openvenues / libpostal

A C library for parsing/normalizing street addresses around the world. Powered by statistical NLP and open geo data.
MIT License
4.09k stars 421 forks source link

Discrepancy Between Java libpostal and Python libpostal Results #650

Open shlomishasho27 opened 10 months ago

shlomishasho27 commented 10 months ago

Hey,

I encountered a discrepancy in the results obtained from the Java libpostal ("1.1-1.5.8" version) and the Python libpostal (postal==1.1.10). Specifically, when processing the address "33 elm st 5a" the outputs differ as follows:

Java libpostal result: [{“key”: “house_number”, “value”: “33”}, {“key”: “road”, “value”: “elm st”}, {“key”: “house_number”, “value”: “5a”}]

Python libpostal result: [{“key”: “house_number”, “value”: “33”}, {“key”: “road”, “value”: “elm st 5a”}]

I'm reaching out to seek clarification on this variance. Is this a version-related issue, and could you provide guidance on which Maven version corresponds to postal==1.1.10 in pip?

Thanks!

shlomishasho27 commented 10 months ago

@albarrentine @saudet FYI

saudet commented 10 months ago

I don't see any version 1.1.10 so that's probably an unrelated fork of this project: https://github.com/openvenues/libpostal/releases

shlomishasho27 commented 10 months ago

@saudet So do we want to use the 1.1 in pip? Is it equivelant to1.1-1.5.8 in maven?

matanzil commented 10 months ago

@saudet Hi, no Python 1.1 version - only 1.0, 1.1.1, 1.1.2... which one as far as you know is equivalent to 1.1-1.5.8? Thanks!