rinigus / geocoder-nlp

Geocoder library based on libpostal normalization of libosmscout generated database
MIT License
21 stars 1 forks source link

Failure to find address in NL #64

Closed rinigus closed 2 years ago

rinigus commented 2 years ago

From https://github.com/rinigus/osmscout-server/issues/400

When searching as "4 Groenweg" the address below is not found:

4, Groenweg, Esch, Boxtel, Noord-Brabant; https://osm.org/go/0Eu0HnYT?m

However, it is in the database and can be found by typing "4 groenw" - adding "e" breaks it. Parsing looks to be correct, just something is off in the search.

Similar issue was reported for some other addresses by the same user - this is the address for testing.

rinigus commented 2 years ago

When running with geocoder-nlp test code:

./geocoder-nlp ../Maps/geocoder-nlp/europe-netherlands "" "" "4 groenweg"

Address parsing before full normalization:

house_number 4 
road groenweg 

Normalization:

house_number 4 
road groenw gro enw

h-0 4 groenw 4 gro enw 

Geocoder loaded
Search hierarchies:
Postal code: 

Levels: 2 -> {groenw, gro enw} / {4} / 

Levels: 1 -> {4 groenw, 4 gro enw} / 
Skipping hierarchy since search result already has more levels (2) than provided

Search results: 

4, Groenweg
4, Groenweg, Esch, Boxtel, Noord-Brabant

So, something is odd with normalization which prevents getting the results with "4 groenwe"

julianfairfax commented 2 years ago

I seem to now be having this issue for an address in Switzerland, which I don't recall having had as an issue before

rinigus commented 2 years ago

I can try to look into it. Just send the example, if needed by private email or message at SFOS forum. Email can be found in the server rpm changelog

julianfairfax commented 2 years ago

I can try to look into it. Just send the example, if needed by private email or message at SFOS forum. Email can be found in the server rpm changelog

I have sent it to you along with accompanying details on Matrix, from @juliannfairfax:matrix.org