>>21
Not related to Haskell, but natural language parsing is pretty difficult as most of them evolved naturally, and to truly be able to obtain the true meaning of a sentence in an unmbiguous way requires "understanding" and "context". There's some promising statistical methods for dealing with this, but they're still quite imperfect. I think the best solution is to just aim at creating a human-level intelligence general AI and teach it human languages (such an AI could be based on a high-level model of our brain or a more low-level model such as a huge neural net that tries to model the neocortex and other related structures. Too bad such projects are still quite expensive these days).