[Openstreetmap] GEOnet Names Server
Richard Fairhurst
richard at systemeD.net
Sat Sep 3 20:12:24 BST 2005
On 2 Sep 2005, at 19:49, Jon Stockill wrote:
> I noticed this appear on the "to look at" page.
>
> It'll give you a huge amount of information very quickly, but isn't
> without errors. There are spelling mistakes, and examples of someone
> doing far too much historical research (Gildersome near Leeds for
> example is also listed as "Home of the Gilders").
>
> Another problem is that the data on the rankings of "populated places"
> appears to be missing for the uk, making the selection of place names
> to display a bit of a problem (you can end up with small village names
> displayed, and large city names missed because there's no room if you
> try to handle the layout automatically).
>
> Having said all that, it's a lot of data, it appears to be available
> under sane terms, and weeding out the small inaccuracies will likely
> require far less work than building such a database from scratch.
I've used the GEOnet data for the UK quite a lot. It's the underlying
gazetteer for the search function at Waterscape.com, for example. A few
experiences:
1. It's no good as a source for mapping at any decent scale, because
the point resolution is so low - 0.0166667 of a degree.
2. There are way too many 'variant' (read: wrong) spellings of
placenames. For example, there's a Bangon and a Bangor at the same
lat/long.
3. As you say, there's no ranking of place size.
None of these are serious problems for a search function, as long as
the returned map is fairly large-scale (1:50,000's about right). But we
then tried to use it as the source data for placenames in our dynamic
PDF mapping (www.waterscape.com/boating/guides). The results were
pretty gruesome.
(YMMV outside the UK - I've no experience of other country data, I'm
afraid.)
Instead, we found another source of placenames... our old friend the
New Popular Edition (out-of-copyright 1940s OS maps). I manually traced
a bunch of placenames from around the waterway system using
Illustrator, then used a Perl script to parse the resulting files into
grid references.
For Openstreetmap UK, I'd suggest a different, but related, source: an
Ordnance Survey paper gazetteer from about the same time. This has
pretty complete coverage of placenames at 100m accuracy (and even
Milton Keynes existed then... albeit a bit smaller than it is now!).
The ranking issue still remains, but that can be solved manually.
I do own a copy of the gazetteer and would be happy to scan it
(slowly). Unfortunately, all the OCR software I've tried for OS X
(whether Mac-specific or generic Unix) is pretty feeble. If I scan it,
are there any volunteers out there to OCR the resulting files?
Richard
More information about the talk
mailing list