[Talk-GB] Schools Progress Tracker Update
edloach at gmail.com
Sun Jan 24 10:43:04 UTC 2016
> Just a quick note to say that I've updated the matching used in my
> tool at http://robert.mathmos.net/osm/schools/progress/ so that
> objects with a ref:edubase, ref:seedcode, or ref:deniirn that
> an entry on the official list will now always be 'matched' in my tool.
> (Previously the match would only be recorded if the OSM object was
> within 1km of the postcode centroid recorded in the official list.
> Also official list entries with a missing or invalid postcode could
> not be matched at all. Both of these problems are now fixed.)
> This new looseness in the matching is needed to cope with some
> that have missing postcodes in the official lists, and multiple sites
> and/or use PO boxes for their official address. However, it may lead
> to some false positives, so there is a new report at
> lists possible problems with the matching that would benefit from
> manual checking.
I think this is an improvement (certainly it has boosted the CO matching by a couple of % - though that might also be partly down to the surveys and edits I made yesterday).
But the above and my edits yesterday make me wonder if we can settle on a way to map two schools that share grounds. I have been tagging the grounds as amenity=school, and the separate buildings for the two schools with the different ref:edubase etc tags on. These don't get matched - I end up with a proximity match for one of the two schools to the school grounds area, and a blue dot for the other.
Similarly there are schools with two sites. For these I put the ref:edubase tags etc on the main site, being the one matched by address from the data. The second site I don't add anything to, but that leaves it as something which will possibly cause a false match based on proximity. Should I add just the ref:edubase tag to the second site as well?
example: school 114705 is being matched to a second site, though in this case that site is further away and without a ref:edubase than the actual school which is closer and does.
False match school: http://www.openstreetmap.org/way/35732246#map=16/51.8778/0.9127
Second site: http://www.openstreetmap.org/way/183160516
Primary site: http://www.openstreetmap.org/way/305233410
Can we agree best practice for these two cases, so Robert is able to detect them and I can remap where necessary?
I also have one situation where I have added the school as a node within another building, as a school seems to be something they offer in the building with the primary purpose being health related (perhaps hospital isn't the exact correct tag - maybe one of the social_facility ones).
I'm fairly sure though we don't want to start matching to nodes, so am happy to leave this showing as unmatched.
Incidentally the Braiswick Primary School on the new possible problems list is a new build school (not sure the larger building is finished yet) and either the postcode is wrong in the DfE data or the centroid location you have for it is way out - I suspect it is using the CO4 centroid rather than anything more accurate.
Thanks again for the matching tool. That and OsmAnd made surveying both easier and interesting yesterday (as I was letting OsmAnd guide me down lanes I didn't know just moving from one point to the next - was interesting to see the GPS trace this morning to find out where I'd been). I still managed to miss two of the points I'd aimed to survey - those are likely to have to wait until next weekend now. I've also found out how to get OsmAnd speaking directions in background and have Mapillary taking photos in foreground. Those 2700+ photos I think are currently processing.
More information about the Talk-GB