ISCA Archive Interspeech 2022
ISCA Archive Interspeech 2022

Production federated keyword spotting via distillation, filtering, and joint federated-centralized training

Andrew Hard, Kurt Partridge, Neng Chen, Sean Augenstein, Aishanee Shah, Hyun Jin Park, Alex Park, Sara Ng, Jessica Nguyen, Ignacio Lopez-Moreno, Rajiv Mathews, Francoise Beaufays

We trained a keyword spotting model using federated learning on real user devices and observed significant improvements when the model was deployed for inference on phones. To compensate for data domains that are missing from on-device training caches, we employed joint federated-centralized training. And to learn in the absence of curated labels on-device, we formulated a confidence filtering strategy based on user-feedback signals for federated distillation. These techniques created models that significantly improved quality metrics in offline evaluations and user-experience metrics in live A/B experiments.


doi: 10.21437/Interspeech.2022-11050

Cite as: Hard, A., Partridge, K., Chen, N., Augenstein, S., Shah, A., Park, H.J., Park, A., Ng, S., Nguyen, J., Lopez-Moreno, I., Mathews, R., Beaufays, F. (2022) Production federated keyword spotting via distillation, filtering, and joint federated-centralized training. Proc. Interspeech 2022, 76-80, doi: 10.21437/Interspeech.2022-11050

@inproceedings{hard22_interspeech,
  author={Andrew Hard and Kurt Partridge and Neng Chen and Sean Augenstein and Aishanee Shah and Hyun Jin Park and Alex Park and Sara Ng and Jessica Nguyen and Ignacio Lopez-Moreno and Rajiv Mathews and Francoise Beaufays},
  title={{Production federated keyword spotting via distillation, filtering, and joint federated-centralized training}},
  year=2022,
  booktitle={Proc. Interspeech 2022},
  pages={76--80},
  doi={10.21437/Interspeech.2022-11050},
  issn={2958-1796}
}