usaddress is a python library for parsing unstructured address strings into address components, using advanced NLP methods. Try it out on our web interface! For those who aren't python developers, we also have an API.

What this can do: Using a probabilistic model, it makes (very educated) guesses in identifying address components, even in tricky cases where rule-based parsers typically break down.

What this cannot do: It cannot identify address components with perfect accuracy, nor can it verify that a given address is correct/valid.

1. Install usaddress with pip, a tool for installing and managing python packages (beginner's guide here)

In the terminal,

pip install usaddress

2. Parse some addresses!

Note that parse and tag are different methods:

import usaddress
addr='123 Main St. Suite 100 Chicago, IL'

# The parse method will split your address string into components, and label each component.
# expected output: [(u'123', 'AddressNumber'), (u'Main', 'StreetName'), (u'St.', 'StreetNamePostType'), (u'Suite', 'OccupancyType'), (u'100', 'OccupancyIdentifier'), (u'Chicago,', 'PlaceName'), (u'IL', 'StateName')]
usaddress.parse(addr)

# The tag method will try to be a little smarter
# it will merge consecutive components, strip commas, & return an address type
# expected output: (OrderedDict([('AddressNumber', u'123'), ('StreetName', u'Main'), ('StreetNamePostType', u'St.'), ('OccupancyType', u'Suite'), ('OccupancyIdentifier', u'100'), ('PlaceName', u'Chicago'), ('StateName', u'IL')]), 'Street Address')
usaddress.tag(addr)

usaddress uses parserator, a library for making and improving probabilistic parsers - specifically, parsers that use python-crfsuite's implementation of conditional random fields. Parserator allows you to train the usaddress parser's model (a .crfsuite settings file) on labeled training data, and provides tools for adding new labeled training data.

git clone https://github.com/datamade/usaddress.git  
cd usaddress  
pip install -r requirements.txt  
python setup.py develop  
parserator train training/labeled.xml usaddress  
nosetests .  

If there are address formats that the parser isn't performing well on, you can add them to training data. As the usaddress parser continually learns about new cases, it will continually become smarter and more robust.

parserator label [infile] [outfile] usaddress  

Our main training file is training/labeled.xml so you can do

parserator label [infile] training/labeled.xml usaddress  

This will start a console labeling task, where you will be prompted to label raw strings via the command line. For more info on using parserator, see the parserator documentation.

If you've added new training data, you will need to re-train the model.

parserator train [traindata] usaddress  

So, you could do

parserator train training/labeled.xml usaddress  

To set multiple files as traindata, separate them with commas (e.g. training/foo.xml,training/bar.xml)

Contribute back by making a pull request with your added training examples.

• Web Interface: https://parserator.datamade.us/usaddress
• Python Package Distribution: https://pypi.python.org/pypi/usaddress
• Python Package Documentation: http://usaddress.rtfd.org/
• API Documentation: https://parserator.datamade.us/api-docs
• Repository: https://github.com/datamade/usaddress
• Issues: https://github.com/datamade/usaddress/issues
• Blog post: http://datamade.us/blog/parsing-addresses-with-usaddress/

• Forest Gregg, DataMade
• Cathy Deng, DataMade
• Miroslav Batchkarov, University of Sussex

Report issues in the issue tracker

If an address was parsed incorrectly, please let us know! If possible, please send over a few real-world examples of similar address patterns, along with some info about the source of the data - this will help us train the parser and improve its performance.

If something in the library is not behaving intuitively, it is a bug, and should be reported.

• Fork the project.
• Make your feature addition or bug fix.
• Send us a pull request. Bonus points for topic branches.

Copyright (c) 2014 Atlanta Journal Constitution. Released under the MIT License.