Parsing an email to find a mailing address?

[Terry Olsen]

I have a very interesting request. A customer receives orders via email.
The email contains the shipping address, shipping method, email address
and phone number. This information is not all together, nor is it in the
same format in each email as the emails come from different order
houses.

The customer wants me to scan the email and pull the information out so
that it can be used to import into their shipping software.

I've been able to copy the body of the email to the windows clipboard
and then into a string variable in VB. But now the hard part, finding
the shipping address.

Has anyone done something like this? How do I recognize a shipping
address in a page full of text?

Some of the different variations on the address are:

Name: Name
Address: 123 Nowhere St
City/State/Zip: Glendale, CA 90000

or

Name
123 Nowhere St
Glendale, CA 90000

or

Name
123 Nowhere St
Glendale, California
90000 USA

or

Name: Name
Address: 123 Nowhere St
City: Glendale
State: ca
Country: usa
Zip: 90000


*** Sent via Developersdex http://www.developersdex.com ***

[Spam Catcher]

Terry Olsen <to******@hotma il.comwrote in news:uQQgrdHVHH A.3592
@TK2MSFTNGP06.p hx.gbl:

Has anyone done something like this? How do I recognize a shipping
address in a page full of text?

Is the e-mail address passed through the e-mail header? Can you extract it
there?

If not, you can use regular expressions to parse the email address.

[Pritcham]

Hi Terry

What you're basically trying to do is Information Extraction from free
text which is a common challenge/problem in the Natural Language
Processing/Machine Learning arena and is something that can be
incredibly complex due to the fact that it's a person adding these
addresses in the first place and therefore they could be added in
many, many different ways. If you're willing to go down that route
then I'd suggest doing a search on "Named Entity Recognition" and/or
Information Extraction / Natural Language Processing etc.

However, in your instance you state that the emails come from a number
of different order houses - if that's the case then it'd be a whole
lot easier if you could either agree a standard format for these
emails to use, or failing that, if the emails from each order house is
in the same format (but a different format to emails from other order
houses) then it'd be a whole lot simpler to get a sample of an email
from each different sender and work on parsing them one at a time with
regex or your particular weapon of choice.

Hope that helps
Martin