in reply to De Duping Street Addresses Fuzzily
it is a tedious problem, but you could start by splitting each record into its components -- streetnum, streetpredir, streetname, streettype, streetpostdir, etc... There are accepted standard values available. Just search on the web. Then match the split-ted values against your reference lookup table. Once you have done that, you could come up with some kind of scoring to flag the most likely to least likely. Depending on how many records you have to scan and how often you have to do this, as well as how important it is for you to not have false positives/negatives in the match, you can decide if this exercise is worth your time. Otoh, if your employer is paying you for this...
|