From: Masaharu Kobashi (mkbsh_at_cs.washington.edu)
Date: Wed Oct 22 2003 - 09:48:48 PDT
Paper titie: Two Theses of Knowledge Representation
Language Restrictions, Taxonomic Classification,
and Utility of Representation Services
The paper proposes that the general purpose KR should provide
the service of providing rational or optimal conclusions rather
than logical soundness, and completeness.
First, the view of the authors is worth serious considerations since
they value the practicality rather than formal cleanliness. Such a view
can have positive impact on the real world applications of AI more
than the conventional ones.
Second, they propose in a sense a mixed systems which do not completely
eliminate the constructs which can have the worst case exponential
complexity. They propose the systems can be more realistically
tuned by limiting the behavior of such constructs rather than
completely liminating them.
The largest flaws in the paper is that authors tend to focus only on the
cases where their proposals better fit than the conventional ones.
But it seems there are still substantial number of cases where the
conventional view of KR has their value.
Secondly, the paper's assertion is not backed by any real implementation
of their philosophy. Also their bashing of the conventional view
occasionally seems to be making a strawman.
One open question is "can there be an ideal general purpose KB?
Or can there be a general purpose KR which surpasses each specialized
KR in its particular domain? Is being more general always a better way?
This archive was generated by hypermail 2.1.6 : Wed Oct 22 2003 - 09:48:51 PDT