HITSZ's End-To-End Speech Translation Systems Combining Sequence-to-Sequence Auto Speech Recognition Model and Indic Large Language Model for IWSLT 2025 in Indic Track

TL;DR

This paper introduces an end-to-end speech translation system combining Whisper ASR and Krutrim LLM for English-Indic translation, achieving competitive BLEU scores and exploring Chain-of-Thought prompting to improve translation quality.

Contribution

The paper presents a novel integration of pre-trained ASR and Indic-specific LLMs for low-resource speech translation, and investigates Chain-of-Thought prompting effects.

Findings

Achieved BLEU scores of 28.88 (English-Indic) and 27.86 (Indic-English)

Chain-of-Thought improved Tamil-to-English translation BLEU by 13.84

Challenges in maintaining consistent CoT output format

Abstract

This paper presents HITSZ's submission for the IWSLT 2025 Indic track, focusing on speech-to-text translation (ST) for English-to-Indic and Indic-to-English language pairs. To enhance translation quality in this low-resource scenario, we propose an end-to-end system integrating the pre-trained Whisper automated speech recognition (ASR) model with Krutrim, an Indic-specialized large language model (LLM). Experimental results demonstrate that our end-to-end system achieved average BLEU scores of $28.88$ for English-to-Indic directions and $27.86$ for Indic-to-English directions. Furthermore, we investigated the Chain-of-Thought (CoT) method. While this method showed potential for significant translation quality improvements on successfully parsed outputs (e.g. a $13.84$ BLEU increase for Tamil-to-English), we observed challenges in ensuring the model consistently adheres to the required CoT output format.