English

Token-Weighted RNN-T for Learning from Flawed Data

Computation and Language 2024-06-27 v1 Machine Learning Sound Audio and Speech Processing

Abstract

ASR models are commonly trained with the cross-entropy criterion to increase the probability of a target token sequence. While optimizing the probability of all tokens in the target sequence is sensible, one may want to de-emphasize tokens that reflect transcription errors. In this work, we propose a novel token-weighted RNN-T criterion that augments the RNN-T objective with token-specific weights. The new objective is used for mitigating accuracy loss from transcriptions errors in the training data, which naturally appear in two settings: pseudo-labeling and human annotation errors. Experiments results show that using our method for semi-supervised learning with pseudo-labels leads to a consistent accuracy improvement, up to 38% relative. We also analyze the accuracy degradation resulting from different levels of WER in the reference transcription, and show that token-weighted RNN-T is suitable for overcoming this degradation, recovering 64%-99% of the accuracy loss.

Keywords

Cite

@article{arxiv.2406.18108,
  title  = {Token-Weighted RNN-T for Learning from Flawed Data},
  author = {Gil Keren and Wei Zhou and Ozlem Kalinli},
  journal= {arXiv preprint arXiv:2406.18108},
  year   = {2024}
}
R2 v1 2026-06-28T17:19:32.686Z