DocumentCode :
3530691
Title :
Acoustically discriminative training for language models
Author :
Kurata, Gakuto ; Itoh, Nobuyasu ; Nishimura, Masafumi
Author_Institution :
IBM Res., IBM Japan, Ltd., Yamato
fYear :
2009
fDate :
19-24 April 2009
Firstpage :
4717
Lastpage :
4720
Abstract :
This paper introduces a discriminative training for language models (LMs) by leveraging phoneme similarities estimated from an acoustic model. To train an LM discriminatively, we needed the correct word sequences and the recognized results that automatic speech recognition (ASR) produced by processing the utterances of those correct word sequences. But, sufficient utterances are not always available. We propose to generate the probable N-best lists, which the ASR may produce, directly from the correct word sequences by leveraging the phoneme similarities. We call this process the ldquoPseudo-ASRrdquo. We train the LM discriminatively by comparing the correct word sequences and the corresponding N-best lists from the Pseudo-ASR. Experiments with real-life data from a Japanese call center showed that the LM trained with the proposed method improved the accuracy of the ASR.
Keywords :
speech recognition; training; Japanese call center; Pseudo-ASR; acoustically discriminative training; automatic speech recognition; language models; Acoustic applications; Acoustic transducers; Automatic speech recognition; Decoding; Equations; Laboratories; Natural languages; Telephony; Testing; Discriminative Training; Finite State Transducer; Language Model; Phoneme Similarity;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Acoustics, Speech and Signal Processing, 2009. ICASSP 2009. IEEE International Conference on
Conference_Location :
Taipei
ISSN :
1520-6149
Print_ISBN :
978-1-4244-2353-8
Electronic_ISBN :
1520-6149
Type :
conf
DOI :
10.1109/ICASSP.2009.4960684
Filename :
4960684
Link To Document :
بازگشت