Transsion Multilingual Speech Recognition System for MLC-SLM 2025 Challenge

TL;DR

This paper introduces a multilingual speech recognition system combining pretrained models and fine-tuning, achieving competitive error rates across 11 languages for the MLC-SLM 2025 Challenge.

Contribution

It presents a novel architecture integrating large pretrained speech and language models with trainable modules for multilingual ASR.

Findings

Achieved 9.83% WER/CER on evaluation set

Ranked third among global participants

Effective integration of pretrained models with task-specific adaptation

Abstract

This paper presents the architecture and performance of a novel Multilingual Automatic Speech Recognition (ASR) system developed by the Transsion Speech Team for Track 1 of the MLC-SLM 2025 Challenge. The proposed system comprises three key components: 1) a frozen Whisper-large-v3 based speech encoder, leveraging large-scale pretraining to ensure robust acoustic feature extraction; 2) a trainable adaptor module using Linear-ReLU-Linear transformation mechanisms to effectively align speech and text representations; and 3) a frozen Qwen2.5-7B-Instruct large language model (LLM) integrated with trainable LoRA for optimized contextual linguistic decoding. By systematically combining pretrained models with task specific fine-tuning, the system achieved a word/character error rate (WER/CER) of 9.83% across 11 languages in the evaluation set and ranked third place among global participants.