From: Tyler (trobison_at_cs.washington.edu)
Date: Wed Oct 22 2003 - 09:38:36 PDT
Two Theses of Knowledge Representation
Jon Doyle & Ramesh S Patil
This paper attacks the conclusions of Levesque and Brachman, who
propose a knowledge representation system that accepts some restrictions
in order to avoid undesirable response times.
An idea at the heart of this paper is the balance between
expressiveness and complexity. The Levesque and Brachman paper holds that
restricting certain aspects of
expression is a legitimate way to efficiency. This paper maintains that
restricting expression makes the overall system less useful, and is not
worth the efficiency gained. They argue that the inefficient constructs
can be used lightly, when it is worth the cost, and that omitting them
altogether is a heavy-handed remedy.
A second idea of importance are the consequences of these opposing
views: if Levesque and Brachman are correct, then there are no
general-purpose knowledge representation systems. If Doyle and Patil are
correct, then we can construct general-purpose knowledge representation
systems. We would of course like to be able to construct a general system
that could be used for any application, so this is a relevant concept.
The most significant flaw in my mind is not necessarily in the
conclusions drawn, nor in the argument itself, but rather in the
presentation. A little conciseness and simplification would have gone a
long way here, as much of the paper is overly wordy, while at the same
time actually conveying little information. At times several pages can be
read without the reader having a sense of progress in terms of the
argument; at least that was my experience.
Since the ideas covered in these two papers are significantly
important (is it possible to construct a general-purpose knowledge
representation system?), it would be nice to see a conclusion drawn by
some objective third-party, taking into account arguments and examples by
both sides.
This archive was generated by hypermail 2.1.6 : Wed Oct 22 2003 - 09:38:48 PDT