But of course computers -do- understand the meaning of words in the same way they understand everything else they understand. They generate look-up tables/references and act accordingly. I think what you meant to say is it's extremely difficult to do this with English, not because of the need for the meta-info on the words, but because of the reliance on probability to determine the most likely use of a word in a sentence. This is how we, as humans, do it, and it's also why we have misunderstandings. Computers would have misunderstandings too, but they wouldn't be nearly as forgivable as ours are, because we're only human but computers are buggy/flawed when they make mistakes. It is certainly possible, especially with modern hardware, to write a program to diagram sentences. This is largely what the field of natural language processing ultimately attempts. But it won't and can't be perfect because -we- can't be perfect either. Once we start dealing with probabilities, there's always a chance of making a valid but incorrect choice.