Active Learning for Speech Recognition: the Power of Gradients

Abstract

In training speech recognition systems, labeling audio clips can be expensive, and not all data is equally valuable. Active learning aims to label only the most informative samples to reduce cost. For speech recognition, confidence scores and other likelihood-based active learning methods have been shown to be effective. Gradient-based active learning methods, however, are still not well-understood. This work investigates the Expected Gradient Length (EGL) approach in active learning for end-to-end speech recognition. We justify EGL from a variance reduction perspective, and observe that EGL's measure of informativeness picks novel samples uncorrelated with confidence scores. Experimentally, we show that EGL can reduce word errors by 11\%, or alternatively, reduce the number of samples to label by 50\%, when compared to random sampling.

arXiv commentpublished as a workshop paper at NIPS 2016

References in corpus (1)

• Deep Speech 2: End-to-End Speech Recognition in English and Mandarin

Cited by in corpus (17)

• Fine-Tuning Language Models from Human Preferences
• Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds
• Discriminative Active Learning
• Deep Representation Learning in Speech Processing: Challenges, Recent Advances, and Future Trends
• Active Learning to Classify Macromolecular Structures in situ for Less Supervision in Cryo-Electron Tomography
• Deep Active Learning for Text Classification with Diverse Interpretations
• Deep Active Learning by Leveraging Training Dynamics
• Multi-class Text Classification using BERT-based Active Learning
• Boosting Active Learning for Speech Recognition with Noisy Pseudo-labeled Samples
• Active Deep Learning on Entity Resolution by Risk Sampling
• Active Learning of Non-semantic Speech Tasks with Pretrained Models
• Ask-n-Learn: Active Learning via Reliable Gradient Representations for Image Classification
• Bayesian Uncertainty and Expected Gradient Length -- Regression: Two Sides Of The Same Coin?
• ALLWAS: Active Learning on Language models in WASserstein space
• Batch Decorrelation for Active Metric Learning
• Deep Active Learning by Model Interpretability
• Loss Prediction: End-to-End Active Learning Approach For Speech Recognition