Locating and correcting erroneously recognized portions of utterances by rescoring based on two n-best lists
作者： Alex H Waibel , Arthur E McNair
DOI: 10.1121/1.424200
关键词:
摘要: A method of repairing machine-recognized speech is comprised the steps receiving from a recognition engine first n-best list hypotheses and scores for each hypothesis generated in response to primary utterance be recognized. An error within having highest score located. Control signals are which input constrain generation second hypotheses, hypothesis, an event independent utterance. The combined with list. selected as replacement located error.
lens.org 参考文章(4)
Wu Chou, Recognition unit model training based on competing word and word string models Journal of the Acoustical Society of America. ,vol. 102, pp. 1284- ,(1993) , 10.1121/1.420029
O. Schmidbauer, J. Tebelskis, An LVQ based reference model for speaker-adaptive speech recognition [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing. ,vol. 1, pp. 441- 444 ,(1992) , 10.1109/ICASSP.1992.225877
B. Suhm, B. Myers, A. Waibel, Interactive recovery from speech recognition errors in speech user interfaces international conference on spoken language processing. ,vol. 2, pp. 865- 868 ,(1996) , 10.1109/ICSLP.1996.607738
Richard M. Schwartz, Stephen C. Austin, Word dependent N-best search method ,(1991)