Abstract

A new scoring algorithm has been developed for generating wordspotting hypotheses and their associated scores. This technique uses a large-vocabulary continuous speech recognition (LVCSR) system to generate the N-best answers along with their Viterbi alignments. The score for a putative hit is computed by summing the likelihoods for all hypotheses that contain the keyword normalized by dividing by the sum of all hypothesis likelihoods in the N-best list. Using a test set of conversational speech from Switchboard Credit Card conversations, we achieved an 81 % figure of merit (FOM). Our word recognition error rate on this same test set is 54.7%. 1.

Citations

37	A tree-trellis based fast search for finding the n best sentence hypotheses in continuous speech recognition - Soong, Huang - 1990
34	GENONES: Optimizing the Degree of Mixture Tying in a Large Vocabulary - Digalakis, Murveit - 1994
21	high-performance algorithms for n-best search - Schwartz, Austin, et al. - 1990
20	Keyword-spotting using SRI's DECIPHER large-vocabulary speech recognition system - Weintraub - 1993
13	Large-Vocabulary Dictation Using SRI’s - Murveit, Butzberger, et al. - 1993
7	Reduced channel dependence for speech recognition - Murveit, Butzberger, et al. - 1992
4	Performance of SRI’s DECIPHER - Murveit, Butzberger, et al.