Sequence training of multi-task acoustic models using meta-state labels

@article{Siohan2016SequenceTO,
  title={Sequence training of multi-task acoustic models using meta-state labels},
  author={Olivier Siohan},
  journal={2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  year={2016},
  pages={5425-5429}
}
In this paper, we describe a multi-task learning approach for acoustic modeling where the multiple output layers are used to predict context-dependent (CD) states from different state inventories. Unlike the traditional multitask learning approach which defines a primary and secondary output layers but discards the secondary output after training, we propose to use all output layers for recognition. This can be achieved by designing a decoding network operating on tuples of CD states and… CONTINUE READING