Storming Media: Pentagon Reports and DocumentsPentagon Reports: Fast. Definitive. Complete.     
New Account »
Forgot Password?
Advanced Search »
Social SciencesLinguistics

Development of a Spoken Language System

Authors: John Makhoul; BBN SYSTEMS AND TECHNOLOGIES CORP CAMBRIDGE MA
Abstract:
The goal of this research project is to integrate speech and natural language technologies into a spoken language system capable of understanding and responding to spoken English for interactive human-machine military applications, such as command and control, training of military personnel, and logistics planning. The system we are building, called HARC (Hear And Respond to Continuous speech) will include a capability to adapt to new speakers and a capability to detect when a user says a new word. Major accomplishments this year include: 1) top performance in the first multi-site competitive evaluation; 2) speedup of the natural language processing component (Delphi); 3) rapid port to the ATIS domain; 4) extending the discourse mechanism, including the handling of more general forms of definite reference; 5) selection of the TRANSCOM domain for our demonstration SLS system; 6) integration of speech recognition and natural language processing; 7) detecting and adding new vocabulary words in speech; 8) several new more efficient search algorithms for speech processing; and 9) demonstration of real-time speech recognition with N-Best sentence output.

Limitations: APPROVED FOR PUBLIC RELEASE
Description: Annual rept. 1 Oct 89-30 Sep 90
Pages: 15
Report Date: 27 SEP 90
Contract Number: N00014-89-C-0008
Report Number: A399822
Keywords relating to this report:
*NATURAL LANGUAGE
*SPEECH RECOGNITION
ALGORITHMS
COMMAND AND CONTROL SYSTEMS
COMPETITION
DELPHI TECHNIQUES
EFFICIENCY
ENGLISH LANGUAGE
INFORMATION PROCESSING
INTEGRATION
INTERACTIONS
LANGUAGE
LOGISTICS PLANNING
MAN MACHINE SYSTEMS
MILITARY APPLICATIONS
MILITARY PERSONNEL
MILITARY TRAINING
PROCESSING
REAL TIME
SEARCHING
SITES
SPEECH
VOCABULARY
WORDS_LANGUAGE_
Email This Abstract