Dialect Matters: Cross-Lingual ASR Transfer for Low-Resource Indic Language Varieties

2026-01-07

arXiv.org

AI Summary

This study addresses the challenges of cross-lingual automatic speech recognition (ASR) for low-resource Indo-Aryan dialects, which are characterized by spontaneity, ambient noise, and code-mixing. Through empirical evaluation, the work systematically assesses the transfer performance of modern ASR models across multiple Indian dialects, with a particular focus on the influence of phylogenetic distance and dialectal data quantity. Using Garhwali as a case study, the research demonstrates that phylogenetic proximity alone does not dictate transfer efficacy; remarkably, even a small amount of dialect-specific fine-tuning data can yield substantial performance gains—sometimes matching models trained extensively on high-resource standard languages. Furthermore, the study uncovers systematic transcription biases in current ASR systems against non-standardized speech varieties, highlighting a critical gap in robustness and inclusivity.

Technology Category

• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Cognitive Modeling & Cognitive Systems: Affective Computing

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

We conduct an empirical study of cross-lingual transfer using spontaneous, noisy, and code-mixed speech across a wide range of Indic dialects and language varieties. Our results indicate that although ASR performance is generally improved with reduced phylogenetic distance between languages, this factor alone does not fully explain performance in dialectal settings. Often, fine-tuning on smaller amounts of dialectal data yields performance comparable to fine-tuning on larger amounts of phylogenetically-related, high-resource standardized languages. We also present a case study on Garhwali, a low-resource Pahari language variety, and evaluate multiple contemporary ASR models. Finally, we analyze transcription errors to examine bias toward pre-training languages, providing additional insight into challenges faced by ASR systems on dialectal and non-standardized speech.

Problem

• dialect
• low-resource
• cross-lingual ASR
• Indic languages
• non-standardized speech

Innovation

• cross-lingual transfer
• low-resource ASR
• dialectal speech
• code-mixed speech
• Garhwali