Language and Speech Meets 
the HPCC Grand Challenge 
Y. T. Chien 
National Science Foundation 
Washington, DC 20550 
At a recent NSF workshop on speech understanding, 
Charles Wayne gave a presentation of Darpa's program 
on Speech and Language research. Towards the end of 
his presentation, he enlightened all of us with a brief 
outline of a possible new technology development pro- 
gram: High Performance Multi-lingual Speech and Text 
Processing. This is a very exciting and timely initia- 
tive, if it comes to that, as many of us in the federal 
government, NSF included, are preparing to launch new 
efforts to implement the President's High Performance 
Computing and Communications (HPCC) program be- 
ginning in this fiscal year. 
The HPCC program is a real challenge and opportu- 
nity for all of us. Many in the research community , 
however, perceive "High Performance" as merely pro- 
viding supercomputer cycles and services or developing 
new computer architectures and software for scientific 
or engineering research. As such language and speech 
research, they feel, will have very limited role to play. 
I beleive this perception is incorrect and unhealthy. In 
fact, language and speech is one of the most critical ele- 
ments of the HPCC program for at least two reasons. 
First, one of HPCC's goals is to address "Grand Chal- 
lenge" problems. These are the fundamental problems 
in science and engineering, with broad economic and sci- 
entific impact, whose solution could be advanced by ap- 
plying high performance computing techniques and re- 
sources. In this context, "language and speech" has 
been identified as one of those grand challenge prob- 
lems, ranking it among such problems of national con- 
cern as weather prediction, drug design, superconduc- 
tivity, human genome, and transportation systems and 
others \[1,2\]. The implication here is that computational 
speech and language is a critical problem for us to solve 
and its solution could be advanced by further research 
in high performance computing. 
Another and perhaps more important reason, in my 
view, is that speech and language can help refine and 
drive HPCC research - ranging from architectures, soft- 
ware, algorithms and future novel applications as we 
move towards a knowledge intensive society. Here again, 
Charles Wayne's list of possible objectives for future 
speech and language research tells us why: Rapid, ef- 
fortless human-machine interaction vis speech, text, and 
other modahties, unlimited vocabulary speech recogni- 
tion, natural speech syntheses, real time translation of 
speech/language, etc. Few other topics in the comput- 
ing and communications field offer an equally rich set of 
research opportunities, with measurable goals and po- 
tential benefits. 
At NSF, we are also beginning the implementation of 
the HPCC program. One of the components is a new 
"Grand Challenge" Applications Groups program focus- 
ing on cross-disciplinary research to address problems 
such as computational speech and language. Other sim- 
ilar HPCC initiatives may further develop, but their 
strategies and contents are likely more driven by the 
visions of the research community. Since HPCC is a 
multi-agency program, it will also provide a rare oppor- 
tunity for Darpa and NSF to coordinate their goals and 
strategies to support research in speech and language. 
This is a challenge not only for the two agencies but for 
the research communities as well. 
References: 
1. "The federal High Performance Computing pro- 
gram", a report submitted to the Congress by the 
Office of Science and Technology, September, 1989. 
2. "Grand Challenges 1993: High Performance Com- 
puting and Communications," a report by the FCC- 
SET Committee to supplement the President's fiscal 
year 1993 budget. 
449 
