Assume you have a database full of addresses.
123 Main St.
Somewhere, USA 12345

Assume that the address is stored in a number of forms:

• Exactly as entered
• Standardized via Lingua::EN::AddressParse
• Encoded as soundex via Text::Soundex (exact/standardized)
• Encoded as metaphone via Text::Metaphone (exact/standardized)
• Broken out into individual pieces (street, city, zip, etc)
• Possibly each individual piece encoded as described above (exact/standardized)

The problem I am trying to solve is this: Given an address as input, find any "similar" addresses in the DB without comparing every address (via String::Approx or Text::Levenshtein for instance). Using just the encoding routines alone seem to have a lot of false positives and false negatives.

Anyone have any expertise in the area have some advice? I am sorry, I don't have any real data I can share.