Re: Roles, again (Pat Hayes)
Message-id: <>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Mon, 11 Sep 1995 14:35:07 -0600
From: (Pat Hayes)
Subject: Re: Roles, again
Cc: "Pierdaniele Giaretta" <>,
        "Massimiliano Carrara" <>
Precedence: bulk
At  1:28 PM 9/8/95 +0100, Nicola Guarino wrote:
......<much omitted>

>>>A final comment deserves Peter's concern regarding the proliferation of
>>>concepts: one of the advantages of a clean distinction between types and
>>>roles is the possibility to superimpose a "skeleton" of types on a huge
>>>network of concepts: such a skeleton (which in most cases turns to be a
>>>tree) would help a lot for indexing purposes.
> When I was speaking of trees, I intended
>to underline the fact that types *tend* to form a tree, ....

Let me suggest that type heirarchies have skeletons consisting of what
might be called forests, ie collections of trees. Tree structures are
desirable and have many nice properties, but most things can be classified
in several ways, and it becomes unweildy to force the whole system into a
single tree. A collection of overlapping trees is much more natural and has
many of the computational advantages of a single tree. However, this allows
everything to be classified as everything higher, in principle, so its only
useful if there arent too many of these overlapping structures. A constant
tactical question for ontologists should be how to keep the number of trees
reasonably small. (This is not to say that the number of overlaps need be

I know many will agree with this. Does anyone find it an unreasonable
restriction, or have any other problems with it? Does anyone have any
examples which suggest that it wont work?

Pat Hayes

Beckman Institute                                      (217)244 1616 office
405 North Mathews Avenue            (415)855 9043 or (217)328 3947 home
Urbana, Il.  61801                                     (217)244 8371 fax