DeepSpeech: end-to-end recurrent recognition and scalable training

What changed

Integrated end-to-end recurrent speech recognition, multi-GPU training and synthesized noisy training data to reduce hand-engineered recognition stages.

Scientific scope

Integrated end-to-end recurrent speech recognition, multi-GPU training and synthesized noisy training data to reduce hand-engineered recognition stages. Inherits recurrent networks and CTC. Source still combines a language model at decoding; end-to-end does not mean no preprocessing, prior data or external decoder. Noise robustness is empirical, not universal invariance.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits recurrent networks and CTC. Source still combines a language model at decoding; end-to-end does not mean no preprocessing, prior data or external decoder. Noise robustness is empirical, not universal invariance.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• DeepSpeech: Scaling up end-to-end speech recognition

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction