Name
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 “Pseudo-ASR”. 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. |
Year | DOI | Venue |
|---|---|---|
2009 | 10.1109/ICASSP.2009.4960684 | ICASSP |
Keywords | Field | DocType |
japanese call center,language model,acoustically discriminative training,automatic speech recognition,probable n-best list,corresponding n-best list,discriminative training,phoneme similarity,correct word sequence,acoustic model,lm discriminatively,testing,data mining,natural languages,speech,accuracy,decoding,speech recognition,language models,telephony,hidden markov models,finite state transducer | Computer science,Natural language processing,Artificial intelligence,Discriminative model,Finite state transducer,Language model,Pattern recognition,Speech recognition,Natural language,Decoding methods,Telephony,Hidden Markov model,Acoustic model | Conference |
ISSN | Citations | PageRank |
1520-6149 | 5 | 0.50 |
References | Authors | |
13 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Gakuto Kurata | 1 | 107 | 19.06 |
| Nobuyasu Itoh | 2 | 65 | 13.19 |
| Masafumi Nishimura | 3 | 112 | 22.77 |