GLoRIA: Gated Low-Rank Interpretable Adaptation for Dialectal ASR

TL;DR

GLoRIA is an efficient, interpretable adaptation method for dialectal ASR that uses metadata to modulate low-rank updates, outperforming existing methods and generalizing well to unseen dialects.

Contribution

We introduce GLoRIA, a novel metadata-gated low-rank adaptation framework for dialectal ASR that achieves state-of-the-art results with minimal parameter updates.

Findings

Outperforms geo-conditioned full fine-tuning and LoRA on GCND corpus.

Achieves state-of-the-art word error rates with under 10% parameter updates.

Generalizes effectively to unseen dialects and enables interpretable geospatial adaptation patterns.

Abstract

Automatic Speech Recognition (ASR) in dialect-heavy settings remains challenging due to strong regional variation and limited labeled data. We propose GLoRIA, a parameter-efficient adaptation framework that leverages metadata (e.g., coordinates) to modulate low-rank updates in a pre-trained encoder. GLoRIA injects low-rank matrices into each feed-forward layer, with a gating MLP determining the non-negative contribution of each LoRA rank-1 component based on location metadata. On the GCND corpus, GLoRIA outperforms geo-conditioned full fine-tuning, LoRA, and both dialect-specific and unified full fine-tuning, achieving state-of-the-art word error rates while updating under 10% of parameters. GLoRIA also generalizes well to unseen dialects, including in extrapolation scenarios, and enables interpretable adaptation patterns that can be visualized geospatially. These results show metadata-gated low-rank adaptation is an effective, interpretable, and efficient solution for dialectal ASR.