2000, Volume 3, Number 3, pp.253--259
It is discussed the problem how to design a machine-programmer, i.e.
an artificial system capable to produce computer programs. Making a
computer program is treated as a translation of a text that describes
an algorithm in a natural language into a text of a program in some
algorithmic language, whereas the essence of the developed approach is
an intermediate stage of this transition, i.e. comprehension of the
meaning of the original text. It is shown that generalized frames can
be used for analysis of meanings, which removes the intractable problem
of finding a set of independent meanings. This enables to suggest an
outline of a dynamical mechanism of thinking based on associative
generation of information in a semantic space which employs hyperbolity
of chaotic attractors. It is a so-called "regularized English-like
language" (RELL) that is stated to describe customer requirements and
to formulate algorithms. The above mentioned points are combined and
embodied in an architecture of a machine-programmer composed of lexical,
semantic and operator neural systems which are implemented as a neural
network, a neural field and a neural network respectively.
Key words: neural systems, language, chaotic processors.
Full text: Acrobat PDF (120KB) PostScript (400KB) PostScript.gz (125KB)
Copyright © Nonlinear Phenomena in Complex Systems. Last updated: June 19, 2001