$150 (1 hour) to Parse Addresses with Python

I recently answered an ad on Craigslist from an accounting firm that had about 2500 addresses that needed parsing. Of course, python was the tool that came immediately to mind. After a little research, I found that the pandas_usaddress was the perfect package to handle this particular gig.

The customer provided the list in Excel and I offered to parse them for $100 — thinking that it would take about an hour as I knew that I would have to clean up a few manually. When I completed the job the customer agreed to pay $150 and also took my information for future jobs — so hopefully this will be a long-term relationship.

So, to get this done, I imported this package:

import pandas as pd
import pandas_usaddress as pda

df = pd.read_csv(filename)

I then used the pda package to parse the address as follows:

dfa = pda.tag(df, ['Address'], granularity='low', standardize=True)
dfa.head()

The package produced output as such:

Finally, I added a little code to join the address so I could produce results in the way that my client wanted.

for index, row in dfa.iterrows():
    df.at[index, 'Street'] = (str(row['AddressNumber']) + ' ' + str(row['StreetTag'])).upper()
    df.at[index, 'City'] = str(row['PlaceName']).upper()
    df.at[index, 'State'] = str(row['StateName']).upper()
    try:
        df.at[index, 'Zip Code'] = str(row['Address']).split(str(row['StateName']).upper())[1].split()[0]
    except:
        print(row['Address'])
        pass

Here is a picture of the code to make it a little clearer:

Then, I eliminated the extra columns:

dfb = dfa[['Business', 'Street', 'City', 'Zip Code']].copy()
dfb.head()

To produce the following result:

More content at PlainEnglish.io. Sign up for our free weekly newsletter. Follow us on Twitter and LinkedIn. Check out our Community Discord and join our Talent Collective.