RE: Request for information

First step should probably be an address cleaning, which I think can be
accomplished with Workright Per/ZIP. Once clean, do your matching on
address, not name.

I took a different approach a few years back that worked quite well, but
took days to process the sql... I created a database of metaphone
tokens (could use soundex) corresponding to each word in a text field
(like name, or address), then ran a huge join to itself to find
addresses with high correlation scores.

This approach was too costly when I originally tried it, so I never
finished the work. That might be less of an issues with today's modern
systems, but I suspect the address cleansing approach will be much
easier and just as effective.

Hth,
-Eric DeLong

-----Original Message-----
From: midrange-l-bounces@xxxxxxxxxxxx
[mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Jim Townsend -
Systems Ideas
Sent: Tuesday, October 20, 2009 8:58 AM
To: midrange-l@xxxxxxxxxxxx
Subject: Request for information

I am looking for any information on Name and Addressing matching
solutions. I am not looking for a DE-DUP solution but one where I an
synchronize two customer lists to come up with common entries. (It is
pretty similar to DE-DUP.)

I am currently using a algorithm for the matching which removes vowels
and spaces then does a comparison based on 3 characters at a time to
figure out a match percent. The problem is we are matching on DBA's
where one person will have it listed as Joe's Snack Bar and Grill and
some one else will list it as Joe's Grill and Bar.

I am looking for either a build or buy solution.

Thank you for your assistance.

Jim Townsend