Using ngram or part-of-speech data could create better mnemonics: https://rmmh.github.io/abbrase/

Filtering the list to require a minimum edit distance of 2 or 3 would be quite easy, too!