eduzhai > Applied Sciences > Engineering >

Future Vector Enhanced LSTM Language Model for LVCSR

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: Language models (LM) play an important role in large vocabulary continuousspeech recognition (LVCSR). However, traditional language models only predictnext single word with given history, while the consecutive predictions on asequence of words are usually demanded and useful in LVCSR. The mismatchbetween the single word prediction modeling in trained and the long termsequence prediction in read demands may lead to the performance degradation. Inthis paper, a novel enhanced long short-term memory (LSTM) LM using the futurevector is proposed. In addition to the given history, the rest of the sequencewill be also embedded by future vectors. This future vector can be incorporatedwith the LSTM LM, so it has the ability to model much longer term sequencelevel information. Experiments show that, the proposed new LSTM LM gets abetter result on BLEU scores for long term sequence prediction. For the speechrecognition rescoring, although the proposed LSTM LM obtains very slight gains,the new model seems obtain the great complementary with the conventional LSTMLM. Rescoring using both the new and conventional LSTM LMs can achieve a verylarge improvement on the word error rate.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×