We developed a string identifier comprised of different elements of a confirmation to train our algorithm. When we completed our system, over 80 percent of the OTC confirmations were perfect matches, proving we could substantially relieve workloads in the middle office.