Alex suggested and approach
to “retrieving ‘raw knowledge’, being captured in the language terminology, via
the means of the hierarchical classification of all terms (nouns), contained in
the given human language based on the verb <-> noun grouping” an idea
borrowed by him from the Software Object Oriented
representation of the class (object). I think, however, there are many issues to
supposing” for each term (noun) available in the language, one can "gather" the
set of all verbs, which could be related to the given term (noun). For some
noun-things like a “hammer” we can associate a focused set of verbs or better
yet relationships expressed by predicate.
However every
human language is a system of
conventions that define for participants with a set of means for encoding an
unlimited class of concepts. A hammer can be used to wedge a door open, for
example and an indefinite number of extended usages in different contexts. Fillmore developed a theory of
lexical semantics called Frame Semantics which stressed the importance of
linking word meanings to the 'frame' of general ideas that define them. This may
be a better way to get at knowledge – not “raw” but in context. So the
relations of nouns and verbs constitute an open set with some
constructed and interpreted on the fly using schemas of knowledge.
As a further example of
issues with a simple verb to noun associate, the class of nouns includes such
concrete types of things as milk, money, furniture, soil and places, which I
have difficulty binding to a small set of verbs. Washington is a place for example. How would one go about establishing the
one-to-one unique relationship with the given specific set of verbs and implied
relations that go with Washington? As a collective noun Washington does many
things and has many relations, like acquiring a ball team and choking on
traffic. There are also problems
with abstract noun concepts like “justice” that might quickly daunt an attempt
to build a verb tree to align to it.
We quickly get into a
discussion of semantics. The units of semantics are concepts which are not
usually in a one-one relationship with noun and verb words:
- one word may contribute several
concepts (lexical semantics),
- one concept may be built on the
basis of two or more words (semantic phrasing).
One of the leading ideas of
cognitive linguistics is that there are no natural boundaries that bound
language - Word meanings fade into encyclopedic knowledge much of which is
individual to a person based on embodied experience.
There are fair amount of cognitive
theories that might be applied to this discussion. Langacker has a psychological theory of Cognitive (after
Gestalt psychology) which uses framework for analyzing meaning in noun verb
terms using notions such as
schemas, profiles and landmarks. So if we are trying to get at raw
knowledge we might consider some approaches that come out of cognitive
linguistics instead of an object modeling procrustean bed.
Gary Berg-Cross Ph.D.
Cognitive Psychologist
Knowledge Strategies
Potomac
Maryland
-----Original Message-----
From:
owner-standard-upper-ontology@LISTSERV.IEEE.ORG on behalf
of Alexander Povolotsky
Sent: Tue 3/15/2005 8:43 AM
To: ontology@LISTSERV.IEEE.ORG
Cc:
standard-upper-ontology@LISTSERV.IEEE.ORG
Subject: nature ->
"human brain" -> "language terms" ==>> knowledge
?
Hello,
I follow the concept that the human brain
reflects on "nature"
(as an objective reality). In its turn, the above
reflection
gets "captured" into the language terminology in some "raw"
form.
Below I propose the method of retrieving this "raw
knowledge",
being captured in the language terminology, via
the means of the
hierarchical classification of all terms
(nouns - see below), contained in
the given human language
(say English Language as most scientifcally
common).
The suggested approach is based on the verb <-> noun
grouping
and is "stolen" by me from the Software Object
Oriented
representation of the class (object). In this
particular
adaption of the OO, the terms (nouns) are analogous to
the
object's data and the verbs are analogous to the "methods"
(aka
"member-functions") which could be applied to (performed on)
the
data.
Suppose for each term (noun) avalable in the language,
we
will "gather" the set of all verbs, which could be applied to
the
given term (noun).
Each set is corresponding to the unique noun, as it
was
described above - so the trees are built around nouns due to
their
one-to-one unique relationship with the given specific
set of
verbs.
Then we could compare each generated (per above description)
set
against all other sets (separately on one-to-one basis) to
find
whether some sets of verbs could share the "common"
subsets.
Then we could attempt to detect whether some sets
were
derived (inherited) from the other sets so we would be able
to
build the hierarchical trees of such related sets.
Actually, the
nodes of the trees should contain the nouns
(rather than
corresponding sets of their verbs).
The top node of each such tree
would contain the set (actually
uniquely corresponding to it noun as
mentioned above), which would contain just the "common" subset of the verbs or
the "minimum" number of verbs.
Such noun with the "minimum" set of verbs
has the highest
level of the "abstraction" in the given tree.
To
complicate the "picture" the two nodes, which belong to
two different
trees, may "act" as "parents" nodes to generate
the "child" node (Multiple
Inheritance), etc.
Further, some quantification of the "abstraction"
value could be
applied to each distinct tree - the "most bottom" node
should
have the "abstraction value" set to 0 (zero) and for each
next
higher level the "abstraction value" should be
incremented by 1
(one).
The above excercise, which I am suggesting to conduct - is not
intended to develop some aid in processing human texts, except may be for
the
very small minor thing of being able to somewhat
quantatively
characterize the average level of abstraction, used in the
text (and
only if the text is recognized, using other cognitive means, as
"scientific", covering certain domain(s) of the science). I
"claim",
however, that the terminology, developed historically in the > human
language, "subconciously" reflects (without much
of "distortion"/"noise")
the objective fundamental reality of the
nature, so the language's terms
are in fact "synonyms" of the
nature's attributes. So in that sense I
believe my excercise could
contribute to studying of the nature itself and
its universal
structure. I guess that the results obtained (as described
above)
could be analyzed using methods developed in the Theory of
Graphs.
Below is my view on classification of the knowledge
:
Currently human race uses two major levels of expressing the
knowledge:
1) Textual expression - historically it was developed/used
first,
but it is rather "primitive": very context related,
non-
quantative, potentially being subject of uncertainties
of
interpretation(s), could be "false" in principle - hence
the
problems being dealt with,
in the
scientific application domain, (objectively speaking, this
not
a fundamental but application domain). This textually
expressed knowledge is "acceptable" for the human
specific utilitarian domains. However, when the textual
knowledge
addresses the fundamental areas/domains - it is (in
my view)
should be considered to be "preliminary", subject of
"upgrading"
into the mathematically expressed knowledge (see
below).
2) Mathematically expressed knowledge (formulas) is:
precise,
quantative,(typically) proven by (based upon
quantative)
experimental data. This form of expressing the
knowledge is used
for describing the fundamentals of the
nature (which is not
necessary utilizable by humans and, by
all means, often could not
be utilized by the humans
immediately). Theoretically, this
mathematically expressed
knowledge could "survive" beyond the
"era" of the human race
existence (dialectically speaking,
everything starts, evolves
and ends/gets "pre-empted").
Thanks,
Best Regards,
Alexander
Povolotsky
|