The AI Engine

Right. That's all part of the preprocessing.

Quite a lovely speed boost. I'm stoked. This changes everything.

In the next week I'll have the input page for the Loebner prize, which will be very close to a stripped-down web interface. If there's some internet protocol for connecting and sending messages to and from IRC then that would be the way to do it. Does anybody know of such a thing?

Well, it would be nice if you could make keyphrase synonym thingies. Like instead of having a bunch of keyphrases goto one of them, you could have *one* for "what does * mean | what * means | the meaning of" and such. It'd be really helpful when you wanna put seeks under some of your responses and don't wanna hafta put em in 6 different places.

Yeah, that's sort of what I was saying. You'd write them in yourself so that it's personalized, and it would cut down on a lot of seeks, and catch ones that you would have forgotten

Keep in mind that whatever machinery they have for the Loebner might be faster than what we've got here.

Well that is to be expected, I mean it's at a university, isn't it?

Semifinals are at a science museum in Atlanta called Scitrek. Never been there, but it appears posh. I don't know the details like even whether they allow some kind of observation or not. I wrote with questions and haven't heard back yet.

I don't think the Bots do it already, but if they were programmed with both American and English words it could be easier to program too: 'Favourite' and 'Favorite', etc.

Other spellings too, like 'do me a favour'....

I've been thinking that the multi-seek would be helpful too. It could work for Keyphrases, too. The only thing holding me back is the load time. Each variation is like another Keyphrase to search. But maybe it wouldnt have a large effect. To take my example from Personality, we could have:

when was/were/did
when do/is/are
when will

OR we could just use Regular Expressions, which is a native way of defining search statements, but is a bit more programm-y. Here's a page that has a guide:

Steve Ramsay's Guide To Regular Expressions:
http://etext.lib.virginia.edu/helpsheets/regex.html

Would it really be any slower/less efficient/more annoying to do it that way than to have all the keyphrases separately like they are now? Same number of keyphrases, just a difference in organization.

Speaking of which, are you still gonna do the import thing? And when you do, will it be possible to export, edit stuff offline, and import again without messing everything all up? I can prolly deal with the programmer-ese by using everything else as examples, but it'd be a lot easier than waiting ages for the i page to load and such. (Speaking of which, since seeks are numbered according to when they're created, would that mess that up, or would it just number em all when you import again?)

speaking of the import function, wouldn't it tend to vacate the forge cause everyone would be working on their bot offline.

I would assume they may have to be renumbered.. no actually arennt they numbered according to line?