[Top] [All Lists]

Re: [ontac-forum] Semantics, Linguistics and Ontology

To: kevinsl@xxxxxxxx
Cc: ONTAC-WG General Discussion <ontac-forum@xxxxxxxxxxxxxx>
From: Ken Ewell <mitioke@xxxxxxxxxxxx>
Date: Thu, 25 May 2006 04:18:20 -0400
Message-id: <4475684C.2000209@xxxxxxxxxxxx>
Kevin,    (01)

You continue to charge in the substance of your statements that I am 
hiding something or there exists something ulterior to what I have 
stated. Your statements attest to this account of mine. The specific 
words you used stand in evidence.    (02)

As John Sowa has reminded us we can talk more precisely about very 
specific things.  I warrant that any reasonable analysis of the facts at 
hand would substantiate my account.  I have no need of unnatural or 
ulterior belief for my warrant.  Let anyone who believes otherwise speak 
out here.    (03)

Are there not consequences to such a continuous charge?   Some people 
would infer I must have ulterior motives; others may infer I am covering 
up or lying. In either case, damaging the character of a significant and 
above all honest and worthy body of work.    (04)

In addition, you implied for anyone to infer that I have supernatural or 
alien beliefs in the form and substance of your statement of alarm.  I 
would not presume to discriminate your beliefs and other than the 
healthy skepticism we should all bring to bear, I would not expect 
unverifiable personal beliefs to be of any significance at all.    (05)

For the record this theory belongs to Dr. Tom Adi.  I cannot defend his 
theory for him, I can only offer my interpretation and speak from my 
direct experience with its explanatory power.  I would be honored to 
introduce you to him. I can attest from a lifetime of working with him 
that he is quite capable of defending this work in a proper circumstance.    (06)

> KL> I guess I could be letting some of my own biases sneak into my 
> 'reading' of your theoretical proposals.  Bringing my biases along 
> with me is somehow part of the 'reading' process, I findů But I don't 
> think I was unkindly attributing beliefs to you that had no warrant in 
> the substance of your statements.      (07)

Speaking of the substance of statements    (08)

>     KL> Actually, I almost forgot, there are some contemporary folks
>     who share your belief that the shape and structure of written
>     language conceals some necessary and eternal truths:  <snip>    (09)

Shall we test which inferences can be drawn from that composition of yours?    (010)

Speaking of warrants    (011)

> KL> I'm trying to see how the following is somehow NOT a statement of 
> belief about how the structure of written language conceals some 
> necessary connections to its referents:  "[W]e believe that the 
> morphology of a word, its structure, is also a sign that refers to an 
> abstract object structure that is somehow related to the structure of 
> the object to which the word refers." ??    (012)

The part you quoted does state a belief though I fail to see a necessary 
connection to something being concealed.  I stated this belief as Tom 
Adi fomulated it to me. It is one part in a process of cognitive growth 
by reinterpretation.    (013)

> KE> The repeated interpretation of the abstract objects to which the 
> phonemes of a word refer, in light of the repeated interpretation of 
> the abstract structure to which the morphology of that word refers, 
> will establish more and more relationships in the human mind to the 
> properties of the object to which that word refers. That is our story 
> of cognitive growth by reinterpretation.      (014)

Shortly after: I perhaps indelicately explained how the abstract object 
structures are defined. A very specific rule was *revealed* and I 
*revealed* precisely which sign structures indicate the stated rule and 
the structural pattern (n,p) + (p,p) that we were selecting as the 
example to look at in real terms.    (015)

Then I listed the many examples Tom was kind enough to prepare together 
with the taxonomy of semantic classification, so that we more readily 
recognize the semantic properties organized by that rule . It also shows 
how the methods map to canonical lexical roots in many but not all cases.    (016)

It seems also, the idea is not so alien after all....    (017)

>LO>  In any complicated application, you will need a lexicon, a thesaurus, and 
>an ontology (or a number of each). Why, because typically you have to go from 
>text (unstructured or structured, e.g., schema labels, metadata terms, etc.) 
>to meaning. Lexicons (especially linked with a morphologizer that can generate 
>or map to canonical lexical roots) enable you to go from terms to word senses, 
>which can then be linked to thesauri, and the latter mapped to ontologies. 
> KL> To take one of your specific examples: "Terms of this type [and 
> perhaps this is where I'm getting confused, because I don't know 
> precisely what is meant by 'this type'] that are related to mental 
> activity include "brain", "mind", "remind", "admonish", "meditate", 
> "medulla", "dementia", "demented", and "cerebral"."    (018)

I would direct your attention to the rule mentioned just previous as the 
type: this type of rule. The terms included in the examples are 
semantically classified by that rule in that way.  The question is does 
that group makes sense?  I did not make any claims in that regards and I 
exaplained one way I figured out for testing such "extraordinary 
theoretical proposals".    (019)

> So, my 'interpretation' of the above is that the shape and structure 
> of these words already reveals a shared, somewhat hidden, but 
> necessary connection to the things -- the abstract object structures 
> -- they signify.  And, presumably there is a neat theoretical 
> explanation for how we've acquired other words in English related to 
> mental activities, like "intellect," "cogitation," "wits," "senses," 
> "rational," "noetic," "psyche," that don't fit the pattern?      (020)

Indeed. Should there only be a single pattern relating all the topics of 
the human intellect, psyche, disease and physiology?     (021)

I do not know how you interpret anything about shape from what I 
stated.  I was not talking about shapes or shape and it has nothing to 
do with anything I wrote.  I do not know what you mean by shape and its 
role here.  That the structures are abstracted or removed from the 
actual case or imagined states of affairs does not imply they are 
hidden.  Though it is clear we must all be very much confused about them.    (022)

The words you named do not fit the pattern of the given example. They 
certainly fit other patterns. The orthography of the words indicate 
different mophology and these patterns inspire different words senses. 
Who would use the term "intellect" to refer to that portion of the 
vertebrate central nervous system that is mostly enclosed within the 
cranium?  There could be many conventional and social reasons why not?  
Because the patterns do not fit is one reason.  Because it is not 
sensible is another.  How many good reasons does one need to make a case?    (023)

> In the spirit of examining a theoretical proposal, I'd like to know 
> how we can be sure these patterns are observed meaningfully and 
> consistently throughout the entire language, and that we are not just 
> focusing on the most convenient examples?      (024)

You are justified in putting that as a significant question.  The 
examples were not chosen for convenience but for their richness. I can 
point you to the TREC reference I supplied at the end of the message and 
I will bring forward these point to make it worth your while to look 
into those references:    (025)

In order for NIST judges to be sure the systems were faced with real 
world circumstances, government analysts prepared fifty different 
topical inquiries.  Statistics were collected on how each of the 
participating methods performed (by retrieving the documents judged 
relevant by the analysts)  in respect to each of the topic areas.  In 
the ad-hoc tests of TREC-8, software using methods implemented from 
Adi's axioms performed best overall in 23/50 topics and we did very well 
in most topics.    (026)

We had some trouble (not as good recall and/or precsison) in a 
half-dozen topics.  Overall, in response to 50 requests for documents 
--each topical inquiry referring to a different subject of discourse-- 
our software implementation returned up to 500 times the total numbers 
of relevant documents returned by any other participating system for any 
of the inquiries.  I read the experience as affirming that these 
patterns are observed meaningfully and consistently throughout the 
entire language.    (027)

> Does this apply only to words with Arabic roots?   And how do we 
> explain words like "cleave," that have two distinctly opposite 
> meanings, but are not morphologically or etymologically distinguished 
> in any way that would allow consistent inferences about the meaning of 
> the word?
I thought the examples offered have what anyone might infer as 
distinctly opposite meanings.   Words like this and their precise 
explainations are quite arbitrary.  There are many dual-use tools, of 
the mind, and that we keep as useful artifacts. A wood screw has a 
capacity to fasten and release. Some of these things are explicit as in 
the two distinct uses of the term 'cleave' and some are implicit, but in 
my view, they could not be hidden in any case.  We are not to forget or 
discount the context, situation or case in which they occur. The 
connection of such objects to a context is what is essential to meaning.    (028)

I cannot really imagine a circumstance where cleave would be used in its 
second more arcane sense, but if I were interested in specific cases I 
can distinguish them idiomatically rather than abstractly.  The idioms 
[cleave to] or [cleaving to] are regularly indexed with our methods.    (029)

I checked our conceptbase and found that cleave had only one sense to 
which it is assigined: crack; that being the common modern use.  It has 
14 relatives listed, including: crevice, fission, fissure, slot, split 
and rift.  My choices to distinguish one sense over another is the 
question.    (030)

The use of idioms seem suficient to me to distinguish the second sense 
of cleave.  If not, I can add the second sense of cleave to the system 
and recompile the conceptbase.  This would take about 15 minutes after 
entering "cleave=fasten" in a text file called the user conceptbase 
extensions and restarting the system.  Once this second sense is added, 
cases will arise where the user will need to  specify which sense should 
be applied.  That is why I would leave it as is.    (031)

I did not mean to imply that any of those words listed arose from Arabic 
roots.  The Arabic root is mentioned merely to show that the semantic 
classification has etymological justification.  Any resemblance due to 
the letters we used is purely incidental as those are not really Arabic 
letters, they are transliterations.  It should be clear though that 
there are sharted cognitive asapects.  Because the language is 
arbitrary, what else explains these cognates, surviving these thousands 
of years,  if not the shared universal abstract objects.    (032)

You will notice some entries without any Arabic roots.  It is because 
Tom Adi could not find one for the entry. If he did not find one, none 
probably exists.  We could not make a find enough Arabic roots for all 
the topics that can be delineated in English or in German.  Tom created 
several hundred constants standing in for canonical lexical roots where 
he deemed them necessary. It was also necessary aritificatally split 
some very large lexical segments to balance the system.    (033)

Morphology does not give us everything.  Neither do our abstract 
categories. They do generate semantic classifications that are rich, 
relevant and reliable. The next level is a uinfied and interrelated 
vocabulary that can also be arranged and linked by the semantic 
classifications provided at the higher level.  This makes a structure 
much like a thesaurus that can be used to make topic maps and can be 
mapped to ontologies. We also have a frames-based type/super-type 
framework for making finer distinctions with more precision-- that I 
have not talked about here.    (034)

Once the vocabulary is linked through the compiler, the upper level is 
immaterial except for theorizing perhaps.  No one reads a text for 
admiring the letters and orthography, well, hardly I suppose..  They  
read it to apprehend the abstract referents indicated by the signs and 
their morphological and grammatical structures.    (035)

I will refer back to the comment by Dr. Leo Orbst.    (036)

>a morphologizer that can generate or map to canonical lexical roots) enable 
>you to go from terms to word senses, which can then be linked to thesauri, and 
>the latter mapped to ontologies. 
I explained that we employed about 2000 lexical roots at the time of 
TREC-8. We have about 2750  now. And we have room in our data structures 
for 4000. The lexical roots link the terms into small collections we 
call concepts.  Some concepts are grouped and called super-concepts.  
Certain lexical stems are linked to one or more canonical lexical 
roots   Many proper names are mapped as constants.     (037)

We have an English lexicon of about 15000 word stems mapped to the 
lexical roots Tom Adi chose.  We have also mapped about 18,000 German 
word stems and roughly 10,000 French word stems to the lexical roots. 
This was a couple years of work all by itself.    (038)

By doing so we are able to compare concepts and ideas expressed in any 
of those languages -- not with three different computatioanl or data 
structures, but with a single unified system and interoperable data 
stucture.     (039)

To be absolutely clear and leave nothing to doubt, I am claiming that 
without any kind of translation and with a singile compilation, a 
topical inquiry expressed in English can be used to locate relevant and 
pertinent passages expressed in German or French, and vice-versa of 
course. Using these methods, we can map the terms of any language in 
this way.    (040)

> I still think there is something alien about this, which, I admit, is 
> a comment somewhat irrelevant to its scientific & linguistic validity. 
> Perhaps one should not be so hasty after all to dismiss ideas merely 
> on the basis of epistemological foreignness?  But theories should be 
> capable of advancing our knowledge by explaining both the convenient 
> and inconvenient real-world examples, which, no doubt, proves to be a 
> high threshold in the vast arena of human language.
> KL    (041)

Current events dictate that we be more capable of interpreting 
situations. The old 80/20 rule does not play in the detailed analysis of 
the situations we are faced with in this era and environment..    (042)

-Ken Ewell    (043)

Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki: 
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG    (044)
<Prev in Thread] Current Thread [Next in Thread>