Effects of Layer Freezing on Transferring a Speech Recognition System to Under-resourced Languages

Onno Eberhard · Torsten Zesch

In this paper, we investigate the effect of layer freezing on the effectiveness of model transfer in the area of automatic speech recognition. We experiment with Mozilla’s DeepSpeech architecture on German and Swiss German speech datasets and compare the results of either training from scratch vs. transferring a pre-trained model. We compare different layer freezing schemes and find that even freezing only one layer already significantly improves results.

@inproceedings{eberhard-2021-effects,
  title = {Effects of Layer Freezing on Transferring a Speech Recognition System to Under-resourced Languages},
  author = {Eberhard, Onno and Zesch, Torsten},
  booktitle = {Proceedings of the 17th Conference on Natural Language Processing (KONVENS 2021)},
  year = {2021},
  url = {https://aclanthology.org/2021.konvens-1.19},
  pages = {208--212}
}