fuzzygeo provides a fuzzy geocoding routine for geocoding at the named entity (city or similar) level. It has been built to resolve messy address and address-in-name data supplied in the PATSTAT international patent file. It computes edit distances between city names and parsed addresses to find appropriate matches. To keep search times reasonable on long city lists, it first tries to intelligently restrict the set of possible cities.
fuzzygeo requires a database of named entities matched to their latitudes and longitudes. For PATSTAT, we used the Maxmind database. But any database that can be coerced to a form country:region:city:lat:lng:population will work.
import pandas as pd
import fuzzygeo
# Assumes city_db has columns city, lat, lng
city_db = pd.read_csv('city_database.csv')
geocoder = fuzzygeo.fuzzygeo(city_db)
address = '1234 n. clark chicago il'
geocoded_address = geocoder(address, 'us', threshold=0.75)