AI that Learns!!!
Root / Submissions / [.]
Version:2Size:2KB
This is an AI/Chatbot That Learns!
At 42 lines long, it is shorter, AND smarter than 12me21's! (with all due respect)
note: this program can run in any folder, as it stores its knowledge in its own source code, by editing itself.
This one's interesting! The thing about this one is the algorithm is known to the user. Whenever you reply a phrase, it immediately gives the response you gave it last for that phrase. This is a correct, and very valid approach.
My request is that you compare every input with a Levenshtein distance to every memory input, and if the Levenshein distance is less that a certain threshold then it will consider the input the same as that memory item. So maybe you have a typo or forget a question mark, it will still find the proper response instead of creating a new one.
You should consider avoiding the save screen every single response, it can get a little tedious for the user.
Very nice program!
Replying to:
Simeon
SimeonThis one's interesting! The thing about this one is the algorithm is known to the user. Whenever you reply a phrase, it immediately gives the response you gave it last for that phrase. This is a correct, and very valid approach.
My request is that you compare every input with a Levenshtein distance to every memory input, and if the Levenshein distance is less that a certain threshold then it will consider the input the same as that memory item. So maybe you have a typo or forget a question mark, it will still find the proper response instead of creating a new one.
You should consider avoiding the save screen every single response, it can get a little tedious for the user.
Very nice program!
That is what mine was supposed to do, but there's a bug that causes it to choose the wrong response
Replying to:Simeon
SimeonThis one's interesting! The thing about this one is the algorithm is known to the user. Whenever you reply a phrase, it immediately gives the response you gave it last for that phrase. This is a correct, and very valid approach.
My request is that you compare every input with a Levenshtein distance to every memory input, and if the Levenshein distance is less that a certain threshold then it will consider the input the same as that memory item. So maybe you have a typo or forget a question mark, it will still find the proper response instead of creating a new one.
You should consider avoiding the save screen every single response, it can get a little tedious for the user.
Very nice program!
I know what Levenshtein distance is, and how it works, but how to make the algorithem eludes me.
Replying to:Simeon
SimeonThis one's interesting! The thing about this one is the algorithm is known to the user. Whenever you reply a phrase, it immediately gives the response you gave it last for that phrase. This is a correct, and very valid approach.
My request is that you compare every input with a Levenshtein distance to every memory input, and if the Levenshein distance is less that a certain threshold then it will consider the input the same as that memory item. So maybe you have a typo or forget a question mark, it will still find the proper response instead of creating a new one.
You should consider avoiding the save screen every single response, it can get a little tedious for the user.
Very nice program!
Also- Thanks, I unwatched this, because I thought that no one was interested. (sorry) I will re-watch it
Replying to:
Lacks
LacksI am impressed. The chat has an AI at the moment and this, so far, is much smarter because it isn't corrupted by other people. You can really focus the bot in on one concept. I am attempting to teach it basic addition and subtraction. Wish me luck.
Thx!! :)
Replying to:
amihart
amihartRather than having it save a file every single time which prompts the annoying dialog box, why not have it temporarily store the data into a slot until the user decides to save it?
Good Idea!! ;)
Replying to:amihart
amihartRather than having it save a file every single time which prompts the annoying dialog box, why not have it temporarily store the data into a slot until the user decides to save it?
Or use the clipboard to temporarily save it
That's what I do to freak the user out, they think nothing saved, but it did (till they restart SmileBASIC)
Is this a big report? If so, please describe the bug. If not, oops.
no, it just doesn't take into account the previous messages, so you can't keep a real conversation past hi and hello.
Yeah, I know, but learning would be much slower if I added what I call 'level 2' or above learning
Perhaps I might get around to making a program that lets you talk to the bot, and when it is sufficently smart convert it to "level 2 memory", but disable learning (it would be 1.5 times more complex)
it's the simplest but moderately effective solution
Replying to:amihart
amihartRather than having it save a file every single time which prompts the annoying dialog box, why not have it temporarily store the data into a slot until the user decides to save it?
Or until they get one of your programs, and it erases the clipboard, much to the dissapointment of the user.
I dont think Ill do that.
But, maybe, If I made it learn AND keyword-based. That would be more complicated, but also smarter.
I'd have to use an algorithem to reconize a keyword, though. THAT would be hard
The key was removed. Sorry.
Thank you for notifying. The page has been updated.