Difference between revisions of "Tutorial for NLP"

Revision as of 02:35, 12 January 2016

Eskarn's tutorial for NLP

Adding Words and Sentences to the NLP

Natural Language Processing

(anywhere i say npl i mean nlp)

This is an advanced tutorial and probably wont have pictures

How im going to do this tutorial is explain the theory on how the NPL works then cheat after you know how the NLP works you will understand why im gonna cheat

i will be using an unpacked dungeon How to unpack

you will need to have the brain enabled to see the responses and debug mode is always good

settings ini in main folder

runtime
{
  start_mode   = FAST;
 
  debug_sys    = BASIC;
  debug_render = NONE;
  debug_phys   = NONE;
  debug_brain  = ENABLED;
 
  save_key     = "";
}

pack_dungeon\ai\nlp

you will see 59 text files each text file handles different parts of the NLP

lets start with something easy

open up

nlp_actor.txt

lets take the top set

#actor = <you>   : you

#actor is the variable

<you> is the word said by the player

: you is the result

this one is a little bit different

#actor = <#name> : &0

#actor is the variable

<#name> is another variable set below so we are seeing if any of those names have been said

: &0 this is saying the first #variable so #name

if the player said tiffany it would return tiffany

if they said aiko it would return aiko

lets move on to nlp_answer.txt

#answerverb = <#actor can>          : yes,&0,can,present

@answerverb is the variable

<#actor can> remember the #actor from before so if the player says any of the #actor words aswell as can so "you can" "we can" "Samantha can" etc

: yes,&0,can,present this will result as

present.can.#actor.yes with #actor being what was said

well this seems easy

nlp_actor.txt

nlp_answer.txt

nlp_attr.txt

nlp_greet.txt

nlp_object.txt

nlp_question.txt

nlp_verb.txt

are the ones that have mostly single words and then a single responses

lets move on to a harder one

and we will see how this all comes together

open nlp_ac_feel.txt

now you will instantly see it becomes a lot harder

so lets take the first response

#nquestion1 = <how * #qis   * #actor>    : ^qn_how_ac_feel(&1,present)

#nquestion1 is the variable

<how * #qis * #actor> #qis is set in npl_question.txt and #actor we know

so #qis can be "is" "are" "am"

"how is he" "how are you" "how am Samantha"

the * mean anything so the sentence could also be "how is [that] Samantha"

"how [nice] is [this] Samantha" "how [goobly] are [pixydustinface] Samantha"

some of these don't make sense but that's why there is many responses

: ^qn_how_ac_feel(&1,present)

^qn_how_ac_feel the ^ means the variable is not set but do it anyway otherwise there would be an error saying that its not set

the qn_how_ac_feel is used later and it would be

#actor.present

now open

nlp_result.txt

and find

// Actor feel

here is where the qn_how_ac_feel is used

qn_how_ac_feel(*,*)                     : get (form),&0(actor),feel(attr),&1(tense)

qn_how_ac_feel(anything,anything) so basicly #actor.present

: get (form),&0(actor),feel(attr),&1(tense)

e=[get.form,#actor.actor,feel.attr,present.tense])"

#nquestion1 = <how * #qwas  * #actor> : ^qn_how_ac_feel(&1,past)
#question  = <#nquestion1>    : question(&0)
#sentence  = <#question> : &0

then it all comes down to #final in npl_sentence.txt

#final = <#answer #name   * #sentence *>  : &1(&0,&2)

Every result has to start with the actor who the sentence is addressed to.

So #final = <#name #greet *> : &0(&1)

Can parse "Saiko Hello" -> Saiko(Hello)

if it does not have #name it must be "you(&0)"

if you understand all of this and want to try to add words and sentences in go for it otherwise im gonna show you how to cheat

only cheat if you are adding an animation command

copy the nlp_actor.txt and rename it nlp_custom

//////////////////////////////////////////
// Custom
//////////////////////////////////////////
 
 
#custom = <#custom0> : animation(&0)
in_custom(*) : &0(animation),animation(type)
#custom0 = <#customval>           : ^in_custom(&0)
#customval = <* handstand *>     : handstand
#customval = <* bend over bench *>     : benchanal
#customval = <* ride me *>     : rideme
#customval = <* cartwheel *>     : cartwheel

open nlp_syntax and add

#include "ai/nlp/nlp_custom.txt"

under #include "ai/nlp/nlp_sentence.txt"

open nlp_sentence.txt

and add

#final = <#custom  *>                       : you(&0)

under nres final 0

this will make our animation response trigger first so don't use already existing words otherwise they will override them

inside the trace.txt after running the dungeon and saying the words

09:39:31 Me: CTALK: "ride me"

09:39:31 BRAIN: "Saiko[TALK]:talk=(closest:2,from:PLAYER,to:you,slen=7,cat=animation,e=[rideme.animation,animation.type])"

09:39:33 BRAIN: "state.code.ref.state.RUN0.animation.rideme"

09:39:31 Me: CTALK: "ride me"// what the player said

to:you// which is the you in #final

slen=7// how many character are in the word "ride me" includes the space

cat=animation// the animation in #custom

I think that's enough on the nlp

If i missed something or something is incorrect then let me know

Good luck