BBN Technologies

Large Vocabulary Speech Recognition for Transforming Speech into Rich Text

San Francisco, CA. February 22, 2005 — BBN Technologies, a leading provider of advanced technology and research and development, announced the availability of its AVOKE™ STX speech-to-text system, which transforms audio into searchable content, enabling a new era of multimedia search. The technology was demonstrated today at SpeechTEK West.

AVOKE STX technology converts the speech in audio and video into a rich transcription that can be searched and analyzed easily. At the core of the system is BBN's very large vocabulary speech recognition engine, which is the most accurate, speaker independent, continuous speech recognition engine commercially available in the world. Applications for this technology include enterprise search, government intelligence, consumer search, broadcast news monitoring, and call center recording, among others.

"AVOKE STX unlocks the information in speech and accelerates access to the wealth of audio and video content on the web, in live radio and television broadcasts, and in stored archives," said Alex Laats, president of BBN Technologies Delta Division. "Our advanced speech technology delivers audio in a format that is easily understood by standard search engines. This allows us to take audio and put it directly into users' hands as searchable content."

The AVOKE STX system accepts a variety of audio input sources, including television, radio, telephone, meetings, webcasts, and speeches, among others, and then outputs a rich transcription of the audio that includes both the words and metadata in XML format. The benefit of this approach is that audio information is delivered to desktops in the same way that any text information is delivered, using familiar web based interfaces.

BBN's AVOKE STX system works by separating speech from non-speech, such as music or laughter, and then processing the speech to identify additional characteristics. Each word is time-stamped, which, when delivered with the search results, allows users to navigate to any point in the broadcast or conversation instantly. The user can listen to the original audio or watch the corresponding video directly from the search result, giving them ultimate control over access to their data.

The system provides further analysis and information of the audio data. The speech is segmented and marked according to speaker turns, the speaker is identified, and named entities, such as location, person or organization, are also identified. This additional information is captured in metadata and indexed so that it can be used by any standard text searching techniques. The software itself is language and domain independent and can be configured to run on many different types of data in the same environment.

BBN's AVOKE Suite of Speech and Language Products
BBN's AVOKE suite of speech and natural language products and services enables solutions that maximize the useful information content in speech. For more information, visit www.bbn.com/avoke.

About BBN Technologies
BBN Technologies, an advanced technology solutions firm, is focused on solving some of the world's most pressing problems. From national security, information security, speech recognition and language translation, to integrating disparate systems and networks, BBN has been at the forefront of technological change for over 50 years.

Known for pioneering the development of the ARPANET, the forerunner of the Internet, BBN continues to create advances in Internet and networking technologies through its work on ad hoc networking, the semantic web, quantum communications, and advanced protocols. Building on its substantial list of firsts, BBN operates the first metro quantum cryptography network, the first real-time foreign broadcast monitoring system, and has developed the world's first stereoscopic digital mammography system. For more information on BBN Technologies, visit www.bbn.com.