Mark asked some interesting questions in his interview.
I’ve already had to post a clarification of our claim to ”free-from unrestricted English’, which is causing the cognoscenti to raise their eyebrows. By ‘free-from unrestricted English’ we mean that there will be no lexical limitations (most English words are understood) and no syntactic limitations (as long as the order of your phrases or clauses is correct, we’ll try to deal with it). The complement (opposite) of ‘free-from unrestricted English’ is called ‘controlled English’ in literature. So what I (mostly) meant was that we won’t require the interface to Bot Colony to be in controlled English.
‘Free-from unrestricted English’ does not necessarily mean that we have to support all the 600,000 entries in a OED, when the norm is about 150,000 - 200,000. Unfortunately, we won’t be able to support (most) slang in the initial Beta version. However, we will support the language as it appears in a normal size dictionary, such as WordNet, with well-formed grammar.
Of course, once you get past parsing, the real problem is with semantics and world knowledge. Should the bots ‘raise their eyebrows’ if a player talks nonsense, like ‘The bacteria were driving the truck with a vengeance’ ? I think so. Would it be legitimate to question the player if she/he says that ? I think so. Otherwise, the robots will look moronic and dialogue won’t be interesting.
On another note, we’re currently working with some Montreal area Beta testers towards a wider December release. We’re currently working with just a few people, since there are still lots of problems and we can’t be responsive to a larger group at this stage. However, if anybody (not necessarily Montreal-based) has hands-on experience with conversational agents and AI and wants to participate in the Beta effort, please e-mail me and we’ll see what we can do.
