Abstract

This paper presents three different techniques applied in or developed during the 1996 Hub-4 broadcastnews transcription task. First, an efficient shortest path graph search algorithm is applied to the word lattice created by Viterbi search, producing a globally optimum result. This reduces the word error rate by about 3-10 % (relative), depending on the test set. The execution time is at or close to real time for most utterances. Second, a segmented N-best list generation algorithm is described for producing compact N-best lists for very long utterances. Finally, a temporal smoothing technique is compared to deleted interpolation. On one test set, temporal smoothing reduces the error rate by 3 % for an 8 % increase in search cost, while the latter improves by 6 % for a 50 % increase in search cost. 1.

Citations

62	Flexible speaker adaptation using maximum likelihood linear regression - LEGGETTER, WOODLAND - 1995
47	Large-Vocabulary Dictation Using SRI's DECIPHER TM Speech Recognition System: Progressive Search Technique - Murveit, Butzberger, et al. - 1993
17	Deleted interpolation and density sharing for continuous hidden Markov models - Huang, Hwang, et al. - 1996
14	A tree search strategy for large-vocabulary continuous speech recognition - Gopalakrishnam, Bahl, et al. - 1995
8	An Improved Search Algorithm for Continuous Speech Recognition - Alleva, Huang, et al. - 1993
8	The Optimal N-Best Algorithm: An Efficient Procedure for Finding Multiple Sentence Hypotheses - Schwartz, Chow - 1990