Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Ling, Tongtao, Lai, Yutao, Chen, Lei, Huang, Shilei, Liu, Yi
Format: Preprint
Veröffentlicht: 2023
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2308.12568
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913407521783808
author Ling, Tongtao
Lai, Yutao
Chen, Lei
Huang, Shilei
Liu, Yi
author_facet Ling, Tongtao
Lai, Yutao
Chen, Lei
Huang, Shilei
Liu, Yi
contents In clinical dictation, utterances after automatic speech recognition (ASR) without explicit punctuation marks may lead to the misunderstanding of dictated reports. To give a precise and understandable clinical report with ASR, automatic punctuation restoration is required. Considering a practical scenario, we propose a fast and light pre-trained model for Chinese medical punctuation restoration based on 'pretraining and fine-tuning' paradigm. In this work, we distill pre-trained models by incorporating supervised contrastive learning and a novel auxiliary pre-training task (Punctuation Mark Prediction) to make it well-suited for punctuation restoration. Our experiments on various distilled models reveal that our model can achieve 95% performance while 10% model size relative to state-of-the-art Chinese RoBERTa.
format Preprint
id arxiv_https___arxiv_org_abs_2308_12568
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle A Small and Fast BERT for Chinese Medical Punctuation Restoration
Ling, Tongtao
Lai, Yutao
Chen, Lei
Huang, Shilei
Liu, Yi
Computation and Language
In clinical dictation, utterances after automatic speech recognition (ASR) without explicit punctuation marks may lead to the misunderstanding of dictated reports. To give a precise and understandable clinical report with ASR, automatic punctuation restoration is required. Considering a practical scenario, we propose a fast and light pre-trained model for Chinese medical punctuation restoration based on 'pretraining and fine-tuning' paradigm. In this work, we distill pre-trained models by incorporating supervised contrastive learning and a novel auxiliary pre-training task (Punctuation Mark Prediction) to make it well-suited for punctuation restoration. Our experiments on various distilled models reveal that our model can achieve 95% performance while 10% model size relative to state-of-the-art Chinese RoBERTa.
title A Small and Fast BERT for Chinese Medical Punctuation Restoration
topic Computation and Language
url https://arxiv.org/abs/2308.12568