ALEX WAIBEL
Professor, Language Technologies, Human Computer Interaction and Computer Science
www

My research interests center around the interpretation and integration of speech, language and other human communication signals in the design of human-computer and human-human communication systems. Two particularly challenging examples of these interests are the JANUS project, a speech-to-speech translation system, that provides translation of spoken language between English, German, Spanish, Japanese, Korean (e.g., a translating telephone). Another, the INTERACT project, attempts to design multimodal interfaces, that incorporate not only speech, but also other communication modalities, such as gesture, lipreading, handwritten character recognition, face- and eye-tracking, in order to derive a robust understanding of user intent.

The JANUS project has provided one of the earliest demonstrations showing that speech-to-speech translation is possible. I am now working on extending the system's toward high quality translation of unlimited spontaneously spoken and ins eom cases conversational human speech. This involves new inroads in robust understanding of spoken language, improved speech recognition methods, ways to automatically learn language as well as interactive and/or mobile ways of deploying such a system to meet natural multi-lingual communication needs. We are also extending JANUS to accept and produce additional languages flexibly.

The INTERACT project is aimed at enhancing human-computer communication to include other communication modalities known to be helpful in human communicative situations, such as observing where a person might be looking or focusing attention, decoding his/her handwriting, gesturing and pointing in conjunction with speech, observing lip movement to enhance recognition of speech, etc. Several real-life human-computer interaction applications are explored to see how these modalities can be usefully captured to make them a natural and reliable part of any human-computer or communication interface. In an effort to achieving these practical goals, computer systems must be able to learn and adapt to a changing environment and growing demands of a user. I am interested in various machine learning and modeling strategies to develop learning and adaptation. I am interested in CONNECTIONIST and STATISTICAL LEARNING and ADAPTATION strategies, particularly as applied to speech, language, visual and interactive signal processing. I am collaborating with other faculty to develop and enhance learning algorithms, that have the required properties for deployment in the real world.