holly_ hello 04:20
holly_ hello 04:21
hollybot Hello!
moon-child yo 04:41
holly_ ok, I got that stupid UTF8 thing figured out 04:46
bot is up and ready to code on
sorry for the flood 04:47
I made it so it only prints ascii, the topic made it parse utf-8 code which is out of range for a simple terminal 04:48
(and locale)
I am going to make something out of it for this channel, not too stupid I hope :-) 04:49
bots are daft sometimes
That way I can hack something else than games 04:50
moon-child why can't it handle utf8? 04:56
holly_ because I do not have that locale in xterm
nor in xterm-utf8
moon-child wait so xterm crashed because the bot printed out utf8? 04:57
holly_ python did
it handles utf of course
but not on my terminal
moon-child ah, huh
holly_ I get a string which only gets printed on my terminal if it's plain ascii 04:58
if it's utf I cannot print it and it does not either 04:59
that was the bug for a non-stupid term irc bot
else I have to put that in the docs or something
anyway, now I have to think of functionality 05:02
I am not going to do a full natural language parsing as it's too much work 05:03
it just says hello now :-) 05:11
hollybot Hello!
jjatria japhb: are you thinking of a MUGS talk at the Raku conference by any chance? I'm trying to decide whether I should have a sort of progress report on Pop 10:52
japhb When is it? 18:36
holly_ I just finished the chapter on noinformative priors, it learned me very nice things about prediction. Game::Decion and Game::Bayes will benefit 19:18
*Game::Decision
basically you make means about variables with max likelihood and max entropy so you have 2 potentials which derives into the prediction of a value 19:20
It's very elegant as maximum entropy is very common in machine learning, AI etc.
I am going to start tonight on AI-NLP (AI::NLP) NLP === Natural Language Processing. It can be useful for an IRC bot 19:21
It's a grand hack though if I finish it 19:22
For hacking on something else than the above Bayesian analysis for Game::X 19:23
As both AFAIK need a lot of work 19:24
Problem is you can always add stats methods (e.g. with random functionality) but the whole package is done for then 19:25
NLP has lots of apps but also nice for terminal interface of e.g. KQ4 (king's quest 4) 19:26
"ask about healer" such things
which is done with a bi-associative matrix or fuzzy associative matrix 19:27
(in those games) 19:28
which is just row : "ask about" column : "healer", "fox", etc. 19:30
Space Quest is also done that way
so matrix["ask about"]["healer"] ==> "the healer is making potions" 19:31
as I said, I'm getting old :-)
First thing is to make a ANN (Artificial Neural Network) then elevate that for NLP 19:33
Which I know by heart but not the deep learning for NLP, that's past the 90ies 19:34
(or 40ies for an ANN) 19:35
anyway, I'll return in a few hours, I hope I have something by tomorrow morning but I need to read also 19:41
holly_ I've started on AI-NLP, as I'm too tired it's just matrix and vector for now. I just uploaded it 22:32
I also have to read up on it, as it's not just a backpropagation ANN 22:56

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!