TITLE: 		TRANSPARENTLY-MOTIVATED METAPHOR GENERATION
AUTHOR:		Mark Alan Jones
COMMENTS:	Ph.D. dissertation. (c) 1994 Mark Alan Jones

ABSTRACT:	
This dissertation is the first attempt at generating metaphors in a
general manner. It introduces and defines a the class of
transparently-motivated (T-M) metaphor, and describes a method for
generating them via a set of top-down rules that transform a literal
expression into one containing a T-M metaphor.  Our aim is to provide
a generation system with the same conceptual groundings that people
often use to communicate. This approach is strongly motivated by
several fields of study, including the fields that comprise cognitive
science: linguistics, computer science, psychology, and
philosophy. Because of this motivation, this dissertation is the first
to investigate the uses for metaphor from the point of view of the
natural language generation paradigm. This work introduces a
computational approach to provide additional metaphorical options for
a text planner to express itself so that the text generated will be
easier for the reader to comprehend. We show these metaphors can be
used to achieve specific textual goals for natural language generation
(i.e., manipulating focus of attention, and increasing
comprehensibility). A computer implementation which embodies many of
the ideas in this work, called Quipper, is described in detail.