Dr. Mazin E. Gilbert
Mazin E. Gilbert, Ph.D., MBA is Executive Director of Technical Research at AT&T Labs. His responsibilities include managing research and development in the areas of automatic speech recognition, natural language processing, web and speech mining, and multimodal voice search. His business areas of focus include product strategy and development, entrepreneurship, and corporate finance. He is the recipient of the AT&T Science and Technology Medal Award (2006).
His major projects include:
- WATSON Speech Recognition: Research in robust large-vocabulary speech processing, acoustic, and language modeling of speech. The project involves software development of next-generation plugin architecture to support a variety of voice applications including those for mobility, IPTV, and call center automation. WATSON was recently licensed to Vlingo, and received the SpeechTechMag.com, Leader award.
- Multimodal Voice Search: The integration of VOIP with graphical browsers on desktop and mobile devices enables a new generation of multimodal services that support user input and system output over multiple modes such as speech and pen. Check out AT&T’s YellowPages Mobile Voice Search Applications from the Apple store including Speak4it!,YPMobile, and ChaCha.
- Natural
Language Search and Web
Mining:
Converting the World Wide Web into a structured set of
information
for the purpose of extracting intelligent information, and the
creation of interactive chat-based or spoken dialog agents. The
project involves research in question/answering, text
summarization,
supervised and unsupervised methods for active learning, active
labeling and active evaluation, and information search from
conversational speech, documents, and
websites.
- Speech
Translation:
Speech-to-speech translation, human/machine translation,
multilingual text, and speech interfaces to existing
applications.
These applications range from human-machine dialog systems
(e.g.,
information access systems) to human-human dialog systems (e.g.
instant messaging).
- Data Analytics: Machine learning, statistics and information search methods for analyzing speech, text and social media data. His division developed the Talkalytics web services tool that is currently supporting AT&T Mobility, and is currently developing Sonar, a social media analytics and visualization tool.
- Spoken Language Services: Research and development into next generation conversational dialog systems including spoken language understanding, dialog management, and large vocabulary speech recognition. This technologies are currently driving AT&T VoiceTone, which specializes in creating sophisticated spoken-language dialog applications for large-business customers. VoiceTone has received many industrial awards including most recently the SpeechTechMag.com, Professional Services award.
He was a Research Professor with James Flanagan at the CAIP Center, Rutgers University during 1991/1992. He has over 100 publications in the area of speech and language processing, holds 36 US patents, and has over 50 patents submitted. He is a member of the IEEE Signal Processing Magazine Editorial Board and the ISCA Advisory Board.
His patents include Method For Building A Natural Language Understanding Model For A Spoken Dialog System, Active Learning Process For Spoken Dialog Systems, Active Labeling For Spoken Language Understanding, Speech Recognition Over Lossy Networks With Rejection Threshold, Reducing time for annotating speech data to develop a dialog application, and Spoken language understanding that incorporates prior knowledge into boosting. Read the full list of his patents!
Watch From the Labs: The Art of the Possible. Read Computers Learn to Listen, and Some Talk Back and Talking to Your Phone. Read his LinkedIn profile.