Training ASR Models By Generation Of Contextual Information

Abstract

Supervised ASR models have reached unprecedented levels of accuracy, thanks in part to ever-increasing amounts of labelled training data. However, in many applications and locales, only moderate amounts of data are available, which has led to a surge in semi- and weakly-supervised learning research. In this paper, we conduct a large-scale study evaluating the effectiveness of weakly-supervised learning for speech recognition by using loosely related contextual information as a surrogate for ground-truth labels. For weakly supervised training, we use 50k hours of public English social media videos along with their respective titles and post text to train an encoder-decoder transformer model. Our best encoder-decoder models achieve an average of 20.8% WER reduction over a 1000 hours supervised baseline, and an average of 13.4% WER reduction when using only the weakly supervised encoder for CTC fine-tuning. Our results show that our setup for weak supervision improved both the encoder aco

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keysingh2019training

Related papers

• Large Scale Weakly And Semi-supervised Learning For Low-resource Video ASR (2020)0.00
• Improving RNN-T ASR Accuracy Using Context Audio (2020)5.84
• Deep Contextualized Acoustic Representations For Semi-supervised Speech Recognition (2019)14.62
• From Weak Labels To Strong Results: Utilizing 5,000 Hours Of Noisy Classroom Transcripts With Minimal Accurate Data (2025)0.00
• Leveraging Acoustic Contextual Representation By Audio-textual Cross-modal Learning For Conversational ASR (2022)0.00
• Learning Asr-robust Contextualized Embeddings For Spoken Language Understanding (2019)12.02
• A Comparison Of Semi-supervised Learning Techniques For Streaming ASR At Scale (2023)2.26
• Improving RNN Transducer Based ASR With Auxiliary Tasks (2020)9.59