Data-selective Transfer Learning for Multi-Domain Speech Recognition

@inproceedings{Doulaty2015DataselectiveTL,
  title={Data-selective Transfer Learning for Multi-Domain Speech Recognition},
  author={Mortaza Doulaty and Oscar Saz-Torralba and Thomas Hain},
  booktitle={INTERSPEECH},
  year={2015}
}
Negative transfer in training of acoustic models for automatic speech recognition has been reported in several contexts such as domain change or speaker characteristics. This paper proposes a novel technique to overcome negative transfer by efficient selection of speech data for acoustic model training. Here data is chosen on relevance for a specific target. A submodular function based on likelihood ratios is used to determine how acoustically similar each training utterance is to a target test… CONTINUE READING
Recent Discussions
This paper has been referenced on Twitter 2 times over the past 90 days. VIEW TWEETS
8 Citations
33 References
Similar Papers

References

Publications referenced by this paper.
Showing 1-10 of 33 references

The HTK book (for HTK version 3.4)

  • S. Young, G. Evermann, +7 authors D. Povey
  • Cambridge university engineering department, vol…
  • 2006
Highly Influential
2 Excerpts

Similar Papers

Loading similar papers…