. This would cause a big mess as the mismatch list shows. I'd oppose this import at all at the moment. There are currently 21533 objects with wikidata tags in the OSM database at the moment. Your algorithm produces 2393 mismatches. This is a error rate of 10 % !! I think it's better to have not so much wikidata-tags in our database than about 11000 wrong tags. Don't be alarmed, this is a work in progress. I'm not going to add any wikidata tags until we have reached a consensus. -- Edward. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
2014-09-05 11:31 GMT+02:00 Simon Poole si...@poole.ch: Example: Sachsendorf, Germany http://www.openstreetmap.org/node/240130457 http://www.openstreetmap.org/relation/2253462 This is rather off-topic in the sense that it has nearly nothing to do with wikidata, but in general using place on admin boundaries is considered bad practice (not the least because the administrative entities very often do not correspond to just one place or to a recognizable place at all). +1, with the exception of very dense areas where the administrative area and the settlement do indeed coincide. in this case there is already a place-tag at an area: http://www.openstreetmap.org/way/169202820 but there is no name on it, so I am not sure if this hamlet is describing the same object than the place=village with name Sachsendorf. It is also not clear to which settlement this area belongs: http://www.openstreetmap.org/way/88233220 (well, you could guess from the administrative relation, but the place tags don't tell it). (that's why it is generally not a good idea to add place tags to a residential landuse, because typically there will be other landuse as well in a settlement). cheers, Martin ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
Edward Betts edward at 4angle.com writes: Don't be alarmed, this is a work in progress. I'm not going to add any wikidata tags until we have reached a consensus. Is there an advantage deriving wikidata tags from existing wikipedia tags? -- Andrew ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
On Fri, Sep 5, 2014 at 11:06 AM, Edward Betts edw...@4angle.com wrote: I started checking addr:housename, some buildings have this tag but are missing the name tag. addr:housename is most of the time improperly used in OSM (should be in the name tag). I understand it is increasing your matching results but it should be reviewed when this tag is corresponding to a wikipedia article. Pieren ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
Pieren pier...@gmail.com wrote: On Fri, Sep 5, 2014 at 11:06 AM, Edward Betts edw...@4angle.com wrote: I started checking addr:housename, some buildings have this tag but are missing the name tag. addr:housename is most of the time improperly used in OSM (should be in the name tag). I understand it is increasing your matching results but it should be reviewed when this tag is corresponding to a wikipedia article. Agreed, here are the numbers for OSM objects where an addr:housename is present, but the name is missing: Apartment buildings: 11 Castles: 2 Commercial buildings: 3 Government buildings: 2 Houses: 9 Museums: 1 Office buildings: 1 Railway stations: 1 Residential buildings: 9 Schools: 1 Shopping malls: 1 Studios: 1 Try searching for 'no name' in this page for examples: http://edwardbetts.com/osm-wikidata/Apartment%20buildings.html -- Edward. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 112k Wikidata tags to add to OSM
:Stonehenge: https://www.wikidata.org/wiki/Q39671 https://www.wikidata.org/wiki/Q587584 both WD objects match the same OSM object: http://www.openstreetmap.org/way/37118074 Berlin Wall: http://wikidata.org/wiki/Q5086 matches only one way of the wall but the Wiki article says there are three parts of the wall and different watchtowers left. Maybe those objects should get each their own Wikidata-object. Yosemite National Park http://wikidata.org/wiki/Q180402 (Q180402) node Yosemite http://www.openstreetmap.org/node/408906523 should match the national park nothing spectacular yet. the main matching problems are villages, other settlements and administrative divisions as Wikipedia doesn't distinguish here. In Wikidata you have to distinguish to get a correct ontology. district -- municipality -- settlements. Much consolidation has to be done in Wikidata yet. - suburbs: Whitchurch, Bristol http://wikidata.org/wiki/Q7994194 (Q7994194) node Whitchurch http://www.openstreetmap.org/node/29674271 maybe should match the relation http://www.openstreetmap.org/relation/2944649 unclear, WP article says village but the map shows the city council ward Pihlajisto http://wikidata.org/wiki/Q3069705 (Q3069705) node Pihlajisto http://www.openstreetmap.org/node/206335421 maybe should match the relation http://www.openstreetmap.org/relation/190373 like the map in WP shows Ringwood East http://wikidata.org/wiki/Q7335020 (Q7335020) node Ringwood East http://www.openstreetmap.org/node/2592650968 maybe should match http://www.openstreetmap.org/relation/2401669 + more examples if needed - villages Cerhonice http://wikidata.org/wiki/Q2140210 (Q2140210) node Cerhonice http://www.openstreetmap.org/node/1599176850 maybe should match http://www.openstreetmap.org/relation/437515 Wikipedia says is a village and municipality. The village and the municipality have different areas and therefore should be split into different objects in Wikidata + mismatch list +++ - towns: Hilpoltstein http://wikidata.org/wiki/Q521132 (Q521132) Hilpoltstein http://www.openstreetmap.org/node/240112461 (node) should match the relation for the municipality Hilpoltstein and not the town: http://www.openstreetmap.org/node/240112461 For the town a different object has to be created in Wikidata. Pocking http://wikidata.org/wiki/Q279629 (Q279629) Pocking http://www.openstreetmap.org/node/63655606 (node) same case as Hilpoltstein, should match the following relation: http://www.openstreetmap.org/relation/958087 + many more examples if needed Lagoa http://wikidata.org/wiki/Q2652722 (Q2652722) node Lagoa http://www.openstreetmap.org/node/25277559 Lagoa Municipality http://wikidata.org/wiki/Q985519 (Q985519) node Lagoa http://www.openstreetmap.org/node/25277559 2014-09-05 11:06 GMT+02:00 Edward Betts edw...@4angle.com: I modified my code, adding more categories and extending the matching distance for some objects. I started checking addr:housename, some buildings have this tag but are missing the name tag. http://edwardbetts.com/osm-wikidata/ There are now 112,278 matches found. I thought the extended range would help reduce the number of mismatches, but I now have 2,393 mismatches. http://edwardbetts.com/osm-wikidata/mismatches.html I've got some ideas about how to fix some of the mismatches. Many of the mismatches are villages represented by both a node and a relation, but the relation isn't tagged with place=village, so my code can't tell it represents the same thing. Maybe relations that represent villages should be tagged with place=village. I'll could modify my code so it rejects nodes inside a way or relation with the same name. Example: Sachsendorf, Germany http://www.openstreetmap.org/node/240130457 http://www.openstreetmap.org/relation/2253462 -- Edward. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk