End-to-End Speech Recognition from Federated Acoustic Models
Yan Gao, Titouan Parcollet, Salah Zaiem, Javier Fernandez-Marques, Pedro P. B. de Gusmao, Daniel J. Beutel, Nicholas D. Lane
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/yan-gao-GY/Flower-SpeechBrainOfficialIn paperpytorch★ 11
Abstract
Training Automatic Speech Recognition (ASR) models under federated learning (FL) settings has attracted a lot of attention recently. However, the FL scenarios often presented in the literature are artificial and fail to capture the complexity of real FL systems. In this paper, we construct a challenging and realistic ASR federated experimental setup consisting of clients with heterogeneous data distributions using the French and Italian sets of the CommonVoice dataset, a large heterogeneous dataset containing thousands of different speakers, acoustic environments and noises. We present the first empirical study on attention-based sequence-to-sequence End-to-End (E2E) ASR model with three aggregation weighting strategies -- standard FedAvg, loss-based aggregation and a novel word error rate (WER)-based aggregation, compared in two realistic FL scenarios: cross-silo with 10 clients and cross-device with 2K and 4K clients. Our analysis on E2E ASR from heterogeneous and realistic federated acoustic models provides the foundations for future research and development of realistic FL-based ASR applications.