The postal code search engine result page (SERP) positions in geonames have been improved for Austria. The most relevant place is now more likely to be on the top position of the postal code search page and the geonames webservice xml or JSON response.
This is best illustated with and example :
A search for code 6020 in Austria returns 124 rows as 124 places share the same postal code. Before this update the top position returned by this search was ‘Rum’ with ‘Sillwerk’ at the second position. Innsbruck was at position 44.
With the improved search algorithm Innsbruck is now on top :
http://www.geonames.org/postalcode-search.html?q=6020&country=at
We are using the Yahoo websearch to determine the relevance of a postal code and placename. A search in Yahoo for ‘6020 Innsbruck’ returns 500,000 hits whereas the search for ‘6020 Rum’ returns only 17 hits. These numbers are now part of the search algorithm for postal codes in geonames.
Improvements for other countries will be visible as soon as the numbers to determine the relevancy are available. A batch job on the developement server is working on this.
This is a good improvement. I think that postalcode-based searches are most efficient when they get one result (at least that’s the model we are thinking for integration in Tagzania), so it’s interesting that the first match in the results is the most important one. Congratulations for this step, as well as for the diplomatic appointment that Tagzania as made in the realms of Geonames 😉
http://www.tagzania.com/doc/about/diplomatic/
The batch job is now working on the Spanish dataset. It will take around eleven days to process over 55'000 records. I will keep you informed.
I think I will also use the results of the msn search engine and calculate a combined relevancy. Google, however, allows only 1'000 requests per day and is pretty useless for this kind of stuff. And yes I am very pleased and honoured.
Marc