KorrBot

  • 88 Replies
  • 4583 Views
*

Art

  • At the end of the game, the King and Pawn go into the same box.
  • Global Moderator
  • ********************
  • Cleo
  • *
  • 4958
Re: KorrBot
« Reply #15 on: April 13, 2018, 09:17:15 pm »
And to think, we haven't even touched on Context or Intent or Inflection, all of which can dramatically affect the meaning or even output/response.
Then there's the double entendre, puns, homonyms, etc... Are you sure you want to do this? Heh!!! ;)
In the world of AI, it's the thought that counts!

*

korrelan

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1010
  • Look into my eyes! WOAH!
    • Google +
Re: KorrBot
« Reply #16 on: April 13, 2018, 09:23:36 pm »
haha... I did say it was a side project... something different to think about when I need a rest from my AGI.

I'll leave the full chatbots to the chatbot gurus.

 :)
It thunk... therefore it is!

*

infurl

  • Trusty Member
  • ********
  • Replicant
  • *
  • 610
  • Humans will disappoint you.
    • Home Page
Re: KorrBot
« Reply #17 on: April 13, 2018, 10:13:12 pm »
English is one of the few modern languages which has been completely analyzed and documented. It was a massive undertaking which took scores of linguists several decades but the results have all been collated and published in one magnificent volume, The Cambridge Grammar of the English Language. It could well be the thickest, most expensive, and most interesting book you ever own.

https://www.amazon.com/Cambridge-Grammar-English-Language/dp/0521431468

If you live outside the USA you can get it more cheaply from Book Depository in the UK.

https://www.bookdepository.com/Cambridge-Grammar-English-Language-Geoffrey-K-Pullum/9780521431460

Note that this is a descriptive grammar. It is all about how people actually write and speak the English language in all its dialects. It is not a prescriptive grammar or style guide written by some snotty academic who thinks they're better than you.

If $400 is too much to spend on a book then there is the student version which is much abridged and much cheaper. I have both and highly recommend them to anyone who is remotely interested in how natural language really works.

*

korrelan

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1010
  • Look into my eyes! WOAH!
    • Google +
Re: KorrBot
« Reply #18 on: April 13, 2018, 10:28:04 pm »
Cambridge-Grammar of the English-Language…

I’ve read that book… terrible story, but at least they explain each word as you read along… hehe… the old puns are the best puns.

No but seriously… it’s an excellent resource.

I regularly use this, though you wouldn’t tell by my spelling lol.

https://dictionary.cambridge.org/

 :)
It thunk... therefore it is!

*

8pla.net

  • Trusty Member
  • *********
  • Terminator
  • *
  • 986
    • 8pla.net
binary
« Reply #19 on: April 14, 2018, 02:00:50 am »
Yet "binary code" is another
of the "levels of abstraction".

SYMBOL TABLE:
A one.               A zero ...
A vertical line.  An oval ...

Symbols have no (single) meaning.


________________________________________
    What you've done is very neat!
    Thanks for  the Korrbot  video!
________________________________________
« Last Edit: April 14, 2018, 02:29:08 am by 8pla.net »
My Very Enormous Monster Just Stopped Using Nine

*

korrelan

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1010
  • Look into my eyes! WOAH!
    • Google +
Re: KorrBot
« Reply #20 on: April 14, 2018, 11:42:06 am »
I've not had long to work on it but now it understands ownership, etc...

Code: [Select]

U: does infurl own a car
B: i'm not sure if infurl owns a car

U: what colour is arts car
B: i don't know what colour arts car is

U: what colour is korrelans car
B: i don't know what colour korrelans car is

U: art owns a car
B: sweet... what else does art own?

U: the car is silver
B: silver is a nice colour.

U: what colour is arts car
B: i think arts car is silver

U: the blue car belongs to infurl
B: so infurls car is blue

U: does infurl own a car
B: of course infurl has a car

U: what colour is infurls car
B: i think infurls car is blue

U: korrelans car is green
B: ok... so korrelans car is green

U: what colour is korrelans car
B: i think korrelans car is green

U: what colour is arts car
B: i think arts car is silver

U: what colour is infurls car
B: i think infurls car is blue

« Last Edit: April 14, 2018, 12:21:59 pm by korrelan »
It thunk... therefore it is!

*

spydaz

  • Trusty Member
  • *******
  • Starship Trooper
  • *
  • 295
  • Developing Conversational AI (Natural Language/ML)
    • SpydazWeb
Re: KorrBot
« Reply #21 on: April 14, 2018, 04:17:13 pm »
English is one of the few modern languages which has been completely analyzed and documented. It was a massive undertaking which took scores of linguists several decades but the results have all been collated and published in one magnificent volume, The Cambridge Grammar of the English Language. It could well be the thickest, most expensive, and most interesting book you ever own.

https://www.amazon.com/Cambridge-Grammar-English-Language/dp/0521431468

If you live outside the USA you can get it more cheaply from Book Depository in the UK.

https://www.bookdepository.com/Cambridge-Grammar-English-Language-Geoffrey-K-Pullum/9780521431460

Note that this is a descriptive grammar. It is all about how people actually write and speak the English language in all its dialects. It is not a prescriptive grammar or style guide written by some snotty academic who thinks they're better than you.

If $400 is too much to spend on a book then there is the student version which is much abridged and much cheaper. I have both and highly recommend them to anyone who is remotely interested in how natural language really works.

While its true that there are a lot of tools to analyse the it has not been completely solved ..... During my research and travel i believe that many other languages would be easier to decipher for an artificial intelligence chat bot. the problem is that NLP is primarily focused in the English domain.   the grammar and meaning of the English language has so many nuances. where as other languages maybe Arabic  grammatically are simpler.
Norm Chomsky actually talks about these issues of the growth of the language, the colloquial usages and meanings. newly created words. in reality there is no secret answer to NLP. the techniques we currently have merely scratch the surface of data extraction.Yet in grammatical analysis only theoretical and traditional paper approaches exist. (linguistics) . 

korrelan... I like the progress;

what else does art own?

It seems as though the conversation was one sided and the question posed by the AI was not responded to... how did it handle that event? how was it sure that the next response was not an answer to the question... obviously it was handled well. Lol. I Like the way it can build the Ownership for each person. !Great!


*

8pla.net

  • Trusty Member
  • *********
  • Terminator
  • *
  • 986
    • 8pla.net
Re: KorrBot
« Reply #22 on: April 14, 2018, 07:12:57 pm »
Korrelan in part said, "AI's using high level languages, they lack the resolution/ tools/ finesse".

I feel that A.I.'s using high level languages benefits the AI at least until computing power catches up one day.

Let's run a test to see how far a high level language can get by itself without A.I.

Test Source code:
Code: [Select]
<?php

$I
=array(
"Does person own a thing?",
"What color is persons thing?",
"Person owns a thing.",
"Persons thing is colored.",
"The thing is colored.",
"The colored thing belongs to person.",
);

$O=array(
"I'm not sure if person owns a thing.",
"I don't know what color persons thing is.",
"What else does person own?",
"I think persons thing is colored.",
"Colored is a nice color.",
"Of course person has a colored thing.",
);

foreach(
$ID as $i){
  
$input=$I[$i];
  
$output=$O[$i];
  
$response AI($input,$output);
  echo 
$response;
}


?>


Test Program Output:
Quote
I: The spacesuit is green.
O: green is a nice color.

I: The red computer belongs to Adam.
O: Of course Adam has a red computer.

I: Does Adam own a computer?
O: I'm not sure if Adam owns a computer.

I: Donna owns a spacecraft.
O: What else does Donna own?

I: What color is Bettys spacesuit?
O: I don't know what color Bettys spacesuit is.

I: Adams computer is green.
O: I think Adams computer is green.

Then the A.I. takes over, and it is less work for the A.I.   In short, high level languages help neural networks look normal.

Live Demo Site:  http://aihax.com/topic/?id=22

The use of any content from this forum post is strictly limited to discussion purposes only.
« Last Edit: April 14, 2018, 08:29:44 pm by 8pla.net »
My Very Enormous Monster Just Stopped Using Nine

*

korrelan

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1010
  • Look into my eyes! WOAH!
    • Google +
Re: KorrBot
« Reply #23 on: April 15, 2018, 10:47:43 am »
Cool.

Yes basic substitution is definitely part of my chat bot, it’s just a bit of fun, I’m not a chatbot guru by any stretch of the imagination…

it’s the AI($input,$output) I’m ultimately interested in.

A high level language can be used to create the illusion of intelligence, but that’s not the same thing as actual intelligence (IMO).  I just want to see how far I can go creating the illusion of intelligence with high level constructs.

It’s just another string I’m adding to my bow, so I can better understand you guys when you’re talking about chatbots and high level language parsing, etc.

Writers block occurs when a person concentrates too much/ long on a particular subject, the GTP networks/ pathways that cover the topic become over used/ they become too sensitive to the internal/ sensory stimulus of that particular problem space.  This then puts the brain at a disadvantage; it stops new ideas forming from other pattern combinations because the subject’s main over used GTP is swamping all other patterns out.  I use side projects like this to stop this happening with my brain.  I like to add to my pattern combinations rather than burn one particular subject into my cortex lol.

 :)
It thunk... therefore it is!

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • ************
  • Bishop
  • *
  • 1927
  • First it wiggles, then it is rewarded.
    • Enter Lair
Re: KorrBot
« Reply #24 on: April 15, 2018, 10:54:48 am »
In reply to korrelan; With my moderation of burn-in:
I work on all of AGI and what will happen soon....I'm not just studying ex. trees.....I think bout everything from self-fires to generation to search to performance to computers to connections to databases to [put anything here].....ok u get it.

I don't implement yet. But we ALL should already be um, scrambled like eggs everywhere, there's so much.......
Emergent

*

8pla.net

  • Trusty Member
  • *********
  • Terminator
  • *
  • 986
    • 8pla.net
Re: KorrBot
« Reply #25 on: April 15, 2018, 02:24:12 pm »
Korrelan,

Suggestion:

Build an original chatbot as a puppet for your AGI to pull the strings.  Korrelan said, "It’s the AI($input,$output) I’m ultimately interested in.".  Let's make it more interesting by introducing a beginner level rudimentary A.I. technique:

Test Source Code:
Code: [Select]
$decisions = array(
0=>array(0,2),
1=>array(1,3),
2=>array(2),
3=>array(3),
4=>array(4),
5=>array(5),
);

Test Run 1 Program Output:
Quote
I: Person owns a thing.
O: What else does person own?

I: Persons thing is colored.
O: I think persons thing is colored.

I: Person owns a thing.
O: What else does person own?

I: Persons thing is colored.
O: I think persons thing is colored.

I: The thing is colored.
O: Colored is a nice color.

I: The colored thing belongs to person.
O: Of course person has a colored thing.


Test Run 2 Program Output:
Quote
I: Does person own a thing?
O: I'm not sure if person owns a thing.

I: What color is persons thing?
O: I don't know what color persons thing is.

I: Person owns a thing.
O: What else does person own?

I: Persons thing is colored.
O: I think persons thing is colored.

I: The thing is colored.
O: Colored is a nice color.

I: The colored thing belongs to person.
O: Of course person has a colored thing.

Or... Ultimately your AGI may be interested in such an entry point to your chatbot A.I. being under its control.

In short, an AGI may pull the strings on a chatbot puppet and report its confidence in the chatbot response.

Otherwise:

Yes.  What you have suggested about building a new chatbot is very cool.

It was fun, my chatbot airing internationally on cable news in prime time!

For fame and glory, often AGI developers casually take part in chatbot A.I.




_________________________________________________________________________________________________________________________________________________

The use of any content from this forum post is strictly limited to discussion purposes only.  Any test source code snippets or test  program output, etc. are just test runs for the sake of a conversation about merging different A.I. techniques.
« Last Edit: April 15, 2018, 03:40:12 pm by 8pla.net »
My Very Enormous Monster Just Stopped Using Nine

*

ivan.moony

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1058
    • Structured Type System
Re: KorrBot
« Reply #26 on: April 15, 2018, 03:40:00 pm »
Korr, may I ask is it possible to get NN to parse a natural language sentence? It has to covert a string of characters into a parse tree. My knowledge on NN reaches extracting a single response from a flat list of possible responses to some input. Simply, upon some stimulus, a single neuron at the top gets lit up. Is it even possible to extract a structured response?

My best shot is to recognize something as a sentence (single neuron). Then we can look back down the tree to see what is the sentence consisted of.
« Last Edit: April 15, 2018, 04:12:32 pm by ivan.moony »
Dream big. The bigger the dream is, the more beautiful place the world becomes.

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • ************
  • Bishop
  • *
  • 1927
  • First it wiggles, then it is rewarded.
    • Enter Lair
Re: KorrBot
« Reply #27 on: April 15, 2018, 04:54:29 pm »
LSTM does that no? It considers past words+order. 4 nodes in layer1 predict layer2 node.
Emergent

*

8pla.net

  • Trusty Member
  • *********
  • Terminator
  • *
  • 986
    • 8pla.net
Re: KorrBot
« Reply #28 on: April 15, 2018, 04:58:15 pm »
Ivan,

Oh, that's a good question for Korrelan.  It may be of interest that an MLP (Multi Layer Perceptron) has some related Natural Language Processing (NLP) applications, such as speech recognition and machine translation.
My Very Enormous Monster Just Stopped Using Nine

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • ************
  • Bishop
  • *
  • 1927
  • First it wiggles, then it is rewarded.
    • Enter Lair
Re: KorrBot
« Reply #29 on: April 15, 2018, 04:59:17 pm »
see attachement for what I mean

it only uses 2 layers to like store many words and sentences
Emergent