LINGUIST List 9.1247

Wed Sep 9 1998

Sum: "UNIQUE NLP" (and CONCLUSION)

Editor for this issue: Martin Jacobsen <martylinguistlist.org>


Directory

  1. Anne Sing, "UNIQUE NLP" (and CONCLUSION)

Message 1: "UNIQUE NLP" (and CONCLUSION)

Date: Tue, 8 Sep 1998 14:57:33 -1000 (HST)
From: Anne Sing <anneshtdc.org>
Subject: "UNIQUE NLP" (and CONCLUSION)

SUMMARY

About two weeks ago I requested information concerning NLP software
that offered the ability to do a q&a exchange between animations and a
user similar to the "ChatterBox" software (free) we have at
http://www.haptek.com. Here is a summary of the responses I got and a
short commentary on each. In addition, I have added a few that I have
found via Microsoft concerning a parrot named "Perdy" and similar
characters. Take a look for yourself and you can see the degree to
which these different companies and research institutions are making
it possible to chat with computers and animations for fun, internet
searches, and information exchange.

1. You should have a look at the work by Boris Katz at the MIT AI
lab. He's been working on a system called START for some time that
does such things and more. See the web page at
http://www.ai.mit.edu/projects/infolab/ for more info.

2. [From Eduard Hovy (hovyisi.edu)]
There is, and has been for about two decades, an active research
group in NLP at USC's Information Sciences Institute (ISI). The
PENMAN system, a natural language *generator*, was developed from
circa 1979 to 1986 in a project led by Bill Mann and Christian
Matthiessen; Mann has since retired and Matthiessen is a professor of
Linguistics at Macquarie University in Melbourne, Australia
(cmatthiepip.engl.mq.edu.au).
 
The NLP Group at ISI currently contains several complementary
projects. More recent research at ISI includes:

A. Machine Translation (the Japangloss/GAZELLE system): Japanese,
Arabic, and Spanish to English, using a hybridization of statistical
and symbolic techniques. Contact Kevin Knight (knightisi.edu).
 
B. Text Summarization (the SUMMARIST system): multilingual
summarization using several complementary techniques, including
discourse structure. Contact Eduard Hovy (hovyisi.edu) and Daniel
Marcu (marcuisi.edu).
 
C. Multilingual Information Management (the MuST and C*ST*RD systems):
this work includes multilingual Information Retrieval, clustering, and
so on; contact Chin-Yew Lin (cylisi.edu).
 
D. Construction of large ontologies and lexicons: the semi-automated
merging of large ontologies, and the enhancement of these using
information extracted out of dictionaries and the web; contact Eduard
Hovy (hovyisi.edu).
 
E. Sentence Planning and Generation (the HealthDoc Sentence Planner
and the NITROGEN system): the pre-realization stage of microplanning
sentence structure and content, using a variety of methods; contact
Eduard Hovy (hovyisi.edu). For realization, NITROGEN combines a
phrase structure grammar with a bigram model of English to generate
from underspecified input; contact Kevin Knight (knightisi.edu).
 
More information on all this is available from
 http://www.isi.edu/natural-language/nlp-at-isi.html

[End Hovy]

3. Kevin Lenzo has a bot named url which stores information and
answers questions online, from multiple users, phrased similarly.

url hangs out in MUSHes, but I don't remember which ones. It seems
that you developed a kind of chatterbot or digital secretary. There
are many chatterbots developed. You can see the other chatterbots
'http://www.toptown.com/hp/sjlaven/'

4. The Microsoft and several other industry sites for this research
are at:

http://www.research.microsoft.com/research/ui/persona/home.htm

http://www-csli.stanford.edu/csli/projects/interface9495-srct.html

http://merl.co.jp

http://www.csl.sony.co.jp

http://www.cs.cmu.edu/afs/cs/cmu.edu/misc/mosaic/common/omega/web/frontdoor.html

http://www.cc.gatech.edu/gva/gvatop.html

http://www-ksl.stanford.edu/projects/cait/index.html

CONCLUSION While all these sites have interesting applications for
speech and animations they do not have the ability to put in factual
information and then query for that information. Usually they have a
key word search ability which allows them to return particular
paragraphs from a body of data based on the key words of a query.
However, to ask and answer questions like the following only seems to
be possible with ChatterBox at http://www.haptek.com. There were no
announcements of the development of such technology at these other
companies.

Who was the first president of the United States?
Who invented the telescope?
When did Columbus come to America?
Hey Mickey, Where did you find that treasure map?
What is your email address?
What is your fax number? 

and so on. 

Philip A. Bralich, President
Ergo Linguistic Technologies
2800 Woodlawn Drive, Suite 175
Honolulu, HI 96822
tel:(808)539-3920
fax:(880)539-3924
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue