TABLE OF CONTENTS

INTRODUCTION					     ?
	The problem of recognizing natural language
	dialogue expressions in real time. Previous
	approaches. The problem of simulating paranoid
	linguistic behavior in a psychiatric interview.
	Summary of a method for transforming natural
	language input expressions until a pattern
	is obtained which completely or fuzzily
	matches a more abstract stored pattern.


OVERVIEW					 ?
	The paranoid model (PARRY2) consists of a
	RECOGNIZE module which performs the task of
	recognizing the input and a RESPOND module
	which decides how to respond. The RECOGNIZE
	module functions independently of the RESPOND
	module except in the case of anaphoric
	references which it provides on request from
	the language recognizer.


PREPROCESSING					  ?
	Dictionary lookup and translations. How
	misspellings and typing errors are handled.


SEGMENTING					  ?
	Bracketing the pattern into shorter segments.
	A "simple" pattern contains no delimiters; a
	"compound" pattern is made up of two or more
	simple patterns.


MATCHING INDIVIDUAL SEGMENTS                      ?
	Negations and anaphora. Matching the pattern
	with stored patterns having pointers to
	response functions in memory. If a complete
	match is not found, a fuzzy match is attempted
	by deleting elements from the pattern one at
	time. If no match is found, the RESPOND module
	must decide what to do.


COMPOUND-PATTERN MATCH                                ?
	Complete and fuzzy matching when the pattern
	contains two or more segments.


ADVANTAGES AND LIMITATIONS                         ?
	The advantages of ignoring as irrelevant some
	of what is recognized and what is not recognized
	at all. The complete language recognition process
	of the algorithm requires less than one second
	of real time. How the data base "learns". The
	measurement of linguistic improvement.


REFERENCES                                         ?


APPENDIX 1                                          ?
	A diagnostic psychiatric interview which
	illustrates some of the model's linguistic
	capabilities.


APPENDIX 2                                           ?
	A listing of the dictionary illustrating
	the algorithm's recognizable input words and
	the word class names they are translated into.


APPENDIX 3                                           ?
	A listing of the simple patterns.