LINGUIST List 2.385

Monday, 5 August 1991

Disc: Military uses of text and speech

Editor for this issue: <>

Directory

1. Elizabeth A. Hinkelman, military uses of text and speech
2. Larry Reeker, Ervin-Tripp's Query on Military Uses
3. Koenraad De Smedt, DARPA LDC

>From: ervin-trcogsci.Berkeley.EDU (Susan Ervin-Tripp)
>Subject: Re: The Linguistic Data Consortium
>Can anyone clarify what the military uses are for the linguistic
>texts, exactly? There is a long history of military support of linguistic
>research. Please give some examples even artificial which can account
>for this interest. Some linguists have argued that getting military
>funding is a good way to remove funds from more malign uses.
Intelligence.
The NSA, CIA, and military intelligence groups would like to know
what's happening in the world, in detail. They would like
technology that could track the world's communication media, and
locate threats, insurgents, potential instabilities. For
instance, the DARPA-sponsored TIPSTER initiative, which is aimed at
automation of labelling document topics and answering questions
about their contents. It uses newspaper articles in English and
Japanese as the texts. Another example would be the DARPA-sponsored
message understanding workshops, the most recent of which used
newspaper articles about terrorism and asked, for example, what the
target of the attack was. Contact Beth Sundheim
(sundheimcod.nosc.mil) for reports on the message understanding
workshops. Current problems include proper names and anaphora.
The LDC in particular is part of Federal sponsorship of several (not
sure if all ten are administered by DARPA) "precompetitive
technologies" which will also be commercially sponsored. In this
case, the technology is robust parsing and speech processing
programs, which need to handle any word or construction that occurs
& need many instances of each for training. So the data we're
talking about might be, say, the entire Congressional Record and 40
million words of office memos from each corporate partner.
Office automation is a major application area. For linguistics
vresearchers, increased public funding will improve communication
on these topics even as military and civilian information
processing are furthered.
Is my voicemail the property of my employer?
Elizabeth Hinkelman
Center for Information and Language Studies
University of Chicago

Although I may not be privy to all military uses of linguistics, I have
been involved in quite a few over the years, and my feeling is that
they the usual areas of applied linguistics, with even a certain amount
of theoretical work funded. As Susan commented, there is a long
history of this interest. The outstanding work by famous linguists on
language pedagogy that was done during and after in WWII is an
example. In the early enthusiasm for mechanical translation during the
early 60's, military and intelligence organizations were major funders
of research; and to a certain extent, this funding was diverted to
theoretical and computational linguistics after the ALPAC report of the
mid-60's pointed in that direction. Both intelligence and military
agencies find it necessary to translatee a good deal of material, and
rapid throughput is more important than in, say, translating a novel,
so their interest is natural.
Aside from translation, the services and other military and
intelligence agencies also handle enormous message traffic, so they,
like other governemntal and business organizations, would like to be
able to process them automatically for purposes of dissemination,
database creation, summarization, etc. Beyond messages, they would
like to do the same for, say, newspaper articles. An example is found
in counter-terrorism, where the synthesis of material processed
automatically may provide one way of preventing incidents. The
services, in particular, are also interested in training applications
and in improving documentation. All such applications that I know of
are ones that many civilian organizations would find useful, too.
As far as I am aware, there are no lethal applications of linguistics
under consideration by the Department of Defense. The DoD, and
particularly DARPA, are funding a number of things that will likely
impact the civilian sector more than the military over the long term.
As far as I am concerned, this support is entirely beneficial to the
field.
		Larry Reeker
	Institute for Defense Analysis

I have received some private responses to my comment on the military
status of the DARPA Linguistic Data Consortium and of DARPA research in
general, so I want to readjust what I wrote in an earlier posting.
I did not want to imply that the net effect of funding research through
DARPA is bad or dangerous. On the contrary, I know that much of the work
funded by DARPA is fundamental rather than military research. I know
that most of the results are very accessible (perhaps more than
company-funded research), and I know that many of the applications are
civilian. In fact, the proposed Linguistic Data Consortium is a good
example of a research project that's primarily intended as
pre-competitive research with litte military but high economic value.
Fine.
But that is precisely why I find it odd that it's done by a military
agency. It's just odd for it to be somehow be associated with the
military. I'm not saying it's bad, because in practice, it may well be
better for the Linguistic Data Consortium to be run by DARPA than by
some quibbling group of computer companies.
Koenraad de Smedt