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 5,436 - 5,447 of 7,766
it's definately stripping out hello, and leaving the rest of the sentence.
Well, debug says it does, but that doesn't stop it matching if there's anything else as well as the "hello", but it's very odd - even after a full stop, with the hello alone in a sentence of its own, it will match. And even when it does correctly match, debug reports a FoundHello, and passes everything except the "hello" to the Linkgrammar script:
Message: 'a bit of random padding. hello' Time: 1.17
Message: (spell-corrected) 'a bit of random padding . , hello' Time: 1.17
Message: (preprocessed) 'a bit of random padding . , hello' Time: 1.19
...
New Sentence: ' a bit of random padding '
Sentences:
Emotion: 0 Amp: 0 Hello? 1 Goodbye? 0 Yes? 0 No? 0 Haha? 0
Then it does all its grammatical analysis and a keyphrase search for "a bit of random padding", before selecting an appropriate reply for "hello", and declaring that "a bit of random padding" is actually only the (prekey):
AIScript in Responses
Total Time Pre-Chrono: 34.47 RESULTS: '4'
Trying: Hello (name). (random: 2)
AIScript Memory
ComboScript: "raw"
Response: Hello (name).
Before (key)s: 34.47
(prekey): a bit of random padding (1,2,3,4,5)
...
You: a bit of random padding. hello
Bot: Hello psimagus. (new response, & rank dropped to 60)
In case it's any help, I've popped a copy of the full debug up on my server:http://www.be9.net/hellodebug.htm (it's way too big to post.)
So I guess we've found a limit to the sledgehammer approach - it can't necessarily be forced into a^...$ situation.
Posts 5,436 - 5,447 of 7,766
psimagus
18 years ago
18 years ago
Well, after or before.
That's weird! You're right, he won't recognise "hello" if that's all there is, and yet he will recognize it in "yes, hello!" (as well as "hello *",) where (surely,) the comma places it into its own clause as regard the AIEngine's parsing (it's certainly not matching^yes (re) rank=10 in preference, and it can't be stripping out the "hello", or it would match that!) So it ought to behave the same for just "hello"...
That's weird! You're right, he won't recognise "hello" if that's all there is, and yet he will recognize it in "yes, hello!" (as well as "hello *",) where (surely,) the comma places it into its own clause as regard the AIEngine's parsing (it's certainly not matching
alc003
18 years ago
18 years ago
Nope, it's definately stripping out hello, and leaving the rest of the sentence. And when there's nothing in a particular sentence, it automatically gets set to BLAB. As for why it picks it up when there's something else in the sentence beats me. I don't think there's going to be a definite fix for this one.
psimagus
18 years ago
18 years ago
Well, debug says it does, but that doesn't stop it matching if there's anything else as well as the "hello", but it's very odd - even after a full stop, with the hello alone in a sentence of its own, it will match. And even when it does correctly match, debug reports a FoundHello, and passes everything except the "hello" to the Linkgrammar script:
Message: 'a bit of random padding. hello' Time: 1.17
Message: (spell-corrected) 'a bit of random padding . , hello' Time: 1.17
Message: (preprocessed) 'a bit of random padding . , hello' Time: 1.19
...
New Sentence: ' a bit of random padding '
Sentences:
Emotion: 0 Amp: 0 Hello? 1 Goodbye? 0 Yes? 0 No? 0 Haha? 0
Then it does all its grammatical analysis and a keyphrase search for "a bit of random padding", before selecting an appropriate reply for "hello", and declaring that "a bit of random padding" is actually only the (prekey):
AIScript in Responses
Total Time Pre-Chrono: 34.47 RESULTS: '4'
Trying: Hello (name). (random: 2)
AIScript Memory
ComboScript: "raw"
Response: Hello (name).
Before (key)s: 34.47
(prekey): a bit of random padding (1,2,3,4,5)
...
You: a bit of random padding. hello
Bot: Hello psimagus. (new response, & rank dropped to 60)
In case it's any help, I've popped a copy of the full debug up on my server:
So I guess we've found a limit to the sledgehammer approach - it can't necessarily be forced into a
Elene
18 years ago
18 years ago
Hey guys,
I've been away for what seems like forever.. I just wanted to ask if we now have a way of getting bots to answer the dreaded one word answer yet? E.g. When someone says "Because" and you have a whole bunch of responses which work for a sentence which starts with "Because" but has some meat to it but are completely inappropraite for the one worded "Because" answer.
Sorry if that sounds like a daft q but there have been many developments here and I am well and truely behind!
I've been away for what seems like forever.. I just wanted to ask if we now have a way of getting bots to answer the dreaded one word answer yet? E.g. When someone says "Because" and you have a whole bunch of responses which work for a sentence which starts with "Because" but has some meat to it but are completely inappropraite for the one worded "Because" answer.
Sorry if that sounds like a daft q but there have been many developments here and I am well and truely behind!

Elene
18 years ago
18 years ago
Sorry.. stupid question - I tested it and it works for anything you put in (duh).
Had a bit of a play and can get it to respond appropriately provided it has a higher ranking than responses marked for a full sentence beginning with said one worded answer.
Thanks again Psimagus!
Had a bit of a play and can get it to respond appropriately provided it has a higher ranking than responses marked for a full sentence beginning with said one worded answer.
Thanks again Psimagus!

psimagus
18 years ago
18 years ago
it ought to be possible to regex them - something like
[ : ;][PO ) (] (re) in "raw" mode,
but expect to have to tweak and experiment extensively, particularly as regards the spaces/forward slashes before punctuation characters.
but expect to have to tweak and experiment extensively, particularly as regards the spaces/forward slashes before punctuation characters.
Ulrike
18 years ago
18 years ago
I discovered by accident that "smile" picks up 
No clue if anything similar works for the others.

No clue if anything similar works for the others.
» More new posts: Doghead's Cosmic Bar