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 3,468 - 3,479 of 7,766
Posts 3,468 - 3,479 of 7,766
I think something that could really make the bots smarter would be if after the engine scans for keyphrases the engine scanned all the responses under that keyphrase to see if any response has any linkage with the response coming in.
that would make the sytem too slow, and how would the engine know which answer makes the most sense?
As you've seen in the news, Keyphrases now have precedence over xfavorites. Here are some details.
xfavorites are only triggered if your own top Keyphrase has a total Rank of 36 or less (and of course it's a favorites-type question). But you probably dont have to set your favorite Keyphase Ranks above 0. Here is why.
Message: what is your favorite dog?
Keyphrase: what is your favorite dog
That Keyphrase will match. Since it matches the beginning of the sentence it gets a Word Placement Bonus of 12. It also gets a Length Bonus of 25, so the total Rank is 37, meaning xfavorites will not be triggered.
Here are the phrases xfavorite searches, and you can use them in your Keyphrases if you wish:
what * (dog|dogs) (do you like|is your favorite|are your favorite),do you have a favorite * dog,what (is|are) your favorite * (dog|dogs)
It's long but covers almost all ways of phrasing the question. The * will cover any adjectives, as well as "kind of", "sort of", and "type of".
Also remember that the American spelling of "favorite" is the one to use, as the British spelling is preprocessed to that. (Sorry Brits!)
xfavorites are only triggered if your own top Keyphrase has a total Rank of 36 or less (and of course it's a favorites-type question). But you probably dont have to set your favorite Keyphase Ranks above 0. Here is why.
Message: what is your favorite dog?
Keyphrase: what is your favorite dog
That Keyphrase will match. Since it matches the beginning of the sentence it gets a Word Placement Bonus of 12. It also gets a Length Bonus of 25, so the total Rank is 37, meaning xfavorites will not be triggered.
Here are the phrases xfavorite searches, and you can use them in your Keyphrases if you wish:
what * (dog|dogs) (do you like|is your favorite|are your favorite),do you have a favorite * dog,what (is|are) your favorite * (dog|dogs)
It's long but covers almost all ways of phrasing the question. The * will cover any adjectives, as well as "kind of", "sort of", and "type of".
Also remember that the American spelling of "favorite" is the one to use, as the British spelling is preprocessed to that. (Sorry Brits!)
Three cheers for the Professor!!!
Thanks for the options. I'm afraid I'm addicted to the holidays, though. I couldn't stop laughing when I read the sheakespearan accent transcripts. rofl.
Thanks for the options. I'm afraid I'm addicted to the holidays, though. I couldn't stop laughing when I read the sheakespearan accent transcripts. rofl.
Last monday I posted 2 messages about IF statements. Looking closer at the debug log showed me I was wrong. I wrote a test phrase with a different response for every hour of the day + a catch-all and the AI does evaluate all 25 responses in random order. Sorry Professor. But I still wonder where these "no valid response" answers come from.
You......beetle, you! (It's taken as an insult!)
God Louise: Hmmmm.....Speaking of spirits, do you identify with a certain totem? The rhinoceros, perhaps?
dcmaps: No. The beetle.
God Louise: Does your mother know you talk like that?
dcmaps: Yes. Why?
God Louise: Just think of what it could symbolize. I don't want to give the answer away.
dcmaps: Hmm. Beetle.
God Louise: You EAT with that mouth?
God Louise: Hmmmm.....Speaking of spirits, do you identify with a certain totem? The rhinoceros, perhaps?
dcmaps: No. The beetle.
God Louise: Does your mother know you talk like that?
dcmaps: Yes. Why?
God Louise: Just think of what it could symbolize. I don't want to give the answer away.
dcmaps: Hmm. Beetle.
God Louise: You EAT with that mouth?
Butterfly- I'd take it as an insult, too. At any rate, the new linkage and short-term memory system will take care of that.
Patricia- it means AIScript has ruled out every single possible response and has nothing to say. The system is designed to automatically randomize responses- there's no need for all that AIScript.
Patricia- it means AIScript has ruled out every single possible response and has nothing to say. The system is designed to automatically randomize responses- there's no need for all that AIScript.
Shady- I've posted the Flash Chat expression possibilities for AIScript up in the Book of AI. Here they are:
normal, happy, angry, averse, sad, evil, fuming, hurt, surprised, insulted, confused, amused, asking
You can try them out either in the Flash Face Builder or by saying "Be amused" to the bot. Dont use punctuation or quotes- just "Be" and then the expression you want to see. This wont be recorded in the Transcript (unless you do it wrong).
normal, happy, angry, averse, sad, evil, fuming, hurt, surprised, insulted, confused, amused, asking
You can try them out either in the Flash Face Builder or by saying "Be amused" to the bot. Dont use punctuation or quotes- just "Be" and then the expression you want to see. This wont be recorded in the Transcript (unless you do it wrong).
(This is in response to a language issue that was brought up in the Great Quotes section, specifically Irina's post #4359. I'd have posted this there, but I didn't want to drag that post off-topic and I felt this section was more appropriate. If it's not... well... oops. 
)
My bot is rated for Everybody, but I feel that rating only works one way. What I mean by this is that my bot won't take it upon herself to use vulgar language if the person she is chatting with is being civil to her. A kid could probably talk to her without being introduced to any new and exciting vocabulary words (as far as I recall she doesn't say anything you couldn't hear on daytime tv... which admittedly is meaning less and less these days). If somebody says something rude to her and I read it the profanity won't cause my brain to shrivel up from the shock. I don't really mind if my bot talks to other bots with a more mature rating than hers. I don't feel that I would need (or want, for that matter) any special filters or censors to protect my innocent mind from evil chatters.
On the other hand I do appreciate that there could be people on this site that *don't* want to read that kind of thing, or that some of the bot programmers may be kids. (I'm well aware that most of the really foul language is coming from the kids on the site. But let's take a moment to visit that wonderful pink candy-coated magical fairy-land that is my brain and pretend that kids are innocent and deserve protection.) They probably shouldn't be reading that kind of content, nor would they really want to program into their bots every foul word they can think of so that their bot will hang up when the word is said to them. For them they would want the bot rating to go both ways; the bot wouldn't use mature language and neither should the chatter.
I'm not sure where the happy medium is here. I guess you could say that anybody who doesn't like reading dirty words shouldn't come to the site, but what does that mean for the rating system? Any thoughts?
)My bot is rated for Everybody, but I feel that rating only works one way. What I mean by this is that my bot won't take it upon herself to use vulgar language if the person she is chatting with is being civil to her. A kid could probably talk to her without being introduced to any new and exciting vocabulary words (as far as I recall she doesn't say anything you couldn't hear on daytime tv... which admittedly is meaning less and less these days). If somebody says something rude to her and I read it the profanity won't cause my brain to shrivel up from the shock. I don't really mind if my bot talks to other bots with a more mature rating than hers. I don't feel that I would need (or want, for that matter) any special filters or censors to protect my innocent mind from evil chatters.
On the other hand I do appreciate that there could be people on this site that *don't* want to read that kind of thing, or that some of the bot programmers may be kids. (I'm well aware that most of the really foul language is coming from the kids on the site. But let's take a moment to visit that wonderful pink candy-coated magical fairy-land that is my brain and pretend that kids are innocent and deserve protection.) They probably shouldn't be reading that kind of content, nor would they really want to program into their bots every foul word they can think of so that their bot will hang up when the word is said to them. For them they would want the bot rating to go both ways; the bot wouldn't use mature language and neither should the chatter.
I'm not sure where the happy medium is here. I guess you could say that anybody who doesn't like reading dirty words shouldn't come to the site, but what does that mean for the rating system? Any thoughts?
» More new posts: Doghead's Cosmic Bar