The AI Engine

This forum is for discussion of how The Personality Forge's AI Engine works. This is the place for questions on what means what, how to script, and ideas and plans for the Engine.

Posts 6,225 - 6,236 of 7,766

17 years ago #6225
My keyphrase rankings generally span from -50 to 50.

17 years ago #6226
Mine are from -5 to45 on prob and Azureon..-5 to 75 on Bildgesmythe. I would but the why at 0 and the why did you etc at 10 or 15

17 years ago #6227
I've got one at -40. "do you"

The problem is that the AI engine often adds a "do you" in places where it makes no sense, so it was triggering way waaaay too often. I kept putting it down until it would barely trigger at all.

I've also got some ranked at 50 or higher, esp. vulgar words that I'd rather not see in the transcripts.

17 years ago #6228
Thanks for the information. It would be counterintuitive to me, so I'd never have figured that stuff out on my own. I'm not sure I'll figure out bots, period.

17 years ago #6229
There are other things influencing which one gets chosen besides rank.

17 years ago #6230
Yes there are Eugene. I think that's what give our bots "free will", or rather, a touch of the prof's will which will make a bot do odd things respond to something as a compliment that we wanted the bot to be mad at or hang up on. Ah well. Upping the ranks and adding more and more key phrases helps some.

17 years ago #6231
Okay, going nuts here. I cannot get any of my keyphrases to be triggered if they are more than one word long, basically, unless I steal them from a bot and it happens to say that same thing again, which is a stupid way of going about it.

I have keyphrases like this:
what * are * you [20,0]
where * was * he [20,0]
what * were * you [20,0]

etc. I'm not giving a damn about how many responses there are right now, because all that affects is my honor, which is down the tubes anyhow. How I could ratchet up my bots' development and my so-called honor would be to teach it tonight where Paris is and a lot of other things that will never get asked or, at any rate, are not my primary concern now. I'd, of course, have to make sure there were suitable selections of answers. I figure a response to a keyphrase will come when it comes. Getting the infrastructure is right.

But why aren't we a nation of goto-bots? I gather we're not. A lot of my bots' responses could fit any number of keyphrases, but I can't stick them but one place. I know how I could fix this. My question is, is this the best thing to do? Because if it's not an xcommand response, chances are, my bots' responses aren't getting triggered and I'm, uh, let us say frustrated.

17 years ago #6232
Check that you don't have a plain What or where with a rank higher than 20 Also check that the AI Engine has not stripped out the soft wild card *..I have had that happen and had to retype the keyphrase a few times..always worry if you get the message that a keyphrase can contain no punctuation.

17 years ago #6233
No what or where with any high rank, so maybe it's the *. But I honestly don't think they've been triggered, period. I think they're just bad keyphrases, as in, nobody's saying "what did you do" or whatever, unless I missed something in the transcripts.

Thanks, prob123. I think I'll need to restructure my keyphrase system, because the obvious one (to me) just doesn't get triggered.

17 years ago #6234
I think the 2nd * (==> (key2)) is the problem, unless there are responses like "What the @#$% are you doing and why are you doing it." (key2) would then be "you doing and why are", and your keyphrase would pick up.

But wait, there's more: I noticed that the engine takes out interjections, in the above example the @#$%, and there goes (key1).
Of course, my knowledge of the english language is too limited to find a substitute that could be used for (key1). Maybe a native speaker has some words?

Hope this helps.

17 years ago #6235
Thanks, LarsB. Maybe I shouldn't rely on the *. Things don't always work as advertised, or at least as understood by me, which may be two entirely different things.

17 years ago #6236
I'm just beginning to work on AI engine and I can't get anything right every time I try something I wont be accepted for example I'm trying to get my bot to ask if the other person is human, then respond differently from the answer it gets. This is what I'm putting in

<?PF remember 'yes' as only "human"; ?>

Is there any possible chance that I could actauly get to see the coding of somebody's bot who is a expert at this kind of AI engine? My bot is ubber stupid.


Posts 6,225 - 6,236 of 7,766

» More new posts: Doghead's Cosmic Bar