[postgis-users] Adding to geocoder data for better results?

[Stephen Woodbridge]

You might also try this address:

46896 state highway 1, Big Sur, CA 93920

http://imaptools.com:8080/demo/tiger-demo2.html?zoom=16&lat=36.25934&lon=-121.78893&layers=BTT&geo=46896%20state%20highway%201%2C%20Big%20Sur%2C%20CA%2093920

And see if it works. This works in my geocoder, which is not the postgis 
geocoder, but is based on Tiger data. The problem has more to do with 
how names are standardized. For example:

California 1
CA 1
Calif 1
State Hwy 1
sthwy 1
Hwy 1
State Route 1
Route 1
1
etc

If your query does not match or is not standardized the same way as the 
reference data the is being searched then it will fail to find that 
object and then it probably does a fuzzy search based on some fallback 
algorithm trying to find something that looks like it might be what you 
want and hence some strange answers.

-Steve W

On 4/25/2014 7:30 AM, Stephen Frost wrote:
> * Kent Richards (info at kentrichards.net) wrote:
>> When geocoding some addresses (perhaps, unincorporated areas), the returned results are not even in the ballpark.
>
> That really depends on what parts of the address you trust.  If the zip
> codes in your data are 100% accurate then you could modify the geocoder
> to fall back to using that if the street isn't found (or the city, of
> course), but...
>
>> select * from geocode('46896 California 1, Big Sur, CA 93920');
>>                           addy                          |                      geomout                       | rating
>> -------------------------------------------------------+----------------------------------------------------+--------
>>   (,,California,Via,,,"Dana Point",CA,92624,t)          | 0101000020AD10000073368950A46A5DC082C23F6CC5BB4040 |     38
>>   (,,California,Via,,,"Dana Point",CA,92675,t)          | 0101000020AD100000CCE9BF1EA76A5DC05EB5F600C9BB4040 |     39
>>   (,,California,Via,,,"San Juan Capistrano",CA,92675,t) | 0101000020AD1000006679E728A16A5DC0467CB31001BC4040 |     40
>>
>> Dana Point, CA is ~350 miles from Big Sur, CA.
>>
>> The search zip (93920) doesn't appear to be in zip_lookup or zip_lookup_all.
>>
>> "Big Sur" doesn't appear to be in place.
>
> It sounds like this is a large part of the issue to begin with- if the
> zip code and place aren't known to the geocoder, I'm not really sure how
> you expect it to be able to produce a better answer based on them...
> Unfortunately, the USPS still (afaik) wants to charge for public
> information like zip code data.  I'm not working with the USGS data much
> any more, unfortunately, but you could also make sure that you're using
> the latest data from them (they release once/year, last I checked).
>
>> I have voter registration data that includes full addresses for registered voters.  Is there a way to add this data to the geocoder tables so that geocoding returns something in the ballpark for addresses such as this?
>
> If you have geolocation information (POINT data) for those addresses
> then you might be able to incorporate it into the geocoder data, but I'm
> guessing you don't or you wouldn't be using the geocoder to begin
> with, and I don't see how just addresses could help with geocoding...
>
> Another option would be to look at using the OpenStreetMap data.  I had
> intended to and more-or-less just never got around to it.
>
> 	Thanks,
>
> 		Stephen
>
>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at lists.osgeo.org
> http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-users
>