Semi-Supervised Model Training for Unbounded Conversational Speech Recognition

@article{Walker2017SemiSupervisedMT,
  title={Semi-Supervised Model Training for Unbounded Conversational Speech Recognition},
  author={Shane Walker and Morten Pedersen and Iroro Orife and Jason Flaks},
  journal={CoRR},
  year={2017},
  volume={abs/1705.09724}
}
For conversational large-vocabulary continuous speech recognition (LVCSR) tasks, up to about two thousand hours of audio is commonly used to train state of the art models. Collection of labeled conversational audio however, is prohibitively expensive, laborious and error-prone. Furthermore, academic corpora like Fisher English (2004) or Switchboard (1992) are inadequate to train models with sufficient accuracy in the unbounded space of conversational speech. These corpora are also timeworn due… CONTINUE READING
Tweets
This paper has been referenced on Twitter 26 times. VIEW TWEETS

From This Paper

Figures, tables, results, and topics from this paper.

Key Quantitative Results

  • We show relative word error rate (WER) reductions of {35%, 19%} on {agent, caller} utterances over our seed model and 5% absolute WER improvements over IBM Watson STT on this conversational speech task.

References

Publications referenced by this paper.
SHOWING 1-10 OF 31 REFERENCES

english conversational telephone speech recognition system

  • S. Thomas, M. L. Seltzer, K. Church
  • 2016

Similar Papers

Loading similar papers…