Dynamic Context-Aware Streaming Pretrained Language Model For Inverse Text Normalization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ho, Luong, Le, Khanh, Pham, Vinh, Nguyen, Bao, Tran, Tan, Chau, Duc
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912403800719360
author Ho, Luong
Le, Khanh
Pham, Vinh
Nguyen, Bao
Tran, Tan
Chau, Duc
author_facet Ho, Luong
Le, Khanh
Pham, Vinh
Nguyen, Bao
Tran, Tan
Chau, Duc
contents Inverse Text Normalization (ITN) is crucial for converting spoken Automatic Speech Recognition (ASR) outputs into well-formatted written text, enhancing both readability and usability. Despite its importance, the integration of streaming ITN within streaming ASR remains largely unexplored due to challenges in accuracy, efficiency, and adaptability, particularly in low-resource and limited-context scenarios. In this paper, we introduce a streaming pretrained language model for ITN, leveraging pretrained linguistic representations for improved robustness. To address streaming constraints, we propose Dynamic Context-Aware during training and inference, enabling adaptive chunk size adjustments and the integration of right-context information. Experimental results demonstrate that our method achieves accuracy comparable to non-streaming ITN and surpasses existing streaming ITN models on a Vietnamese dataset, all while maintaining low latency, ensuring seamless integration into ASR systems.
format Preprint
id arxiv_https___arxiv_org_abs_2505_24229
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Dynamic Context-Aware Streaming Pretrained Language Model For Inverse Text Normalization
Ho, Luong
Le, Khanh
Pham, Vinh
Nguyen, Bao
Tran, Tan
Chau, Duc
Computation and Language
Sound
Audio and Speech Processing
Inverse Text Normalization (ITN) is crucial for converting spoken Automatic Speech Recognition (ASR) outputs into well-formatted written text, enhancing both readability and usability. Despite its importance, the integration of streaming ITN within streaming ASR remains largely unexplored due to challenges in accuracy, efficiency, and adaptability, particularly in low-resource and limited-context scenarios. In this paper, we introduce a streaming pretrained language model for ITN, leveraging pretrained linguistic representations for improved robustness. To address streaming constraints, we propose Dynamic Context-Aware during training and inference, enabling adaptive chunk size adjustments and the integration of right-context information. Experimental results demonstrate that our method achieves accuracy comparable to non-streaming ITN and surpasses existing streaming ITN models on a Vietnamese dataset, all while maintaining low latency, ensuring seamless integration into ASR systems.
title Dynamic Context-Aware Streaming Pretrained Language Model For Inverse Text Normalization
topic Computation and Language
Sound
Audio and Speech Processing
url https://arxiv.org/abs/2505.24229