Dual-Pipeline with Low-Rank Adaptation for New Language Integration in Multilingual ASR

TL;DR

This paper introduces a dual-pipeline approach with low-rank adaptation to efficiently incorporate new languages into pre-trained multilingual ASR systems, minimizing performance loss on existing languages.

Contribution

It proposes a novel dual-pipeline framework with LoRA for seamless integration of new languages into multilingual ASR models, maintaining performance on existing languages.

Findings

Effective extension of Whisper to 19 new languages

Minimal performance degradation on existing languages

Language-agnostic operation enabled

Abstract

This paper addresses challenges in integrating new languages into a pre-trained multilingual automatic speech recognition (mASR) system, particularly in scenarios where training data for existing languages is limited or unavailable. The proposed method employs a dual-pipeline with low-rank adaptation (LoRA). It maintains two data flow pipelines-one for existing languages and another for new languages. The primary pipeline follows the standard flow through the pre-trained parameters of mASR, while the secondary pipeline additionally utilizes language-specific parameters represented by LoRA and a separate output decoder module. Importantly, the proposed approach minimizes the performance degradation of existing languages and enables a language-agnostic operation mode, facilitated by a decoder selection strategy. We validate the effectiveness of the proposed method by extending the pre-trained Whisper model to 19 new languages from the FLEURS dataset