Add training data for multi word states

[xmedr]

Overview

The parser was having trouble labelling states like Rhode Island and New Jersey. We've added some training data for that.

• Closes https://github.com/datamade/parserator.datamade.us/issues/189

Testing

• Open a virtual environment and download the latest version of usaddress
  • pip install usaddress --upgrade
• Open a python shell and try to parse some addresses with two word state names
  • I used some samples from Zillow

>>> import usaddress
>>> usaddress.parse("6774 68th St S, Horace, North Dakota 58047")
[('6774', 'AddressNumber'), ('68th', 'StreetName'), ('St', 'StreetNamePostType'), ('S,', 'StreetNamePostDirectional'), ('Horace,', 'PlaceName'), ('North', 'StateName'), ('Dakota', 'StateName'), ('58047', 'ZipCode')]

[derekeder]

nice. i had to take a few additional steps to get it working locally with the latest training data.

this ended up working for me:

pip install setuptools
python setup.py develop  
parserator train training/multi_word_state_addresses.xml usaddress 

[derekeder]

this didn't work without quotes. see mu-editor/mu#852