I might be able to do something with a word list. This is actually a medical glossary, so the number of distinct terms may be higher than common documents, but probably not too bad. The situation is complicated by the fact that I need to support phrase searching as well as and/or boolean searching. I could use a word list just to qualify records for further checking. I already have a dbm file for fast random access to the records once I know which ones I want. I'll try that and see how much of a difference it makes.