Spoken form address and closest match support

[sergproua]

I was testing different addresses with Extracting US street addresses from arbitrary text and could not make it work with spoken form input or with typos, typical in speech recognition.

Below are few examples:

Works:
My address is 11629 Charter Oak Ct. Reston Virginia

Does not work (street number in spoken form):
My address is eleven six two nine Charter Oak Ct. Reston Virginia

Does not work (Reston misrecognized with Easton or Western):
My address is 11629 Charter Oak Ct. Easton Virginia
My address is 11629 Charter Oak Ct. Western Virginia

Is spoken form supported and if not, do you plan on implementing it?
Is it possible to receive nearest matches and probability?

[RyanLCox1]

Spoken form is not currently well supported, but you can use a library to convert the spoken form numbers into actual numbers. Something like https://pypi.org/project/text2num/ to convert your input before passing it along. There will probably still be issues with spaces between numbers. This is something we are looking into adding.