Robust Speech Recognition via Large-Scale Weak Supervision

Abstract

We study the capabilities of speech processing systems trained simply to predict large amounts of transcripts of audio on the internet. When scaled to 680,000 hours of multilingual and multitask supervision, the resulting models generalize well to standard benchmarks and are often competitive with prior fully supervised results but in a zero-shot transfer setting without the need for any fine-tuning. When compared to humans, the models approach their accuracy and robustness. We are releasing models and inference code to serve as a foundation for further work on robust speech processing.

Cited by in corpus (13)

• 14 Examples of How LLMs Can Transform Materials Science and Chemistry: A Reflection on a Large Language Model Hackathon
• Augmented Datasheets for Speech Datasets and Ethical Decision-Making
• Design and Evaluation of a Socially Assistive Robot Schoolwork Companion for College Students with ADHD
• Automated speech audiometry: Can it work using open-source pre-trained Kaldi-NL automatic speech recognition?
• LanSER: Language-Model Supported Speech Emotion Recognition
• Considerations for Ethical Speech Recognition Datasets
• A Multimodal Approach to Device-Directed Speech Detection with Large Language Models
• Advancing Audio Emotion and Intent Recognition with Large Pre-Trained Models and Bayesian Inference
• Single-Channel Robot Ego-Speech Filtering during Human-Robot Interaction
• Parameter Selection for Analyzing Conversations with Autism Spectrum Disorder
• Computational analyses of linguistic features with schizophrenic and autistic traits along with formal thought disorders
• Remote Inference of Cognitive Scores in ALS Patients Using a Picture Description
• Modular Speech-to-Text Translation for Zero-Shot Cross-Modal Transfer