>A coworker wants me to build a routine that will search a dbf for duplicate addresses. A simple task if we were looking for an exact match. However he wants me to use "fuzzy" logic so that the match doesn't have to be exact, just "pretty close". I haven't a clue how to go about this. Is there a routine around that I could use as a sample to get started?
You need to establish a pointing system based on various criteria in matching related items. It took us several months to fine tune such routine in the earlier '90s for a project I worked on. Then, based on corelation with the data and the point system, you can decide the percentage of acceptance which would then be consider a duplicate. If you fine tune that a lot, you can obtain a 99.5% percentage of efficacity on those which will be removed.
This is usually applied on mailing list procedures.