End-to-End Speech Recognition from Federated Acoustic Models
This work addresses the gap in realistic federated learning for speech recognition, which is important for developers and researchers aiming to deploy privacy-preserving ASR systems, though it is incremental as it builds on existing FL methods.
The paper tackled the problem of training end-to-end automatic speech recognition models in realistic federated learning settings with heterogeneous data, by constructing a challenging experimental setup using CommonVoice datasets and comparing three aggregation strategies, achieving results that show WER-based aggregation performs best in cross-silo and cross-device scenarios with up to 4K clients.
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.