With knowledge representation we face more or less the same problem as Augustine
(354–430) when thinking about time: if nobody asks what it is, it seems
clear enough, but being asked it proves to be very difficult to provide an answer.
At the beginning of our research we thought that a solution for the problem
of knowledge representation depends on a solution for the problem of natural
language processing. But this was wrongly taken. Our experience with practical
applications of grammars developed by linguists has shown that formal
grammars, at least those we had access to, cannot capture the full complexity
of language. As a consequence of an analysis of our ‘results’ we drew the conclusion
that a reason for the fiasco we experienced must reside in the formal,
i.e., static, detached character of our approach. Natural language and human
communication in general are basically dynamic processes. We surmised that
the sources for the dynamics of language can be found in two fields: in human
information processing and in human communication. We also realized that any
modeling of the first requires a theory of cognitive activity, and any understanding
of the second needs a theory of signs. As the two fields are torn asunder, in
the beginning it was not clear whether it is possible to establish a link between
them. It was clear, however, that if we could find a solution for the problem of
language interpretation, we would be able to find an answer for the problem of
human knowledge representation as well.
This was the situation when we started our research in order to develop a
model of knowledge representation that respects the properties of human information
processing as well as the properties of signs. Soon we found out
that a proper understanding of the relation between the theory of signs and
the cognitive model of knowledge representation taken is harder than expected.
Similar to experiences with problem solving in other fields, for instance, chemistry,
which requires some understanding of the properties of chemical elements,
we found that the definition of a model of knowledge representation asks for
some understanding of the rules pertaining to the propagation of signs. We had
to realize that signs are not some special kind of ‘things’, but consist in complex
systems of dependencies.
A promising result of our research is the development of a theory that, in
our view, provides an answer to our original problem. Results of various tests
and experiments of our theory show that the approach taken could be correct.