[Textop-en-phil] Top level outline headings

[Philippe MARTIN]

Larry,


> > Clearly. If you keep using relations in the outline, their meanings 
> > need to be respected, even the meaning of the default 
> > relation (the specialization relation): "Human being" cannot 
> > be a specialization of 
> > "Things in general" (since "a 'human being' is not a 'things 
> > in general'") but it can be a specialization of "Thing" 
> > (since "a 'human being' is a 'thing'). The following may or 
> > may not make sense: "a 'right and wrong' is a 'thing'".
> 
> This really does point up a problem with specifying relations clearly at
> this point: it makes it harder to edit the outline, because you have to edit
> the relationship metadata too.  Is there any clear advantage, that would
> justify that inconvenience, to wanting to keep relationship metadata
> up-to-date from this early stage in the project?  I'm afraid I'm not seeing
> one: I don't see why we shouldn't just add in the relationships, and
> processes for keeping relationship data accurate, once the outline is better
> developed?
> 
> What would be much more useful at this point is not the analysis of a rough,
> rough draft of an outline, which is what my work is, but instead a
> discussion of some *practical, usable* principles involved in actually
> structuring *this sort* of information into an outline.  I would greatly
> appreciate such a discussion from you, Philippe, if you feel up to it!

Not in a written form, since
- we already had many discussions on and off this mailing list and
  I have given the rationale and pointers for knowledge representation 
  principles;
- it is clear that what we have very different views on what is practical,
  usable and necessary;
- even the meaning of the specialization relation you are not prepared to 
  respect (as noted in your paragraph below) and this is the "least 
  painful" (and the most important) practical and usable principle 
  that you can follow;
- I am terribly late on other tasks that I have to do.

However, you can phone me (in the evening for you, around 11am or noon for 
me), for example via my skype identifier (which I once emailed you).

As far as specialization relations are concerned, some well known 
methodological advices to make sure that you respect their meanings 
are provided by the Ontoclean (ontoclean.org) methodology, e.g., read
http://www.cs.vassar.edu/faculty/welty/papers/er2000/LADSEB05-2000.pdf
As you will see this is way "heavier" (from your perspective) than 
anything I have so far refered to you but it is a good read that 
might encourage you to follow (lighter) ontological principles.

Regarding the use of categories/headings for domain fields, I 
strongly suggest the reading of Welty & Jenkins (1999)
http://www.cs.vassar.edu/faculty/welty/papers/subjects/subject.html
then my extension and application: 
http://www.webkb.org/kb/classif/_field_of_study.html
http://www.webkb.org/kb/classif/IT.html
In essence, categories/headings for domain fields should be avoided
since anything can be placed "under" them (any subdomain, task,
physical object,statements, etc.). It is a matter of "preferences".
They do not provide scalable organising guidance for inserting or 
searching information.



> As to whether "Human being" is a specialization of "Things in general":
> well, aren't you being a little pedantic here?  It's just a matter of
> formulation.  I can use the name of a set ("things in general"), the name of
> an abstraction ("thinghood"), or a singular formulation ("thing").  The
> reason "things in general" sounds more natural than "thing" (or "the
> functions of language" instead of "function of language") is that the former
> is what you say when asked what you're talking about.  It's purely a matter
> of style and readability.  Will it make it the slightest bit more difficult
> to identify the proper relationship between nodes if we do not fastidiously
> insist on the singular form?

I do not see anything pedantic and it is about meaning, not about style.
WebKB-2 (my knowledge server) would allow you to use cryptic "identifiers"
for categories (e.g., numbers instead of English nouns and, why not,
plural nouns to refer to individual entities, so that for example your
categories "persons" or "people" would refer to what I would call a "person")
but in your project
1) you have to be systematic or otherwise anything goes,
2) you will loose precision (e.g., in your case, you will not be able to 
   distinguish between the categories "human beings" (collectivelly) 
   and "human being" (one "human being"),
3) it will be confusing for other people.


Philippe