-=learning from text files =-

[fracty]

hello all

i have seen the feature "learning from text file."

1-what can hall learn by himself ?(i mean infering on new things that we did not taught him )
2-is there any plugins or modification related to this ?
3-i'm not talking about Q/Answer for this is just building a libray, right ?
4- did you try to put things like fuzzy logic and neural net in this soft? i do not know much now but when i was student i heard of these topics.

to go back about text file learning:
i do not understand this feature. where is the info stored ? can we access it for review ? thanks for guiding me

i have read that people have made hall digest information but can he extract it and learn new things from this information.
that would be interesting.

Best,[^]

[Bill DeWitt]

quote:Originally posted by fracty
1-what can hall learn by himself ?(i mean infering on new things that we did not taught him )

Nothing.

[fracty]

quote:Originally posted by Bill DeWitt

quote:Originally posted by fracty
1-what can hall learn by himself ?(i mean infering on new things that we did not taught him )

Nothing.

well it is a pity ! well, we will wait for things to improve then.
i have jsut discovered a plugin for text file in ULF or something like this, wait for validation and download...will test this soon.

[Bill DeWitt]

quote:Originally posted by fracty
well it is a pity !

On the chance that you mean something other than what you asked, there is a function that can do simple inferences. If A = B and B = C then A = C.

But it cannot make up new information based upon observation or experimentation. It simply compares and stores strings of text, then uses a fairly wide set of rules to deliver them in a conversational way.

I have worked on ways for Hal to search the web for information about subjects brought up in conversation, but did not get very far before running into huge parsing problems.

[fracty]

thanks fo lettin me know of your work.
well, i'm supporting you in your research...

for the "first degree inference" i did read this feature and what was interesting was just to go beyond this simple rule...

and may be to crosscheck the proposition for logic so it does not contradict itself or fall into logic trap or false conclusion.

so when it found a new fact, no previous conclusion/observation must contradict the "discovery" or more supple-more real life,the fuzzy logic manner: the more it is "improbable"  the less it will catch and use the new fact...

that would be great !