Monday, October 15, 2007

‘Intelligent Machine’ Learns Human Psychology

The foll is an abstract of the paper i had submitted on Artificial Intellegince in JNU ,New Delhi in August 06 .

‘Intelligent Machine’ Learns Human Psychology
Help AI Help People…………..Swiss Knife Of Communication.

“If a child is in a dangerous situation and you say ‘Don’t fall down,’ in order for him to understand what you have said, he has to access some representation of ‘falling down.’ That internal representation, especially if it is kinesthetic, will usually result in the behavior that the parent is trying to prevent.
However, if you give positive instructions like ‘Be careful; pay attention to your balance and move slowly,’then the child will access representations that will help him cope with the situation.”

Neuro-Linguistic Programming is the Study of the Structure of Subjective Experience and what can be calculated from it. In simple terms it’s the study of modeling patterns of communication. Though this field has been widely utilized in many areas like recruiters, therapists, sales persons to enhance creativity and elicit learning strategies, cure phobias and make the million dollar sale nothing has been done to guide the practitioner to train him and calibrate his own ability. The proposed work is an engine upon which applications that can train simulate and evaluate a practitioner’s performance can be built. There are fivecommunication channels -visual, auditory, kinesthetic, smell and taste and these are expressed as digital when writing or using a language to express them. And this is the only channel that is taken under consideration as it is beyond the scope of this project to consider non-verbal communication.
Processing verbal language requires machine learning and knowledge about the language itself. So this engine will be built on already developed API on natural language processing, neural networks and fuzzy logic , wordnet and other natural language ontology.

This paper is divided into three sections which explain :
1 Neuro linguistic programming as a meta-model of communication which is process oriented
2 The function of the proposed engine to understand the meta-model of communication so as
to simulate,analyze and detect communication patterns in dealing with therapy or sales
or even the truth in a conversation.
3 Design and Implementation of the engine

Neuro Linguistic Programming -
The Meaning Of Communication Is The Response

That You Get Meta-Modeling Most often we come across people who are so uniquely talented with their communication skills that they succeed in their task by establishing a close rapport with their clients in a magical fashion. Apart from what they talk or communicate about ,what really matters is how good they are at eliciting responses. This magical intuition with which they work is defined by a structure and it is explained by the meta-model ofcommunication. Once the meta-model has been explained , it is possible to learn how a successful communication is established and the same patterns can be thought to another person.

Process Oriented Vs. Content Oriented

The focus is here on the process rather than the content .By changing the process and giving more options over what patterns they can use , the person still maintains his own skill and creativity. He will also be able to adapt to new situations as what he learnt was process oriented and not content oriented.

The Function of the Engine
Psychologists and salespersons have practiced and succeeded in their professions, without realizing howthey did it. Our engine learns the patterns that make them successful using the Meta – model analysis. This can be used to teach others and arrive at generalizations
The engine learns and evaluates its own mathematical representation from further conversation and it adapts new representations accordingly. To start with the engine looks for the well defined generalized nominalizations in the meta-model like sampling . deletion , distortion , quotes , unspecified verb to start generating its mathematical meta model of effective communication and critically analysis the communication using the following evaluation pattern

What is the goal of every communication?
How close is a particular pattern helpful in taking you closer to the goal?
Evaluating how successful the pattern is?
How accordingly a different pattern or method is being used ?

Engine Design and Implementation
The engine has to be designed to understand natural language and process the information and analyzewhat’s going on and match it with the meta – model. For this it learns using a back propagation method.
It also needs to construct sales and a mathematical representation which it uses Quinine c4.5 algorithm andfuzzy sets. In order to comprehend the language and goal of communication the engine relies on wordnet framework and language anthology.
The engine has to parse the conversation and understand the lexical meaning of the sentence for which natural language processing algorithms using backward propagation and reverse propagation is used. To maintain a meta –model and analyze the results , a fuzzy set is generated from which it is compared and moreover ontology or knowledge of the language is necessary for which wordnet framework is used. The engine is built on a generic design , which allows any application to be build on it , its main purpose is to analyze and synthesis the meta-model of communication from given examples and match any given
conversation. In additi on the engine is also designed to learn continuously from every analysis andsimulation it performs besides its learning from the given set of data.