Probabilistic Speech Recognition       


Artificial Intelligence Accomplishment | 1974

IBM researchers: Robert L. Mercer, Lalit R. Bahl, Raimo Bakis, Fred Jelinek

Where the work was done: T.J. Watson Research Center

What we accomplished: IBM developed the core approach to probabilistic speech recognition based on ideas from Information Theory. From Bob Mercer's address at the 2014 Lifetime achievement Award from the American Association of Computational Linguistics:  "Rather than emphasizing the linguistic understanding of speech, the group advocated probabilistic and information theoretic approaches. Despite their lack of linguistic background, the group in rapid succession produced breakthrough after breakthrough in various areas of computational linguistics, including few statistical approaches in speech recognition, developing some of the first large probabilistic language models, and finally pioneering the use of statistical approaches to machine translation, the problem of translating from one language to another."

Related links:

Bahl, Lalit R., and Frederick Jelinek. "Decoding for channels with insertions, deletions, and substitutions with applications to speech recognition." Information Theory, IEEE Transactions on 21.4 (1975): 404-411.

 

Jelinek, Frederick, Lalit R. Bahl, and Robert L. Mercer. "Design of a linguistic statistical decoder for the recognition of continuous speech." Information Theory, IEEE Transactions on 21.3 (1975): 250-256.
 

BACK TO ARTIFICIAL INTELLIGENCE

BACK TO IBM RESEARCH ACCOMPLISHMENTS