@inherentlee @Anke

I have a python tool that ingests open text medications names and spell checks them vs FDA and also our cloud of misspellings so that it learns "all possible" misspellings and eventually be used to feed back curated matches for all of those strings. If you approached it that way then the up front costs suck but the long term outputs get better fairly quickly I'd think
I'm a user though so I don't actually know with evidence.