ISCA Archive Interspeech 2021
ISCA Archive Interspeech 2021

Investigating Methods to Improve Language Model Integration for Attention-Based Encoder-Decoder ASR Models

Mohammad Zeineldeen, Aleksandr Glushko, Wilfried Michel, Albert Zeyer, Ralf Schlüter, Hermann Ney

Attention-based encoder-decoder (AED) models learn an implicit internal language model (ILM) from the training transcriptions. The integration with an external LM trained on much more unpaired text usually leads to better performance. A Bayesian interpretation as in the hybrid autoregressive transducer (HAT) suggests dividing by the prior of the discriminative acoustic model, which corresponds to this implicit LM, similarly as in the hybrid hidden Markov model approach. The implicit LM cannot be calculated efficiently in general and it is yet unclear what are the best methods to estimate it. In this work, we compare different approaches from the literature and propose several novel methods to estimate the ILM directly from the AED model. Our proposed methods outperform all previous approaches. We also investigate other methods to suppress the ILM mainly by decreasing the capacity of the AED model, limiting the label context, and also by training the AED model together with a pre-existing LM.


doi: 10.21437/Interspeech.2021-1255

Cite as: Zeineldeen, M., Glushko, A., Michel, W., Zeyer, A., Schlüter, R., Ney, H. (2021) Investigating Methods to Improve Language Model Integration for Attention-Based Encoder-Decoder ASR Models. Proc. Interspeech 2021, 2856-2860, doi: 10.21437/Interspeech.2021-1255

@inproceedings{zeineldeen21_interspeech,
  author={Mohammad Zeineldeen and Aleksandr Glushko and Wilfried Michel and Albert Zeyer and Ralf Schlüter and Hermann Ney},
  title={{Investigating Methods to Improve Language Model Integration for Attention-Based Encoder-Decoder ASR Models}},
  year=2021,
  booktitle={Proc. Interspeech 2021},
  pages={2856--2860},
  doi={10.21437/Interspeech.2021-1255}
}