Mitigating Structural Noise in Low-Resource S2TT: An Optimized Cascaded Nepali-English Pipeline with Punctuation Restoration
This work addresses translation quality degradation from punctuation loss in low-resource language speech translation systems, establishing an optimized baseline for Nepali-English S2TT.
This paper tackled structural noise in low-resource Nepali speech-to-English text translation by developing an optimized cascaded pipeline with punctuation restoration, achieving a 4.90 BLEU point gain over the baseline (36.38 vs. 31.48 BLEU) and validating improvements through human assessment scores of 3.673 for Adequacy and 3.804 for Fluency.
This paper presents and evaluates an optimized cascaded Nepali speech-to-English text translation (S2TT) system, focusing on mitigating structural noise introduced by Automatic Speech Recognition (ASR). We first establish highly proficient ASR and NMT components: a Wav2Vec2-XLS-R-300m model achieved a state-of-the-art 2.72% CER on OpenSLR-54, and a multi-stage fine-tuned MarianMT model reached a 28.32 BLEU score on the FLORES-200 benchmark. We empirically investigate the influence of punctuation loss, demonstrating that unpunctuated ASR output significantly degrades translation quality, causing a massive 20.7% relative BLEU drop on the FLORES benchmark. To overcome this, we propose and evaluate an intermediate Punctuation Restoration Module (PRM). The final S2TT pipeline was tested across three configurations on a custom dataset. The optimal configuration, which applied the PRM directly to ASR output, achieved a 4.90 BLEU point gain over the direct ASR-to-NMT baseline (BLEU 36.38 vs. 31.48). This improvement was validated by human assessment, which confirmed the optimized pipeline's superior Adequacy (3.673) and Fluency (3.804). This work validates that targeted punctuation restoration is the most effective intervention for mitigating structural noise in the Nepali S2TT pipeline. It establishes an optimized baseline and demonstrates a critical architectural insight for developing cascaded speech translation systems for similar low-resource languages.