Write a python script (that can be a used as a module too) that counts the stems in a bidix for a language pair over svn.  The idea is that it can be used as a module for this script (you should add its functionality there—should be quite easy), but it should also be usable as a stand-alone script.  Have a look at any pair's lg1-lg2.dix file and the bidix docs, but basically you should probably be doing a quick count of <e>...</e> entries.