> > The implementation is nothing like NLI (it's pattern matching stuff), but > > that's the way to go :-) > > From what I have read in various places, there are very few sophisticated > NLI's. Most simply look for verb and objects and then ignore everything > else. The effect is sentences like this: > > "Today I think that I would like to take that box and put it on that shelf." > > Are exactly the same as "sentences" like this: > > "box put (on) shelf" > > Think about it, when NLI was all the craze (for games like Zork), these > programs were less than 256 kb. They couldn't be very complex, but they sure > *SEEMED* to understand what you typed. It largely is a trick, but it works. Sure. But if that's all it does, you don't get much and appear to do stupid things. Depends on what degree of "understanding" you want to exhibit. > Do you think there is some way that Narval and Piper could share an NLI > "engine"? Yes, it may even be called www.alicebot.org :-) It's one of the best NLI stuff available in free software I've seen so far. But I could dig other things out of my bookmarks stack (ThoughtTreasure maybe, hmm... were was that URL?) -- Nicolas Chauvat http://www.logilab.com - "Mais où est donc Ornicar ?" - LOGILAB, Paris (France)