Message 12148 of the SUO list

Subject: Re: Model of Activity and Action in SUO ontology
Date: Tue, 16 Dec 2003 05:50:54 +1000 
From: Philippe Martin
In reply to: msg12147 by Richard Cooper and msg12146 by Murray Altheim


Richard Cooper wrote:
> Philippe, do you have a URL that defines the single characters you're
> using to describe aspects of each concept?  Characters such as
> ^ ! = M p s L < and so on.

Sorry for not refering to that in my last two files.

1) You can see the taxonomies with "long link names" by adding
"&longLinkNames" at the end of the URLs I give or by using the
"informal" option of the "category search interface" at
http://www.webkb.org/interface/categSearch.html

2) Alternatively, the encoding of the link is listed in 
http://www.webkb.org/doc/w2doc.html#link
and http://www.webkb.org/doc/wn/wnIntegration.html
and in one of the menus of the "category search interface"
but to simplify, here is the current list:

  < (subtypeOf),   > (subtype),    ^ (instanceOf), : (instance),
  = (equal),       - (inverse),    ! (simple exclusion),
  / (closed exclusion: the two linked types are exclusive and
            either are respectively identical to pm#thing and pm#nothing
            or they subtype a same type and form a complete subtype partition),
  ~ (closely similar from an Information Retrieval viewpoint; such links are
     currently only used between categories for Greek gods and their Roman
     counterparts, and between some types from the 3D (endurantist) approach
     and their counterparts from the 4D (perdurantist) approach or the
     ?D (vague/unspecified) approach)
  L (locationOf),  l (location),   M (memberOf),   m (member),
  S (substanceOf), s (substance),  P (partOf),     p (spatial_part/subprocess),
  N (nounTypeOf),  n (nounType),   O (objectOf),   o (object/experiencer),
  U (urlOf),       u (url), 
  I (inputOf),     i (input),      R (resultOf),   r (result)
  A (agentOf),     a (agent),      T (toolOf),     t (tool/technique/instrument)

When two or more names are given for a link (as in "spatial_part/subprocess"),
the types of the linked categories permit to distinguish the relevant relation
type (e.g. pm#spatial_part or pm#subprocess).
Note: users are allowed to introduce new kinds of links as long as their
encodings is an alphabetic letter.


> Do your characters have the same meaning as those in the WordNet database?

The rationale for my use of links comes from my re-use of WordNet (and because
they are more readable and much quicker to parse than full CGs) but
I use a different encoding because the encoding of WordNet is neither
standard, nor intuitive, nor economic (it often uses many characters).
WordNet also has no encoding for equal, reverse, instance, location, url,
input, result, agent and tool. Cardinalities may also be associated to links.


Murray Altheim wrote:
> I note in that document you mention a "Partology"

Yes, for the sub-headers of Section 3, I re-used the categorization of NSM.

> ... I just happen to be working on something in
> this area right now and was wondering if in your work on this you
> either already have or plan to further differentiate the subtypes
> of:  nsm#part_of = pm#part_of (pm);

The relation type pm#part_of has no subtype but its inverse, pm#part,
has currently 23 subtypes; see
http://www.webkb.org/bin/categSearch.cgi?categ=pm%23part&recursLink=%3E&hyperlinks
and there are various concept types for the word "part", see:
http://www.webkb.org/bin/categSearch.cgi?categ=part&hyperlinks    See also
http://www.webkb.org/bin/categSearch.cgi?categ=divisible_entity&hyperlinks

In CGKAT, I had Sowa's classification (lumpy, continuous, ...). For some 
unknown reason, they are not currently in WebKB-2 but it would be quite easy
to re-enter them (and some other ontology you might like).

Some structuring needs to be done there. Before doing it, I'd like to
remove the distinction between concept types and relation types, which leads
to partial duplications and waste of possibilities. Within WebKB-2, the same
identifier should be able to be seen as an identifier for a role type and for
a relation type (and thus have a signature). When exporting to other languages,
the distinction will be made. This will require some upgrades to WebKB-2,
especially in the controls. So, probably end of January since I'll have a
break from next week to mid-January.


> It would seem important to be able to differentiate between
> the relation between some sand on a beach, the piston of an engine,
> members of a family, etc.,

Sub-situation (+ sub-process), physical-part, substance and member 
relationships are distinguished.

Philippe