RE: some notes

 

-----Original Message-----
From: pat hayes [mailto:phayes@ihmc.us]
Sent: Friday, September 26, 2003 9:26 PM
To: LYNN,JAMES (HP-USA,ex1)
Cc: public-sw-meaning@w3.org
Subject: RE: some notes



> Dictionary definitions are not really definitions in any foundational
sense:
> they are kind of sketches of a meaning which themselves rely on the
> same connected web of shared knowledge (some of it about language
> itself) which they set out to explain. URIs don't have this
> surrounding context of shared beliefs and so on; and in any case,
> URIs are not NL words.

I disagree.


? So you are saying that URIs *are* NL words?
No, I was simply disagreeing more with the statement that dictionary
definitions are not really definitions. I would concede that they are more
like indications (of usage?) more than definitions in a strict sense. Just
for the sake of argument though, in what way are they different from words?
If I see RDF as playing the role of NL for machines, how is the URI's role
in RDF different from a word in a true NL?


Although dictionary definitions do not reflect the same rigor as
a mathematical definition, they certainly provide enough definition to allow
a "common usage"


By human native speakers, sure. Although there is a great children's game
where one person gets to read out a definition from a dictionary (omitting
obvious clues such as synonyms and keywords) and the other has to guess the
word. It can be quite remarkably difficult: try this for example, from the
OED:

"One of the epidermal appendages of ...., usually in the form of a central
shaft or midrib, of a horny nature, in part tubular, for the rest square in
section and solid, fringed on either side with a .... row of thin narrow
plates mutually adpressed, which form a rounded outline at the end."

But the key point here is that a dictionary, like any other book, is
readable only by someone who already knows the language: and *any* adult
human being brings to bear an incredibly rich context of meaning and
interpretation to any linguistic act.  
 
How would you explain the Chinese Room scenario? I'm not necessarily
disagreeing here, just thinking out loud.
 
 There is no hope of having software do anything remotely like this in the
forseeable future; and even if we include human beings in the relevant
society of users, human native speakers don't have URIs attached to
information in their heads.


 According to http://www.merriam-webster.com/ a
pharmaceutical is defined as a medicinal drug. So even without context, I
should be able to glean enough knowledge to know that using the word

'pharmaceutical' could be used to refer to an aspirin, but not to my dog.


Not at all: think of the contextual knowledge in that very example. Asprin
is a drug; people don't eat dogs, drugs are used for medical purposes and
usually must be ingested in order to work, .... (For all I know, there may
have been times and places where dogs, or parts of dogs, may have been used
as pharmaceuticals.) Even the simplest act of linguistic comprehension is
based on so much context that no attempt to write it out formally has ever
succeeded. 
I'm still having trouble with how to handle anything but the "right now"
case. I've gotten different answers on this issue - the context of time,
maybe I just don't get it. 


While I certainly do't claim that if we had enough "URI dictionaries" we
could avoid all misuse, abuse, and disagreement, I can't help but think that
it would be beneficial to have registries play an authority role (perhaps
more descriptive than prescriptive).


I agree with you there.


As someone else already stated, isn't
the difference between words and URI's (simply viewed as tokens for the
moment) that while a NL word may have different meanings represented by the
same string (token), we insist that URI's reflect the different meanings by
assigning different strings to each meaning.



Well, this kind of claim is often made, but I don't think it stands up to
close examination, even when applied to existing examples of URI use on the
conventional Web. 
Not sure I understand you point without an example. I'm saying that if I
want to represent two different objects, let's say 
    1) the case where I shoot a hole of golf at 1 under par and 
    2) a winged animal,  
both of which might be names as a "birdie", I would be and should be
expected to use two different URIs, perhaps I should say that from a "social
meaning" perspective, it is deliberately (or at least negligently)
misleading to use the same URI. Are you saying that this doesn't stand up to
close examination?



..not so different from what is
used in referencing dictionaries as in pharmaceutical[1], pharmaceutical[2].


I think it is VERY different. For a start, the dictionary definitions don't
fix the meanings of words: they simply indicate them in a concise way.



I also disagree that

> NL takes ages to build up meaning through usage,

Certainly language as a whole has taken ages to reach its present state, but
new words pop into common usage overnite, btw.



NOt overnight, but they do appear, as do new usages of old words.  But this
hardly makes the case for the dictionary analogy: these new meanings are
only found in dictionaries later, if at all.  (There are some fascinating
studies of regional variations in everyday terminology, eg there are
something like a hundred different words for woodlice, none of them in
standard dictionaries and all used only locally in a geographical area.)


 OK, maybe one shouldn't
consider 'btw' a word, but there are plenty examples that are, not all from
the internet/web either. If a group of KB engineers start using some URI
'.../theNewFoo', it will quickly assume some meaning among that group and

any who have a need to converse with that group regarding theNewFoo.


But why would this group of engineers use a URI to converse? I don't know
anyone *that* geeky. Even engineers do usually talk using at least a  kind
of English.  Unless of course it was a genuine URI which actually did
something on the Web (or better, which the Web could do something with) and
that is why they are using it.  So its still not like a word: NL words don't
do anything on the Web.  
 
I guess I was thinking more along the lines of a conversation by proxy,
using semweb agents, where the engineers are the puppetmasters and the
conversations, and reasoning, is done by the agents. Maybe an e-commerce
example would be better? Anyway, the idea is that the engineers and agents
(I guess the engineers might converse at some point, not in RDF but perhaps
about the RDF - meta-RDF?) would have a conversation to negotiate or "come
to an understanding" about the meaning of 'theNewFoo'. Hey this stuff might
be good for Web Services afterall. ;)

BTW, even the WGs who have to be constantly be talking to one another about
URIs tend to distinguish a URI from what might be called its NL word-form.
Nobody actually says, and hardly anyone even writes,
"http://www.w3.org/2000/01/rdf-schema"; what we say to one another are
things like "are-dee-eff-ess", which itself tends to get shortened to a
monosyllable, and we invent shorthands like rdfs: to use in emails. 


But in any case, the analogy with NL suggests that any attempt to control or
even influence this by external fiat or authority isn't likely to succeed,
if this kind of thing is indeed anything at all like language. If you want
some real-world experience in trying to attach meanings to words by writing
authoritative registries, just ask the Academie Francaise about things like
"rosbif" and "bigmac". 
 
I agree, my original main point (lost in the trees, no doubt) was that
dictionaries are useful not because they fix definitions by fiat, but
because they record some sort of a consensus of meaning. I see this whole
meaning issue as something that should be addressed in a way which mirrors
what we do in the real world, where meaning is a complicated mix of
influences - cultural, psychological, linguistic, social, with the authority
of academia and dictionaries playing just a small but key part to making it
all work. In anticipation of anyone taking issue with whether it works, I
simply mean that results are realized, not that difficult issues never arise
again. Even set theory can't claim that!
 
James

Strategic Coordination - Enterprise Standards 
HP Software Services 
610 595 4995