English

Discriminative Self-training for Punctuation Prediction

Computation and Language 2021-09-02 v2

Abstract

Punctuation prediction for automatic speech recognition (ASR) output transcripts plays a crucial role for improving the readability of the ASR transcripts and for improving the performance of downstream natural language processing applications. However, achieving good performance on punctuation prediction often requires large amounts of labeled speech transcripts, which is expensive and laborious. In this paper, we propose a Discriminative Self-Training approach with weighted loss and discriminative label smoothing to exploit unlabeled speech transcripts. Experimental results on the English IWSLT2011 benchmark test set and an internal Chinese spoken language dataset demonstrate that the proposed approach achieves significant improvement on punctuation prediction accuracy over strong baselines including BERT, RoBERTa, and ELECTRA models. The proposed Discriminative Self-Training approach outperforms the vanilla self-training approach. We establish a new state-of-the-art (SOTA) on the IWSLT2011 test set, outperforming the current SOTA model by 1.3% absolute gain on F1_1.

Keywords

Cite

@article{arxiv.2104.10339,
  title  = {Discriminative Self-training for Punctuation Prediction},
  author = {Qian Chen and Wen Wang and Mengzhe Chen and Qinglin Zhang},
  journal= {arXiv preprint arXiv:2104.10339},
  year   = {2021}
}

Comments

Accepted by INTERSPEECH 2021

R2 v1 2026-06-24T01:23:22.244Z