IBM T.J. Watson Research Center

IBM T.J. Watson Research Center’s pioneering techniques and continued innovations over the last 35 years have made speech recognition applications an indispensable part of everyday life. IBM T.J. Watson Research Center was established in 1961 and has its main laboratory in Yorktown Heights, New York. Its research ranges from exploratory work in the physical sciences to semiconductors and systems technology to software for security, programming, mathematics, and speech technologies.

IBM commercialized the technology in the 1990s with its Simply Speaking and ViaVoice products, allowing PC-based dictation enabling users to create documents without typing. In 2003 IBM introduced an embedded version of ViaVoice providing fully integrated automatic speech recognition to enable interaction for small mobile devices such as voice-directed applications in automobiles. In early 2000s IBM delivered Websphere Voice Server, a telephony version of its speech recognition technology, for call center interaction applications such as directory assistance. IBM’s worldwide R&D labs have ported these products to a variety of languages and domains.

IBM has also combined speech technologies with other technologies in innovative fashions. IBM’s speech translation applications are used by the U.S. military to more easily communicate with the native population using portable computers. IBM has augmented speech recognition with video information that captures lip movements to enhance recognition performance in noisy environments. IBM continues to advance the speech recognition area by developing promising research directions such as discriminative training concepts, statistical language model estimation and discriminative feature extraction to improve overall performance.