[Photon] Higher literal sensitivity

Christoph Lingg christoph at lingg.eu
Wed Feb 18 17:24:53 UTC 2015


Good catch, I forgot to include the mailing list mail address!

There are a number of possibilities, disabling AddressRow.isUsefulForContext should be the easiest!

Good luck!

> Am 18.02.2015 um 17:48 schrieb Andreas Karlsson <andreas.karlsson at appello.com>:
> 
> Hi Christoph,
> 
> Thanks, it was the context field that contained something about
> 'Kungsbacka', since the abundant hits in this case were all from that
> municipality. One example is 'Kyrkobyn'
> (http://nominatim.openstreetmap.org/details.php?place_id=23140010).
> 
> The reason I don't want the administratively related documents is because
> I want to use Photon strictly as a typeahead feature, with location bias.
> 
> After letting AddressRow.isUsefulForContext always return false I get
> something more to my liking.
> 
> Thanks for the help!
> 
> PS.
> Any special reason not replying to the mailing list?
> 
> 
> Best regards,
> --
> Andreas Karlsson
> 
> 
> 
> 
> 
> On 2015-02-18 08:17 , "Christoph Lingg" <christoph at lingg.eu> wrote:
> 
>> Hi Andreas,
>> 
>> It's possible that kungsba got translated to its English counterpart.
>> There is another possibility: we save additional information in the
>> context field of our documents, this is content we user to index but we
>> don't show it to the user. However those hits have a small priority and
>> should come last in the result list. Are you sure you don't want that
>> results?
>> 
>> Cheers, 
>> Christoph
>> 
>> Am 17.02.2015 17:25 schrieb Andreas Karlsson
>> <andreas.karlsson at appello.com>:
>>> 
>>> Hi,
>>> 
>>> When searching for ³Kungsba² (http://photon.komoot.de/api?q=Kungsba) you
>>> get a lot of results located near the city of Kungsbacka (Sweden), but
>>> nowhere in their documents do they have a string looking like
>>> ³Kungsbacka²
>>> or ³Kungsba². So it seems that ES responds with hits placed in the
>>> vicinity of more highly ranked hits.
>>> 
>>> In this case, with no bias and no tags, the query will be built from the
>>> template query.json. What is it in query.json that selects those
>>> documents
>>> with no mention of the search phrase? How would you go about tuning the
>>> query to avoid getting hits that doesn¹t match (or fuzzy match)
>>> literally?
>>> 
>>> Thanks!
>>> 
>>> Best regards,
>>> --
>>> Andreas Karlsson
>>> 
>>> _______________________________________________
>>> Photon mailing list
>>> Photon at openstreetmap.org
>>> https://lists.openstreetmap.org/listinfo/photon
> 




More information about the Photon mailing list