The method is interesting and, if
refined, will surely find some application in linguistics, but I don't
think I'll be out of a job very soon. First, there is more to historical
linguistics than drawing family trees. Secondly, the method has
yielded no novel classificatory proposals as yet; it has merely reproduced the
familiar clusters. All that we have at present is a promissory note
that may or may not be convertible into "impressive results" in the future.
Thirdly, the method is not reconstructive; it has no ambition to
generate historical narrative (protoforms, sound changes), and seems to be in
principle unable to factor out convergence (borrowing, areal traits). This
shortcoming is much more serious in linguistic applications (where
large-scale lateral diffusion has to be taken into account) than in
reconstructing DNA cladograms. Fourthly, it seems to work best with long
parallel texts, and unfortunately we have no Thracian, Luwian or Tocharian
translation of the Universal Declaration of Human Rights.
Piotr
----- Original Message -----
Sent: Saturday, March 02, 2002 5:44 PM
Subject: [tied] No need for IE database, computers with zipped files
are coming
http://www.economist.com/science/displayStory.cfm?story_id=975770
Analysing
compressed data leads to impressive results in
linguistics
ZIPPING, as
any computer buff knows, enables you to compress a
file so that it can be
stored efficiently, or sent quickly over
the Internet. But Emanuele Caglioti
and his colleagues at the
University of Rome-La Sapienza have found a more
esoteric use for
it. Using zipped files, they can identify the authors
of
documents and reconstruct the family trees of
languages.