DocumentCode
3039581
Title
Further results on the recognition of a continuously read natural corpus
Author
Bahl, L.R. ; Bakis, R. ; Cohen, P.S. ; Cole, A.G. ; Jelinek, F. ; Lewis, B.L. ; Mercer, R.L.
Author_Institution
IBM Thomas J. Watson Research Center, Yorktown Heights, N.Y.
Volume
5
fYear
1980
fDate
29312
Firstpage
872
Lastpage
875
Abstract
Further results have been obtained on the recognition of continuously read sentences from a natural language corpus of laser patents. The vocabulary is limited to the 1000 most frequently occurring words in the corpus. Our model of the task language has a perplexity of 24.1 words (corresponding to an entropy of 4.6 bits/word). This paper describes modifications and improvements to the system which have resulted in the lowering of the word error rate from the previously reported 33.1% to 8.9%.
Keywords
Computer science; Discrete Fourier transforms; Entropy; Error analysis; Laser theory; Lead; Natural languages; Prototypes; Speech recognition; Vocabulary;
fLanguage
English
Publisher
ieee
Conference_Titel
Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP '80.
Type
conf
DOI
10.1109/ICASSP.1980.1170862
Filename
1170862
Link To Document