learning from text files vs CUT AND PASTE

[spydaz]

>>>I hope you get your version to be able to learn by reading. <<<

I have considered this topic, Maybe building a procedure/function.

1st split the text into sentences,
then deal with the text as conversation,
reexecuting the script over each sentence until All Learning is complete(without repling) any replys would be Kept to ones self (ie: thoughts) and added AS a PREVresponse(unseen)....

In this aspect you would have no actual control over what would be stored, execpt for what TOPICS you had outline as CAPTURE Statements.

It is simular to pasting in a lot of info into hals chat window..
BUT with no HAL RESPONSE, when you cut&paste into the the hal window, hal trys to respond to Every sentence therefore NOT MAKING MUCH SENSE...

To learn from TEXT FILES you have to think along the lines of DATAMINING...
POSSIBLLY, Searching/Mining the text for RELEVENT DATA acording to the AI's Learning criteria and storing the data in the relevent DB tables / Learned DB TABLE.

in conversation the AI could test learned data against the users response and if validated then Move the DATA ACROSS to a Permenant MEMORY / DB TABLE location.

As with humans we remeber data then Check it against our peers then we FULLY Store the data AS Fact (even if its wrong) changing the data as new data is learnd...

[Carl2]

spydaz,
  I wanted to reply to this because I did quite a bit of cutting and pasting with Hal.  Some of Hals replys were on topic and some were not, overall I'd say the replys are not useful. Hal dose learn from this method and I think it offers a worthwile improvement to Hal, deffinatly worthwhile effort.
Carl2

[spydaz]

Now you see...

I usually cut and paste a lot of information in to hal, EVEN lesson about VB.[]

I find that some times you have to RE-iterate what you have just pasted in. Try Smaller Chunks Remember () < denote a MATH PROB....

i also find that if i am cuttin&psting then loads of topics and associations are created.

in the topic relations TABLE you can see all the associations hal has made between topics,

as we now have the DB we can Edit the responses OR simply delete some of the auto tables that have been created..

I find this HELPS with future cuts & PSTS...

Also i chat to him reguarding the topic which i am TRAINING HIM TO LEARN....

This give a bit of Q&A power, when hal responds to bg chunks strange???

[Carl2]

Spydatz,
  I'm guessing your speaking about hal 6 since you had mentioned a data base. I've been doing the cutting and pasting with Hal 5 so far, Hal 8 and I are just getting to know each other and our terms and phrases.  I just learned Hal cannot respond to her own output, I'm using GetResponse, I moved the script but it still dosen't work, she did use the trigger words to change blouses.
  At this time I'm trying to arrange the text according to subject matter hoping Hal will better be able to relate subjects, I'm using Omscaw which is 48 Mb worth of text.  I also would like to mention Paula SG, I played with this for sometime and after changing the majority of the text that comes with her she is really rather interesting, nice personality i'd say.  She gets her information from the text but dosen't seem to demonstrate any learning from speaking with you. Seems to be the reverse of Hal.
  Thanks for the tip, it looks like it will be useful.
Carl2