The AI Engine

On the favorites: Someone asked my bot about his favorite color long before I expected to have to deal with it, but the randomizer actually _did_ come up with his real favorite color! (Or as close to it as makes no difference). Of course, I don't want to trust to luck in the future, but I think I've got a workaround for that now, just by changing the xphrase to something that doesn't call up an answer.

On the wildcards: I'm not sure I trust (subj) to be specific enough -- I've already seen bots misusing it aplenty. And what I really want is a within-word wildcard rather than a within-sentence wildcard. For instance, I want to be able to use 2* to mean any number in the twenties, without reference to what follows the space after the *. Or the "reverse plug-in" I described would work even better. But if it's not possible, it's not possible, I suppose. I'll live anyway. :-)

Thanks for the answers, guys.

Okay, new question! My bot has been -- quite uncharacteristically -- greeting guest chatters by calling them "friend". I thought at first that it must be simply plugging in "friend" when it didn't have a (name) to access, and I "fixed" it by taking (name) out of the xhello responses. But I'm going through the transcripts now, and it _is_ using Guest instead of name, the way I expected. So now I'm confused about where the "friend" is coming from. Help?!? My bot likes being antisocial! (Oh, and _that_ subject.... both my bots seem to decide they like people rather easily, with little provocation. How do they decide whether they like someone or not, if they haven't talked long enough to invoke any of the emotion-ranked keyphrases?)

The bots use "friend" for the (name) plugin when talking to guests. If chatting to a member, they say the member's name.
Ex "Hi friend" if I wasn't logged in, or "Hi Shadyman" if I was logged in

The beginning of the conversation is the first line, no matter who says it.

Okay, thanks. Then where does the Guest come in? (As in, where the transcript says "Hello, Guest")

I haven't used guest chat, so I don't know how it works.

I think that "Guest" is used mostly in non-contest times, "friend" is used to suck up to the judges

Ah, of course. Sucking up... now that's an explanation I should have anticipated. *g*

Bedawyn, I know what you mean with (subj) plug-in being misused. I think The Professor was trying to get a short subject plug-in happening at one stage which would only use the one noun or verb phrase immediately following keyphrases, but it's hard to get going.

As far as the reverse plug-in idea, I think it would take a significant amount of recoding that would seriously delay response time. At the moment, the responding bot only deals with the words it has been given. It doesn't know that the speaking bot has used a plug-in in a sentence.

...and bots do talk to humans, who are basically one big plug-in....

Well, the idea would be not to have the bots simply recognise if another bot is using a plug-in, but to see if a word is of a specific type. Thus, you could have the keyphrase
Do you like [foodType]?
== I don't eat, sorry.
Which would be called if the human said "Do you like pastrami?" or something like that.

Right, but that would require a huge amount of preprocessing. Remember when chats were going at snail speed?