Speech Recognition: Where Do We Go From Here?

ABSTRACT:
Speech recognition has recently achieved a major milestone. There are now products available in every retail computer store to do large vocabulary continuous speech dictation on a personal computer. However, although it has taken over 30 years to achieve this milestone, it is important to understand it not as the final goal of all the work we have done so far, but rather as the first step of all the work remaining before us.

Speech is the easiest and most natural means for people to communicate with other people. The goal for speech recognition research is to make speech the easiest and most natural means for people to communicate with computers and other machines and appliances -- in all places and all appropriate circumstances. Continuous dictation software is only the beginning. It does only one task. It does not understand or act on the contents of the speech it transcribes. It does not yet work well enough speaker independently. It is designed to recognized careful, intentionally dictated speech, not ordinary conversational speech. However, these limitations are only temporary.

Over the next few decades, all these limitations and others will be overcome. You will be able to talk to your digital personal assistant just the way you would talk to a human assistant. Your personal assistant will fit in your pocket and will go with you everywhere. It will be your access point to a world wide network of computer resources. It will also be a personal communicator, providing voice, text, data and video communication with other people around the world. It will provide peech-to-speech translation -- you will be able to communicate in other languages almost as easily as in your native language.

All these things are possible -- they all will happen. But to make them happen, we have as much hard work to do in the next 30 years as we have done in the last.

SPEAKER BIO:
James K. Baker, Ph.D., is Co-Founder and Chairman/Chief Executive Officer of Dragon Systems, Inc. Jim oversees Dragon Systems' research and defines new business directions. As the company's chief technical officer, he has been instrumental in positioning Dragon Systems as the industry's premier developer and marketer of speech recognition technology. Jim dedicates considerable time to performing research first hand to advance the methodologies he pioneered 20 years ago.

Jim's background is in applied mathematics. He introduced the efficacy and power of stochastic processing techniques and Hidden Markov Models to the field of speech recognition where they are now widely accepted. Jim was a member of the research staff at the IBM Thomas J. Watson Research Center, where he contributed to the Continuous peech Research project. He was also Vice President of Advanced Development at the Verbex division of Exxon Enterprises, which produced a continuous speech recognition product. Jim received a A.B. in Mathematics at Princeton University, where he was valedictorian of his class, and a Ph.D. in Computer Science from Carnegie-Mellon niversity, where he developed the original DRAGON speech recognition system under the auspices of the government's Advanced Research Projects Agency (ARPA) Speech Understanding Research (SUR) Program, under the supervision of Raj Reddy as thesis advisor.

Return to Inventing the Future Home Page