On 06/17/2014 04:10 PM, Rob Nickerson wrote:
[...] My understanding is that the script would look for objects in wikidata that have a location (lat/longitude) and some wikidata tags that help us to identify the object (e.g. that it's a church and it's name is "St Nicolas's"). The script would then look for a similar object in OSM assessed according to the lat/longitude in OSM and other OSM tags (name=St Nicolas's, and amenity=place_of_worship). It would then flag the match for a human to check, or if accepted, automatically add the wikidata tag if the level of certainty matches some threshold.[...] - What are the risks of introducing bad data and how can we reduce/eliminate this?
I don't think this should ever be done without a human check, because there are often several related objects with similar names near each other.
To continue your example, there could be "St Nicolas's", "St Nicolas's Church Gardens", and "St Nicolas's Gift Shop" all near each other. Which one should the wikipedia page "St Nicolas's Church" match?
Now what if only the gardens exist in OSM because someone imported gardens in the area but no one has mapped the church yet?
In my opinion, the risks of doing this automatically are just too high. I'd be very happy to see a tool to do it *with* human checks, though. --Andrew _______________________________________________ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
