M-CIF: Multi-Scale Alignment For CIF-Based Non-Autoregressive ASR

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mao, Ruixiang, Ma, Xiangnan, Yang, Qing, Zhu, Ziming, Qiao, Yucheng, Ge, Yuan, Xiao, Tong, Gao, Shengxiang, Yu, Zhengtao, Zhu, Jingbo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918171220377600
author Mao, Ruixiang
Ma, Xiangnan
Yang, Qing
Zhu, Ziming
Qiao, Yucheng
Ge, Yuan
Xiao, Tong
Gao, Shengxiang
Yu, Zhengtao
Zhu, Jingbo
author_facet Mao, Ruixiang
Ma, Xiangnan
Yang, Qing
Zhu, Ziming
Qiao, Yucheng
Ge, Yuan
Xiao, Tong
Gao, Shengxiang
Yu, Zhengtao
Zhu, Jingbo
contents The Continuous Integrate-and-Fire (CIF) mechanism provides effective alignment for non-autoregressive (NAR) speech recognition. This mechanism creates a smooth and monotonic mapping from acoustic features to target tokens, achieving performance on Mandarin competitive with other NAR approaches. However, without finer-grained guidance, its stability degrades in some languages such as English and French. In this paper, we propose Multi-scale CIF (M-CIF), which performs multi-level alignment by integrating character and phoneme level supervision progressively distilled into subword representations, thereby enhancing robust acoustic-text alignment. Experiments show that M-CIF reduces WER compared to the Paraformer baseline, especially on CommonVoice by 4.21% in German and 3.05% in French. To further investigate these gains, we define phonetic confusion errors (PE) and space-related segmentation errors (SE) as evaluation metrics. Analysis of these metrics across different M-CIF settings reveals that the phoneme and character layers are essential for enhancing progressive CIF alignment.
format Preprint
id arxiv_https___arxiv_org_abs_2510_22172
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle M-CIF: Multi-Scale Alignment For CIF-Based Non-Autoregressive ASR
Mao, Ruixiang
Ma, Xiangnan
Yang, Qing
Zhu, Ziming
Qiao, Yucheng
Ge, Yuan
Xiao, Tong
Gao, Shengxiang
Yu, Zhengtao
Zhu, Jingbo
Sound
Computation and Language
The Continuous Integrate-and-Fire (CIF) mechanism provides effective alignment for non-autoregressive (NAR) speech recognition. This mechanism creates a smooth and monotonic mapping from acoustic features to target tokens, achieving performance on Mandarin competitive with other NAR approaches. However, without finer-grained guidance, its stability degrades in some languages such as English and French. In this paper, we propose Multi-scale CIF (M-CIF), which performs multi-level alignment by integrating character and phoneme level supervision progressively distilled into subword representations, thereby enhancing robust acoustic-text alignment. Experiments show that M-CIF reduces WER compared to the Paraformer baseline, especially on CommonVoice by 4.21% in German and 3.05% in French. To further investigate these gains, we define phonetic confusion errors (PE) and space-related segmentation errors (SE) as evaluation metrics. Analysis of these metrics across different M-CIF settings reveals that the phoneme and character layers are essential for enhancing progressive CIF alignment.
title M-CIF: Multi-Scale Alignment For CIF-Based Non-Autoregressive ASR
topic Sound
Computation and Language
url https://arxiv.org/abs/2510.22172