openvenues / libpostal

A C library for parsing/normalizing street addresses around the world. Powered by statistical NLP and open geo data.
MIT License
4.04k stars 417 forks source link

Ot senzing datamodel #616

Closed oskar700 closed 1 year ago

oskar700 commented 1 year ago

This adds the option of downloading Senzing data model in lieu of the default data model.

albarrentine commented 1 year ago

Thanks @oskar700, glad the new model training worked out! I downloaded it earlier to do a few spot checks and it looks like it indeed does solve a few issues folks have reported (and probably many more that have not been). Added a few comments on the pull request above. The main initial thought was making it --model=senzing which generally allows for models to be swapped out, know there are some language-specific versions out there, etc. Congrats to you and your team and thanks for contributing back!

brianmacy commented 1 year ago

Al, working on the differences. We plan on documenting the delta on our test set per country on our GitHub with any commentary. I expect we will be doing an updated model every 3-6mo.

oskar700 commented 1 year ago

Thanks @albarrentine for a great review. It is spot on.