From: Daniel J Klein (djklein_at_u.washington.edu)
Date: Wed Oct 22 2003 - 10:48:43 PDT
Summary:
The authors argue that the restricting language, as presented by Brachman and Levesque, results in specialized KR that lack the generality a KR should have.
Review:
The first part of the paper sets up the author's somewhat negative view of restricted language thesis. They specifically focus on generality, soundness, completeness, and efficiency stating that restricted language results in sound, complete, and somewhat efficient representations that are specific. The trade off between generality and these other measure is a theme of the paper.
The authors then present their counterargument to the restricted language representation. They clearly state, "the restricted language and restricted classification theses and their underlying assumptions are flawed". It seems the authors took personal offense to the restricted idea and this paper is their chance to set things right. They continually hammer on the fact that limiting expressive power results in lost generality.
The view the authors is that a language should be fully expressive. However, this comes at the expense of soundness, completeness, and worse-case time complexity. But they argue that these measures are "not right". With a fully expressive language, the authors argue that rational or optimal conclusions are "better" than logically sound ones.
In general the paper is well organized and their statements are well supported. However, the paper was a bit long winded for me. Also, I strongly disliked the general tone of the authors. Sure, they may be right, but they only hurt them selfs by continually nit-picking on Levesque and Brachman.
Despite the long-winded nit-picking, the authors are quite convincing and might actually have a point. Their argument that general purpose systems need to be able to provide a language that can express a variety of concepts is the focus of their argument and the key contribution of the paper.
This archive was generated by hypermail 2.1.6 : Wed Oct 22 2003 - 10:48:45 PDT