Oh yeah they are very good at just ignoring typos, bad grammar, etc. They don't quite *reason* but extracting meaning from imperfect language (whether human or machine) is a strong suit.