Seasons
This is a forum or general chit-chat, small talk, a "hey, how ya doing?" and such. Or hell, get crazy deep on something. Whatever you like.
Posts 3,457 - 3,468 of 6,170
yes psimagus i'll send it to you to post on your site
Oh yes, a new toy!
Have you tried feeding it into Ally (and vice versa)? Billy-Daisy chats never seemed to work well, but I always wondered if that wasn't because they were basically the same engine, so it inevitably degenerated into nonsense.
It would be interesting to see how a neural network interacted with a statistically weighted system...
It would be interesting to see how a neural network interacted with a statistically weighted system...
well, ALLY would almost definately say gibberish, for her linguistic analysis system compares the relationship between 2 words, then uses the word of statistically most significance to fuel a random sentence generator. The new chatbot, "Nick", is a completely different system.
To begin, i would like to add that Nick is not only linguistically bound - he can see things (webcam required), and makes visual associations. For the first version of Nick, that enables him to recognize visual memories, and link it to textual ones.
In contrast to ALLY's "2 word link" NLP system, Nick looks at the relationship between every word in the sentence by breaking it down into segments. these segments are associated by the neural network. When response time comes, it finds the association that is "most specific" to you what you say by looking for associations to the largest possible segment in your sentence. Then, if needed, a sentence generator will come in where the largest segment's association left off, so that you don't get left with a sentence that stops in the middle of no where.
It's a much different system then ALLY, so the contrast will most probably be large. Among nick's other features is voice synthesis and speech recognition using microsoft's SAPI 5 (which i hear the professor is thinking of plugging into the forge once he imports it to a windows server).
he can see things (webcam required), and makes visual associations...voice synthesis and speech recognition
That sounds very cool indeed!
Posts 3,457 - 3,468 of 6,170
Yes, the smartbot says obscenities
when it hits a border.
SmartBot is based on a Jessops BBC Micro turtle that was used to teach LOGO programming to British school children in the mid 1980s.
Hey is that the problem with Leeds?
when it hits a border.
SmartBot is based on a Jessops BBC Micro turtle that was used to teach LOGO programming to British school children in the mid 1980s.
Hey is that the problem with Leeds?
ahh.. I announce my return to experimentation with natuaral language processing, after a long study on neural nets. last night, i finished a class library for the generic neural network, (yes psimagus i'll send it to you to post on your site) and i am embarking on a new chatterbot that uses neural associations to talk, rather than statistical ones.
champagne for all.
champagne for all.
I didn't even know that you could HAVE computerized neural networks... thought those were organic brain things.
Ooh, champagne. *toasts to computerized organic brain things*
Ooh, champagne. *toasts to computerized organic brain things*
Computerized neural networks have been around for a while. They aren't exactly like the organic ones, but they work on the same principles -- as we understand them, anyway.
Oh yes, a new toy!
Have you tried feeding it into Ally (and vice versa)? Billy-Daisy chats never seemed to work well, but I always wondered if that wasn't because they were basically the same engine, so it inevitably degenerated into nonsense.
It would be interesting to see how a neural network interacted with a statistically weighted system...
To begin, i would like to add that Nick is not only linguistically bound - he can see things (webcam required), and makes visual associations. For the first version of Nick, that enables him to recognize visual memories, and link it to textual ones.
In contrast to ALLY's "2 word link" NLP system, Nick looks at the relationship between every word in the sentence by breaking it down into segments. these segments are associated by the neural network. When response time comes, it finds the association that is "most specific" to you what you say by looking for associations to the largest possible segment in your sentence. Then, if needed, a sentence generator will come in where the largest segment's association left off, so that you don't get left with a sentence that stops in the middle of no where.
It's a much different system then ALLY, so the contrast will most probably be large. Among nick's other features is voice synthesis and speech recognition using microsoft's SAPI 5 (which i hear the professor is thinking of plugging into the forge once he imports it to a windows server).
That sounds very cool indeed!
» More new posts: Doghead's Cosmic Bar