Nostratic et al. Part 2: The Challenge

[H. M. Hubey]

----------------------------Original message----------------------------
Alexis Manaster-Ramer wrote:

> (5) Of course, there are many other classification issues that need
> work, but most are not easily accessible.
> There ARE some simple mathematical issues that have not
> been solved and anyone who knows elementary probability
> theory and loves linguistics is invited.

Since my main interest is methodological and I do not
see myself memorizing thousands of words from
tens of languages, I am more than happy to find
this kind of work collected in one spot so more time
can be spent reading instead of driving to libraries.

I am happy to volunteer as someone who does do probability theory.

> most of the work being done in THIS field suffers from
> precisely the methodological problems which are laid by
> Trask, Thomason, et al. (usually incorrectly) at the door of
> the kind of linguistics that does deal with classification.
> The shoe is on the other foot, as I am prepared
> to document in detail if asked (and
> already have in various articles, notably in
> International Jo. of Dravidian Linguistics, IJAL, Georgica,
> JIES, etc.).

The least that should be done is to follow in the footsteps
of those like J. Nichols (and you) who have done work creating objective
measures of family relationship possibilities/probabilities
so that everyone can at least learn to deal with one or several
numbers on strength of classification families so that like
sociologists, psychologists, economists and other social scientists
historical linguists can also compare these numbers instead of their
gut feelings, and other apparently inexplicable ways in which they
get such strong feelings about certain things. (I have already done
a little in this respect in the History of Language journal and
much more in my book.)


> This then is theother challenge: let us
> educate a new generation to clean u[p
> the historical linguistics of well-established
> language families even as we try to test
> and refine theories that posit more far-flung
> families.
>

One of the ideal candidates for this is, fuzzy logic, which was constructed

practically for natural languages by L. Zadeh. There are
many things going for it. The first is that logic is easier
to learn than probability theory so that the basic meanings
of what means what can be learned in logic and then the
meanings extended to fuzzy logic. The second is that fuzzy
logic sort of sits between probability theory (which is
difficult for many) and logic. Furthermore, one can look at a
site which says plainly that probability theory is extended
logic:  http://bayes.wustl.edu. I don't think people exist who
will deny that that historical linguistics should be illogical
and irrational. All three forms of reasoning, above, then
can plainl be seen as about nothing else but logic and reasoning.
I hope that soon the age of having gut feelings on family
relationships will soon be over. I intend to do my part on this
and since I just got a paper accepted (about 2 days ago) to
the Journal of the International Quantitative Linguistics
Association, even those that only respect authority should not
have problems believing the seriousness of my efforts.

> Alexis Manaster Ramer
> Professor of Computer Science
> Wayne State University

--
M. Hubey

Associate Professor of Computer Science

Montclair State University

Email:          hubeyh at Montclair.edu    Backup:hubeyh at alpha.montclair.edu
WWW Page:       http://www.csam.montclair.edu/Faculty/Hubey.html