The product I support actually works somewhat similar to the way the article says to do it. One of the major difference is that the list of excluded words is coded into the proc, not stored in a table. Another is that we do a sequential check on the keywords. IE - Get a list of items with keyword 1 then compare to a list of items with keyword2 etc. It'll be interesting to see what kind of performance bump happens from doing the check in a set based manner instead of procedurally. We do have a benefit in that the list of terms is delivered to us on a quarterly basis so we get updates already split out so no work is required for us to do that.