BBN Technologies

Since the early 1970s, BBN has been performing pioneering research in automatic speech recognition.

Over the years, BBN has had many firsts, including the first demonstration, in the early 1990s, of real-time, large-vocabulary, speaker-independent continuous speech recognition on commercial, off-the-shelf hardware.

BBN's Byblos, our primary speech recognition system, is an automatically trainable system that utilizes probabilistic hidden Markov models, and it continues to represent the state of the art in large-vocabulary, speaker-independent speech recognition.

The Byblos engine forms the core of our application suite that includes Audio Indexer and Audio Monitoring System. It has also been used for automated directory assistance. Current research programs continue to advance the state of our speech recognition technology and deliver significant improvements in recognition accuracy for speech in different environments, including telephony and broadcast news, and in multiple languages, including English, Arabic, Mandarin, and Spanish.