[OSM-talk] Nominatim weakness

[Sarah Hoffmann]

On Mon, Dec 14, 2015 at 08:45:38AM +0100, Maarten Deen wrote:
> On 2015-12-14 01:02, Tom Hughes wrote:
> >On 13/12/15 22:52, John Goodman wrote:
> >
> >>For example, if my map is showing an area of the United States where I
> >>happen to know a mapped Starbucks exists, and I search for "Starbucks"
> >>in the search panel, the entire panel is filled with Starbucks
> >>in Japan.
> >>Do the same on a "competitor's" map, and you get what you expect: the
> >>Starbucks that are closest to the current map view are listed first.
> >
> >That shouldn't happen as we we pass the current view to Nominatim and
> >it is supposed to prioritise results in that area.
> 
> Then that logic is seriously flawed (if not broken).
> I noticed this yesterday, and I did this again just now. I opened
> OSM and I get the map at [1]. I look for Thorn (which for me is a
> town in the Netherlands [2] )
> The first results (now and yesterday) are (in that order)
> - City Toruń, Kuyavian-Pomeranian Voivodeship, Poland
> - Administrative Boundary Toruń, Kuyavian-Pomeranian Voivodeship, Poland
> - County Boundary Toruń, Kuyavian-Pomeranian Voivodeship, Poland
> - Suburb Boundary Thorn, Maasgouw, Province of Limburg, Netherlands,
> The Netherlands
> 
> The first three are basically the same and have a name:de=Thorn.

This is the same problem of importance. Reordering according to
your search query does happen but is weighted against importance.
Torun is much more important than the suburb in the Netherlands,
so it wins.

There is always room for improvement with the ranking but given
that the results at least appear on the list, there are other more
pressing issues.

Sarah