John, I guess it is always good to talk as a data scientist - with numbers and facts. Here's why matching by coordinates would not work. This query calculates the distance between the OSM nodes, and the coordinates that Wikidata has for those nodes. I only looked at nodes, because ways and relations are even more incorrect - Wiki only has a center point. The results are bucketed by the distance (in km) - the bigger the distance, the bigger the mismatch between OSM and Wikipedia. As you can see, only a small number of nodes are accurate to 10 meters. Query: http://tinyurl.com/ybp4tp7a
diff in km number of nodes <0.01 75,027 <0.1 131,644 <0.5 147,637 <1 46,891 <2 28,049 <5 10,792 <10 3,537 10+ 7,239 Is this a convincing argument why we should have a Wikipedia/Wikidata link, as oppose to calculate it? The other issue is why we need Wikidata links - while I have said it many times, let me say it again. Because the current system is badly broken - as is evident by tens of thousands of errors that my approach has uncovered. I am not advocating to delete Wikipedia tag. Only that when you use wikipedia tag, it creates a burden on the community to maintain, and community is clearly unable to keep up with the changes on the Wikipedia side. So instead of using just the bad link (page title), I am advocating to use a good link (wikidata). We are already using it for 90%. Why not fill in the last 10%? It does not change anything of how you do your mapping. It simply helps those who want to fix errors, or view corresponding wikipedia articles even if it gets renamed. On Sun, Oct 1, 2017 at 8:50 PM, john whelan <jwhelan0...@gmail.com> wrote: > >Assuming my above arguments has convinced you > > No I still do not see a requirement here, but there again I'm only part of > the community and that's the concern you appear to be ramming this down our > threats. As for what iD does or does not do, I don't see that is relevant. > > Why does OSM need it and why are you unable to put forth a convincing > argument that is accepted by the community? A ninety percent acceptance > rate will be fine but I'm not seeing it. >
_______________________________________________ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
