fdiff

fdiff -l lowfile -h highfile -a /Data/tmp/slimq
-a scores all 70k words are used
reads from stdin and provides diff scores

defaults to /Data/tmp/slimq (you can request
a different one with the -m flag, though).

You can also specify criterion words on the command line now with
the -L and -H flags (as opposed to the -l and -h which specify
files of words). These flags take comma-separated lists of words,
so you could do something like this:

fdiff -L bad,evil -H good,great