On Tue, October 30, 2012 05:56, Andries van Renssen wrote:
> Should the rule in a formal language to classify every individual thing be
> a requirement or a recommendation? (01)
For a formal language, that should be a requirement. (02)
> I don't see the precise benefits of classifying every 'kind' by a 'kind of
> kinds'. (03)
It is beneficial for argument restrictions on relations and for defining
disjointness. It is beneficial when more than one person is working
on an ontology in helping prevent divergent use of terms. (04)
> See below.
>
> William Frank, on woensdag 26 september 2012 at 20:11, wrote: (05)
> "all things shall??? be classified". See below. (06)
> On Wed, Sep 26, 2012 at 1:15 PM, doug foxvog <doug@xxxxxxxxxx> wrote:
> On Wed, September 26, 2012 08:53, Andries van Renssen wrote:
>> On 10 september 2012 at 20:41 doug foxvog wrote:
>>> On Thu, September 6, 2012 16:39, Andries van Renssen wrote: (07)
>>> > Furthermore, the statement is that all individual things 'shall be'
>>> > classified, whereas that is not required for kinds of things. (08)
>>> This is a rule in your system. Kingsley asked "what is not to like"
>>> about this text snippet.
>>> The lack of requirement that kinds of things need not
>>> be classified is something that falls in this category since the lack
>>> of classification limits the type of reasoning that can be performed. (09)
>> [AvR] The statement that "all individual things 'shall be' classified"
>> is indeed a rule.
>> Semantically it is not a necessity in Formal English,
>> because expressions can be interpreted without it.
>> However, I think it is a valuable rule, because it adds possibilities
>> for verification of the correctness (consistency) of expressions.
>> Therefore, the rule is intended as a (strong) recommendation. (010)
> Agreed. (011)
> [AvR] O.K. We agree that it is a valuable rule to classify all individual
> things. I assume that you are aware that this includes the
> classification of all (individual) relations. (012)
I am aware that many ontological systems consider relations to be
individuals. I am also aware of systems that do not define "individual"
so as to cover relations. (013)
> In Gellish Formal English up to now we apply this rule as a requirement,
> in order to enable verification of the completeness of an
> information model, because otherwise it is not certain that verification
> of possible requirements for things of a kind and
> verification of compliance with definitions can be performed. (014)
Good. (015)
> However, there is an argument that states that also in a formal
> language, users are responsible for the correctness of their expressions
> and they may be convinced that there are no requirements
> nor definitions that are not met, so that such a verification is not
> required. (016)
The only way that users can be responsible for the correctness of their
expressions is if they are totally clear on the meanings of the terms they
use
(especially if there is slack in the definition). How can one expect a user
to be totally clear on the meaning of the term if that term is not
classified? (017)
> Furthermore, humans may interpret classification from
> names or from relations with other things and properties, (018)
And humans can easily make mistaken interpretations because names are
polysemous (both for types and relations) (019)
> whereas it is not always necessary that computers can also
> derive that from the expressions. (020)
Since computers can not always derive classification from the expressions
if such classification is not stated, doesn't that behoove those who are
designing ontologies which are to be interpreted by humans to specify
the classifications? (021)
> On that basis it can be argued that the rule in a
> kind of Formal English to classify every individual thing
> can be a recommendation instead of a requirement.
> Do you agree with the latter? (022)
No. It seems that the argument was headed in the opposite
direction. (023)
>> The statement "all things 'shall be' classified" is a stronger
>> rule. IMO, it is valuable because it adds more possibilities for
>> verification of the correctness (consistency) of expressions than
>> the more limited rule that you state. (024)
> [AvR] I agree that it is also valuable to apply a rule to classify all
> relations between kinds of things.
> But when kinds are arranged in a taxonomy (subtype-supertype hierarchy),
> then they inherit their classification as 'being a kind'
> from the top of the hierarchy. (025)
Their classification of 'being a kind', yes. But not their classification
as to what
kind. (026)
If we look at a Linnean taxonomy, we have a set of kinds of kind, including
the meta-classes Kingdom, Phylum, Class, Order, Family, Genus, and Species. (027)
When the kinds of individual are arranged in a subtype=supertype hierarchy,
there are rules on the subtyping due to the metaclass, e.g., a Species may
be a subtype of a Genus, but unless they have the same extent, a Genus
may not be a subtype of a Species. (028)
> I see only a few occasions where explicit
> classification of kinds provide possibilities for
> additional verification. Please provide evidence by giving some examples
> of such additional verifications. (029)
I've been working with ontologies for material science recently. This
field needs
relations among classes all the time. Arguments for such relations include
types of material, elements, chemicals, etc., each of which are types
which have
their own instances. (030)
It is useful to state that Aluminum is an instance of MetallicElement, and
have that mean that it is also an instance of ChemicalElement a supertype
of the meta-class MetallicElement (031)
> Because of the above, I think that it is not justified to introduce a
> general rule that every kind shall be classified by a kind of
> kinds. (032)
In toy systems, classification is not needed. In real systems used by
a variety of users, classification aids in understanding even if no relation
restricts its arguments to a specific kind of kinds. (033)
> In natural languages we also classify mainly individual things and
> classify kinds only rarely. (034)
It seems to me that in natural languages, types of things are routinely
classified. (035)
> [AvR] The question is also not whether kinds shall be defined or related
> to other kinds, because the specialization relation(s) and
> the relations that form the definition model are precisely the relations
> that make the classification relations of kinds in most
> cases superfluous. (036)
If there is only a single hierarchy below a given class in an ontology, then
there is little need to specify a kind of kinds. However if there are non-
disjoint subclasses, it can be useful to distinguish them as being different
kinds of subclasses (subclass by gender, by species, by age, by ...). (037)
-- doug foxvog (038)
> ...
> --
> William Frank
>
> 413/376-8167
> (039)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (040)
|