[OSM-dev] Bad result from name finder postcode search

Robert (Jamie) Munro rjmunro at arjam.net
Fri Feb 8 17:19:26 GMT 2008

Hash: SHA1

David Earl wrote:
| On 06/02/2008 08:11, Chris Jones wrote:
|> Robert (Jamie) Munro wrote:
|>> Hash: SHA1
|>> I realise that the name finder will never be totally correct, but I
|>> noticed the following error:
|> Beats searching for RG5 3DF (Reading) it gives you Manchester!
|> http://www.frankieandshadow.com/osm/?find=rg5+3df
| :-)
| I can tell you partly why: the first address Google pops up with for
| that postcode is "83, Beechwood Avenue, Woodley, Reading, Wokingham, RG5
| 3DF"
| The Wokingham confuses the search, and what the pattern matching ends up
| looking for is "Beechwood Avenue, Woodley" and there is indeed a
| Beechwood Avenue, Woodley only 12km from Manchester city centre.
| However, I'm surprised it doesn't match the Reading Woodley. I suspect,
| though I haven't checked, that that suburb has been added since November
| when the name finder database was last updated.
| As I said, it's only a first cut, and there is lots of room for
| improvement. Also, I think we have top level copyright free postcodes
| for most areas - certainly RG even if not RG5. I could reject results
| more than X km from that lat/lon.

That would be good, but the current top level postcodes, e.g. OX1 can be
biased because they have lots of correct postcodes in Oxford, but a
couple in the north sea off Scotland that pushes the average way of
course. One solution is to take the median position of the postcodes for
a region rather than the mean. The ultimate solution is to detect and
remove these outliers.

Do you log postcode searches? It might be nice to compare your results
to NPEmaps.

Robert (Jamie) Munro
Version: GnuPG v1.4.6 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


More information about the dev mailing list