I think that most of the reduction was due to me going through the obvious formatting issues last weekend. I used an import from the Geofabrik GB extract into an osm2pgsql database looking at the addr:postcode and postal_code tags. The database enabled me to easily get the appropriate OSM object IDs such that I could download and fix each in turn using JOSM after using some judgement.
I've also fixed a few way-off postcodes when compared to the distances to that which OS OpenData CodePoint thought that their centroids were at -- things like postcodes in Edinburgh (EH) mistakenly typed in with an Enfield (EN) postcode -- also identified using my import into PostGIS. Cases which were less clear I've added notes for, to have local mappers review. Gregory Sent from BlueMail On 16 Nov 2016 20:36, at 20:36, "Robert Whittaker (OSM lists)" <robert.whittaker+...@gmail.com> wrote: >My daily report of addr:postcode value errors at >http://robert.mathmos.net/osm/postcodes/osm-errors.html seems to be >being used by at least one other person, since the numbers of errors >showing there has dropped significantly now. The page is regenerated >daily, but unfortunately the data hasn't been refreshed for a few days >now because the source data on which it relies (The Geofrabrik GB >extract via the GB Taginfo instance) hasn't been updated in that time. > >I've also starting playing with a second report that lists location >discrepancies of postcode-tagged OSM objects compared with the >postcode centroid locations in Code-Point Open. This is less of an >exact science, since postcodes will not all be located at the centroid >for that postcode unit, and the allowable deviations vary depending on >the unit. However, you can find an initial list of postcodes that are >more than 1km from their official centroid at >http://robert.mathmos.net/osm/postcodes/location-errors.cgi -- there >are about 1500 of them, although quite a few are in groups where the >same postcode is on multiple neighbouring objects. Presumably most of >the 1500 will be cases of a typo being made by an editor or in the >data source they used, so they'll need manual checking and updating. > >If anyone fancies looking at any of these please feel free to dive in. >If you find any false positives (i.e. errors in the processing, or >postcodes that genuinely are that far from their centroid), please let >me know, and I'll see if there's anything that can be improved in the >tool, or if they need to be marked manually as ok. > >Robert. > >-- >Robert Whittaker > >_______________________________________________ >Talk-GB mailing list >Talk-GB@openstreetmap.org >https://lists.openstreetmap.org/listinfo/talk-gb
_______________________________________________ Talk-GB mailing list Talk-GB@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-gb