Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

mathematicians looking for things to do

0 views
Skip to first unread message

anal...@hotmail.com

unread,
Jan 16, 2012, 7:26:37 PM1/16/12
to
http://webdocs.cs.ualberta.ca/~kondrak/papers/thesis.pdf

I found this bit interesting:

"An early paper by Smith [1969] provides an example of a derivation
program
that deals with a much larger time distance. The goal was to derive
modern Russian
from reconstructed Proto-Indo-European forms. These two languages are
separated
by at least 5000 years of mostly undocumented changes, as compared to
about 1500
years that have passed since Latin split into Romance vernaculars. Out
of the total
of 650 Proto-Indo-European etyma that were examined, almost 90% have
left no
re
exes in Russian, and so could not be used to verify the correctness of
the program.
Of the remaining 69 etyma, the generated form exactly matched the
corresponding
Russian word in only 9 cases, which prompted Smith to rather gloomily
conclude that
\historical linguistics may have grossly overestimated the
exceptionless character of
sound change." The poor performance of the program, however, was most
likely also
due to the incompleteness of the implemented set of sound changes.
Appendix B
contains some results of my implementation for modeling the
phonological evolution
of Polish that contradict Smith's conclusion."

Nathan Sanders

unread,
Jan 16, 2012, 7:47:08 PM1/16/12
to
In article
<19083cae-a0b0-4dbb...@34g2000yqm.googlegroups.com>,
Which part? The part where Smith 1969 concludes that "historical
linguistics may have grossly overestimated the exceptionless character
of sound change", because he didn't implent enough sound changes in
his program? Or the part where Kondrak's 2002 results for Polish
"contradict Smith's conclusion"?

Nathan

--
Department of Linguistics
Swarthmore College
http://sanders.phonologist.org/

anal...@hotmail.com

unread,
Jan 24, 2012, 8:54:35 PM1/24/12
to
On Jan 16, 7:26 pm, "analys...@hotmail.com" <analys...@hotmail.com>
wrote:
And now this:

http://en.wikipedia.org/wiki/Quantitative_comparative_linguistics

we have come a long way from the explorations of Bopp et al - which,
however misguided, had some charm.

If the current state of the art is roots like "HHdhghwA" (fish) -
math can't make it any uglier. In fact the computer can be programmed
with parameters like how may "laryngeals" are mandaory how many can
occur togther etc. and protolanguages can be spun with one keystroke
and the computer can also be told that etyma that don't sound like
gargling with barbed wire are to be frowned upon.
0 new messages