On 10/4/2011 5:28 AM, Rob Freeman wrote:
> Good luck selling your software John. I always respect your
> advocacy for subjectivity in logical representations. (01)
Thanks. (02)
But I want to emphasize that I did not use the words subjective
or subjectivity. In one sense, anything that anybody thinks is
subjective to them. But neuroscience and psycholinguistics can be
just as objective as any other science. I cited them to explain
how the framework is related to what happens inside the brain. (03)
In any case, I wanted to summarize why meaning (as processed in
the human brain or a computer system) cannot be represented by
statistical vectors, such as LSA: (04)
1. Children and adults can begin to use a word correctly on the
first occasion that they hear it. They can use it even though
they have no backlog of statistical data. (See slide 44 of
http://www.jfsowa.com/talks/ontofound.pdf ) (05)
2. As people learn to use the word in more "language games"
(or contexts or sublanguages), the word acquires a new
"microsense" for every game. (See slides 39 to 42.) (06)
3. The fact that people can create new microsenses implies
that they must have some basis for meaning that is prior
to and more fundamental than the accumulated statistics. (07)
4. Statistical methods, such as LSA and others, can be useful
for finding an appropriate language game (or context), but
the statistical vectors are not the basis for meaning. (08)
5. Implementing the full complexity of the human body and brain
would be necessary for using language in exactly the same
way that people do, but good approximations for many useful
"language games" can be and have been implemented on digital
computers. (See slides 60 and 75 to 105.) (09)
There is nothing subjective about that argument. In effect, I'm
proposing a kind of model-theoretic semantics, but one that is
much more dynamic than Tarski's and resembles human usage more
closely and naturally. (010)
John (011)
PS: If you're using the Adobe reader, you don't have to page through
the slides, just type the page number in the little window. (012)
_________________________________________________________________
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 (013)
|