[treerex]

Here are two papers that describe the techniques used by the FDA system (or that were used in the mid-2000's) to find these confusable names.

"Automatic identification of confusable drug names" (2006, http://goo.gl/W5DK0f PDF)

and

"Identification of Confusable Drug Names: A New Approach and Evaluation Methodology" (2004, http://goo.gl/RziUgf PDF)

Both by Grzegorz Kondrak and Bonnie Dorr.

I've used the BI-SIM in a medical-informatics system and it does quite well. I'm also a big fan of EDITEX, which for some uses is better.