[Top] [All Lists]

Re: [ontolog-forum] Fact Guru, Controlled NLs, and OOR

To: ontolog-forum@xxxxxxxxxxxxxxxx
Cc: Doug Skuce <drskuce@xxxxxxxxx>, John Talbot <johntalbot@xxxxxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Wed, 30 Mar 2011 22:23:34 -0500
Message-id: <4D93F3B6.3080208@xxxxxxxxxxx>
Mike, Todd, Leo, Ed, Doug F,    (01)

Fact Guru has stirred up quite a bit of discussion.  Having an actual
implementation, such as FG, can provide concrete examples that can
clarify the options and show how they could be used.    (02)

> ... ontologies and terminologies are complementary but different,
> so any tool that formally relates one to the other has to be good.    (03)

Professional *terminologists* have been designing terminologies
for science, engineering, business, and international organizations
long before computers were invented.  They try to be as precise as
possible, even though their definitions are stated in natural languages.    (04)

The UN and the EU have professional terminologists whose primary job
is to specify equivalent words and phrases in each of the official
languages.  A useful extension to Fact Guru should allow terminologies
in different languages to be mapped to the same ontology.    (05)

Todd Schneider wrote the following comment to OOR forum:    (06)

> The ability to enforce common vocabularies
> across a domain is beyond the scope of current requirements
> and expectations. However, I could see this as specified via
> policies and enforced as a work flow. Of course the enforcement
> of these more global policies would be 'expensive' (from a
> computational point).    (07)

I'd like to illustrate some of the options in terms of Fact Guru,
and show how systematic ways of using a tool such as FG could
enforce OOR policies.  Those features aren't yet in FG, but when
it is released as open source, people could implement various
extensions to support *and* enforce OOR policies.    (08)

For example, consider the SUMO example, which is the kind of
application that OOR is designed for:    (09)

    http://www.site.uottawa.ca/~tcl/factguru1/sumo/index.html    (010)

The left column contains the names of various KIF terms, such as
Entity, Abstract, Attribute, BiologicalProperty, AnimacyProperty,
Living.  Those names serve two purposes:  they are English words
or phrases, and they are labels of semantic features of KIF.    (011)

I claim that the column on the left is a *terminology*, which
happens to have a one-to-one relationship with the labels used in
the ontology.  Many ontologies happen to have that property,
but others may have very different relations between the English
terminology and the names of the formal terms used in the ontology.    (012)

Whatever policy the OOR adopts could be enforced by a suitable
option in FG.  The current version of FG also has formats for
importing and exporting the terminologies on the left and the
associated information displayed on the right.    (013)

The current format is probably not the one that the OOR would
adopt as a standard, but changing the formats is, as they say,
ASMOP (A Simple Matter Of Programming).  Enforcing the policies
could also be computationally simple:  the software could be
designed to support only those options that are defined in
the standard format.    (014)

> You may also use a term as a label for a concept, which is fine
> if you understand the distinction, but the concept could be
> labeled KDKJKDLK123 in the ontology and that really is just
> as valid, if perverse (though you cannot interpret it
> semantically as an English speaker unless you see the term
> in the terminology that refers to it). And note that there
> is not a 1-1 relationship between terms and concepts, but
> a many-many, depending on the language and of course the
> ontology.    (015)

I mostly agree.  But note that KDKJKDLK123 is no more perverse
than many labels found in computer implementations or the
URIs recommended for the Semantic Web.    (016)

> There is a particularly cute trick in one of the ISO TC37 (Vocabulary
> and Terminology) standards (ISO 1087-1, I think).  It defines
> 'designation' as the relationship between an 'expression' (variously
> called the 'sign' or 'signifier' or 'term') and a 'concept', with the
> statement: "In the speech community that uses the designation, the
> signifier refers to the concept and denotes its instances in a world of
> interest."
> ... sorting out the terminology for terminology concepts seems to be
> a problem for terminologists.   It is turtles all the way down...    (017)

Since a very large number of ontologies are related to terminologies,
we need to accommodate both.  If we require every formal ontology
to be stated in some logic that has a model-theoretic semantics,
we can provide a solid ground to support the bottom turtle.    (018)

By the way, the New Yorker had a cartoon that showed Atlas standing
next to the globe of the earth, which was perched on a three-legged
stool.  The caption read:  "Why didn't I think of that before?"    (019)

Tarski's model theory is that three-legged stool.    (020)

> Natural language semantics (and natural languages) are not well
> founded. The application of the notion of interpretation from
> First Order Logic in these arenas also suffers from this attribute.    (021)

The discussion of NL semantics can get us bogged down forever.
Let's just adopt the policy that a formal ontology can be used
to specify a precise definition for the terms of a terminology.    (022)

> When using multiple defined terminologies, the mapping would be many-to-one
> for concepts that are represented in multiple ontologies.  When using
> ontologies, if the full name of an ontological term includes a unique
> reference to the ontology, then the issue of many-to-many would not
> arise.  If one carelessly uses terms from multiple ontologies without
> designating which ontology the term comes from, one is back to
> many-to-many.    (023)

I agree.  And it would be possible to provide options in Fact Guru to
make it clear which terminology is being mapped to which ontology.    (024)

> But I think we owe some simple enlightenment to a larger community,
> providing a kind of Newtonian physics when we know that, really,
> Einstein rules.    (025)

This example shows why we need the option of mapping the same terms
to different ontologies.  Scientists who know that relativity is more
precise than Newtonian mechanics also know that they can use the
simpler Newtonian calculations for many applications.  In fact, the
popular PSL ontology for specifying processes does *not* use the
relativistic corrections.    (026)

> The basic premise is that as humans we don't have a mechanism to
> (completely) step outside of ourselves and the natural languages
> we use to communicate.    (027)

Yes, but we always have the option of using our language at the
metalevel to talk about the level(s) beneath us.    (028)

In fact, there is no limit to the number of metalevels we might
use, but in practice too many metalevels can get confusing.    (029)

That confusion level is why we don't have to worry about
"turtles all the way up."    (030)

John Sowa    (031)

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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (032)

<Prev in Thread] Current Thread [Next in Thread>