DocumentCode :
2255865
Title :
The use of shibboleth words for automatically classifying speakers by dialect
Author :
Huggins, A. W F ; Patel, Yogen
Author_Institution :
BBN Hark Syst., Cambridge, MA, USA
Volume :
4
fYear :
1996
fDate :
3-6 Oct 1996
Firstpage :
2017
Abstract :
Real-world applications using speech recognition must perform well over a range of dialects. Differences in dialect between the speakers in the training database and the target users often leads to degraded recognition performance. For the BBN Hark hidden Markov model (HMM) based system, we have already developed a reasonably effective technique for dealing with multiple US dialects. The solution involves building separate HMM sets for each dialect from representative training speech data. This requires that training speakers be accurately classified by dialect, which is difficult to do reliably even by hand. In this paper, we describe a recognition-based pseudo-automatic scheme for partitioning a pool of American English training speakers into groups, such that the speakers within each group share the same pronunciation characteristics. Our scheme is speech-data driven, and involves using transcript-level word hypotheses generated by a recognizer to partition the pool of training speakers
Keywords :
hidden Markov models; linguistics; pattern classification; speech recognition; American English dialects; BBN Hark hidden Markov model-based system; automatic speaker clasification; degraded recognition performance; dialect differences; pronunciation characteristics; recognition-based pseudo-automatic scheme; shibboleth words; speaker set partitioning; speech-data driven scheme; training database; training speakers; transcript-level word hypotheses; Character recognition; Control systems; Databases; Degradation; Engines; Hidden Markov models; Probability density function; Speech recognition; Target recognition; Testing;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Spoken Language, 1996. ICSLP 96. Proceedings., Fourth International Conference on
Conference_Location :
Philadelphia, PA
Print_ISBN :
0-7803-3555-4
Type :
conf
DOI :
10.1109/ICSLP.1996.607194
Filename :
607194
Link To Document :
بازگشت