M-CIF: Multi-Scale Alignment For CIF-Based Non-Autoregressive ASR
This work addresses a specific bottleneck in non-autoregressive ASR for multiple languages, offering incremental improvements in alignment stability and error reduction.
The paper tackles the instability of the Continuous Integrate-and-Fire (CIF) mechanism in non-autoregressive speech recognition for languages like English and French by proposing Multi-scale CIF (M-CIF), which integrates multi-level alignment with character and phoneme supervision, resulting in reduced word error rates, such as 4.21% in German and 3.05% in French on CommonVoice compared to a baseline.
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.