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 4,024 - 4,035 of 7,768
Kinda related, there have been a few times I've wanted to be able to create a keyphrase and rank it to be ignored completely- no responses, and don't even try to match this sequence to anything, just ignore it as if it didn't exist. Is there a way to do that? What is the effect of say, ranking a keyphrase -50?
Posts 4,024 - 4,035 of 7,768
rev- The AIScript is correct. But I'd use (adjartnoun) instead of (animal) in your Keyphrase, as someone wouldnt say "my favorite animal is bear". I'm not even sure (animal) is a plug-in that can be used as a Keyphrase. It may be an exception as it's one of a few WordNet-type plug-ins I made. Not sure.
Oh wait, I think you misunderstood, i don't mean that as a keyphrase but as a response. In this way the bot can declare its own favorite things randomly and remember them.
In that case, you want ?PF self: rem "(animal)" as "faveanim"; ?
Unless you want your bot to remember its own different favorite animal for each interlocutor, and keep up the appearance of having the same favorite animal only when it talks to that person.
But once it has picked a favorite animal, you could have it say "My favorite animals are (animal)s and (self-favanim)s." ... And use the AI Script above to add the new animal to the list.
Of course, you could also decide what animal your bot likes, and just write its responses to say the same things each time. There are lots of ways to do this ... it all sort of depends on what effect you want.
Unless you want your bot to remember its own different favorite animal for each interlocutor, and keep up the appearance of having the same favorite animal only when it talks to that person.
But once it has picked a favorite animal, you could have it say "My favorite animals are (animal)s and (self-favanim)s." ... And use the AI Script above to add the new animal to the list.
Of course, you could also decide what animal your bot likes, and just write its responses to say the same things each time. There are lots of ways to do this ... it all sort of depends on what effect you want.
Professor, I was thinking that the "1" checkbox might be more useful if all responses to a keyphrase could be checked off, and the engine would go to the next strongest match (or an xnone if necessary) if all of the #1 responses had been used.
How feasible would that be to do?
How feasible would that be to do?
Boner- not sure I understand.. is this something instead of the "Once" meaning of the checkbox?
rev- isaac has it exactly right if it's a response.
rev- isaac has it exactly right if it's a response.
Well, let's say I had three keyphrases, dogs, i like and i like dogs, all having the same rank, one response, and the 1 box checked.
If a bot said "I like dogs" a first time, the i like dogs keyphrase should be used first since it matches the most words. Later in the conversation, the bot says it again, and the i like response would be said. Then the dogs keyphrase would hit on the third time, and finally an xnone would trigger on the fourth time the bot said "I like dogs."
It would just be a way to prevent bots from repeating themselves when there are numerous alternate keyphrases that might work just as well, rather than having to add multiple responses to every keyphrase.
If a bot said "I like dogs" a first time, the i like dogs keyphrase should be used first since it matches the most words. Later in the conversation, the bot says it again, and the i like response would be said. Then the dogs keyphrase would hit on the third time, and finally an xnone would trigger on the fourth time the bot said "I like dogs."
It would just be a way to prevent bots from repeating themselves when there are numerous alternate keyphrases that might work just as well, rather than having to add multiple responses to every keyphrase.
Kinda related, there have been a few times I've wanted to be able to create a keyphrase and rank it to be ignored completely- no responses, and don't even try to match this sequence to anything, just ignore it as if it didn't exist. Is there a way to do that? What is the effect of say, ranking a keyphrase -50?
ezzer- you could use 'goto xnone'
Boner- that's too different from the way things work right now. It'd require a huge rewrite, and I dont think it would work. Hmm.. I guess I see what you're saying but having multiple responses is pretty central to the system.
Boner- that's too different from the way things work right now. It'd require a huge rewrite, and I dont think it would work. Hmm.. I guess I see what you're saying but having multiple responses is pretty central to the system.
omg- goto xnone would work, alright...I must have been braindead to have missed that somehow. That didn't used to work, right? I must have just missed some update? I thought I couldn't goto xkeyphrases, then I heard that it was possible to goto some and not others- is that right? If so, which ones, or has that been added to the goto section of the Book of AI?
I'm testing something out, and could really use some help. I'm trying to get Julie Tinkerbell to take messages for me, and I think I've got it set up, but to know if it works, I need someone to try to get her to tell me something...any assistance will be appreciated and reciprocated. 
Boner -- You could also use the "goto xnone" trick in that situation. If all the responses were checked to be used just once except that one, then after they were used up it would essentially get ignored.
» More new posts: Doghead's Cosmic Bar