So, as many of you know, i am attempting the impossible.
I am making a learning chatbot on a ti 84+ in ti basic.
So yeah, i will post the source code here asap.
So feel free to help me optimize.
Her name is “Suzan”. I have more work to do, but i feel the main loop could use some optimization.
Thanks in advance,
Wave
Here is the link (i cant get it in program mode, so sorry)
http://sc.cemetech.net/?hash=Poj5EIuJOg/Crez941HwaHlnAz17
(I am unfamiliar with how to do this, sorry that I am such a noob)
Suzan has gone from on she shelf to development hell.
Someone borrowed my calculator, and somehow managed to delete suzan “accidentally”.
She was in a group file! How do you even do that!
In retrospect, it WAS my math teacher, so she may have founf out i program somehow....
...but then again, she only deleted Suzan, so that is sketch.
Well, either i start over, or give up.
Any input?
Wavejumper3 wrote:
Suzan has gone from on she shelf to development hell.
Someone borrowed my calculator, and somehow managed to delete suzan “accidentally”.
She was in a group file! How do you even do that!
In retrospect, it WAS my math teacher, so she may have founf out i program somehow....
...but then again, she only deleted Suzan, so that is sketch.
Well, either i start over, or give up.
Any input?
Your math sounds kinda weird. Kind of weird that your teacher just deleted Suzan.
I think that this isn't a program that you should give up on. Keep working on it!
No worries! I have a recent backup of the file! You should be okay!
I am working on Suzan again!
Don't expect too much, I really set the bar quite high...
But she is (technically) working! Its just that she is super slow at the moment.
To make her faster, I would have to redo most of the code, which would suck, but I will try to over the summer.
Also: Among us is still a go! I am trying to juggle these two projects and my schoolwork.
Hope to be able to report something before summer!
How could I imagine how Suzan learns and how you can "use" her?
(I couldn`t look at the code at the link above, because the link seems to not work)
ok....
as womp put it, answers are hardcoded with a layer of obsfucation.
(so all qa pairs are stored as variables instead of directly written)
she learns when you correct her by storing the q/a pair with "fuzzy" matching.
what mean by that is she can generalize a statement, so you can say "say (anything)" and she will respond with the (anything).
i am working on automating that.
after this, i will make it so she can remember parts of a phrase, and store things that will help with contextualizing what is being said.
i will post a new link soon, i edited the cod a bunch.
Here is a link to the code as it stands:
http://sc.cemetech.net/?hash=9WeDsh53DsVmGuVrvNAuyWJHmcRY
(I just optimized it a lot, it is now less than 4000 bytes!)
Error: Unable to fetch project or project list from server.
Error loading project from database
I found that that happened a lot when I tried to make my own sourcecoder 3 projects public; the link broke after a couple of days or when I reloaded, or when I edited the code. I suggest you upload your code to a GitHub repository.
Could someone look at my code and optimize it a little?
You have until next September to do so (that's when I will be on), so have at it!
out of the blue, i am posting here!
I have uploaded a beta test version of suzan to the file archives.
I sure do hope it gets accepted!
Still a really REALLY basic version rn, without any real ability to do too much, but still a proof-of-concept!
I forgot to include source, but it was programmed in axe, and if you REALLY want source i guess you could dm me.
i honeslty only have source on-calc lol.
Cant wait for you all to test her out!
if you have any questions or bugs post them here.
LES GOOO SUZAN WAS ACCEPTED!!!!
once again, any questions, ask them here.
Also if you have any reccomendations or things i should add, please do tell.
I am planning on adding 2 things next: multiple word matching (but in no particlar order) and allowing for a part of the query to be put into the response.
After that, maybe some short-term memory for stuff like names? Idk...
0.2 update: added priority matching.
I am hoping to add multiple appvar capabilities and a pretrained brain next.
Any reccomendations for stuff?
I was interested in trying your beta version, however when I went to try it out, the calculator gave me a syntax error. And when I try to view the file on Sourcecoder or TI-Connect CE, it tells me there's a bad token. If you can fix that, I'd love to actually try the program.
Hi Michael!
It is an assembly program, so it needs the Asm( token before it.
Also, its compiled, so you cant see the source code.
I will add source in the next update! Just warning you though, it is very optimized and very unreadable.
I also feel like i should clarify: this is for TI 83+ and Ti 84+.
Not CE.
It is in assembly. I programmed it in Axe.
Hope that was a good clarification.
Oh, that's my bad. I should've realized it wasn't for the CE.
I would love to see this updated soon! When do you think will be the next one?
Also, the Readme.txt is almost unreadable, and since I found this on the archives, It took me so long to figure out what it was.
Being an Axe programmer myself (not on cemetech), I would like to see the source code. Is there a GitHub post anywhere with it?
Also, is the bot able to build off the conversation? Or is it just going to say stuff off of the one thing that you said?
Maybe you could add an option where its like
HELLO
HI
HOW ARE YOU?
GOOD
YOU HAVE THREE EYES
NO I DON'T
HOW DO YOU KNOW?
CUZ I KNOW I HAVE NO EYES
All of this would be customized by the user, in an interface that would look like.
if last query = (user input here)
and this query = (user input here)
then say (user input here)
Note: This is not Axe code, this is an interface in the program.
These are just my ideas, not demands. One last thing, is HELLO the only thing the chatbot currently reacts with?