I'm a Senior Systems Scientist in ISRI,
School of Computer Science at
Carnegie Mellon University, currently on leave
of absence, and working at Carnegie Speech.
(Contact info at the bottom of this page.)
In speech recognition, I have worked on real-time recognition algorithms for large-vocabulary, speaker-independent systems. As a result of this work, the CMU Sphinx speech group has made the Sphinx-II and Sphinx-3 recognizers available publicly.
Speech recognition has provided a wonderful setting for a wide range of interesting problems in statistics and estimation. This is important to me in its own right, as is speech recognition as an application of this technology.
The technology has matured sufficiently to allow many commercial applications to be developed and deployed. However, it is still brittle. So that any such deployment requires an inordinate amount of handcrafting and fine tuning. I am interested in understanding the sources of the brittleness in speech recognition systems. This includes acoustic modeling as well as language modeling.
With regards to multiprocessor architecture, my main focus has been in designing large-scale cache-coherent shared-memory processor systems. Most recently, I was associated with the Piranha project at the Compaq Western Research Laboratory in Palo Alto, CA. Unfortunately this project was cancelled owing to business reasons. The main features of this project have been presented in ISCA 2000.