I wonder if you could make the job of a natural language interpreter program easier, by using Bayesian statistical techniques to help determine the subject area of a conversation? For instance, using Bayesian statistical inference to tell whether they were talking about directory trees or the kinds of trees you plant?
Bayesian statistical inference might be faster than directly interpreting each sentence/paragraph, as you should be able to narrow down the search space considerably.
I wonder if the human mind works this way?
