InterCorrect: Intersection-Aware Correction of Demographic Model Merging for Fair ASR

2026-10-06

AI Summary

This study addresses recognition biases in automatic speech recognition (ASR) systems across intersectional demographic groups by proposing an intersectionality-aware correction mechanism. Building upon the SLAM-ASR architecture, the method fine-tunes the connector module and employs the TIES algorithm to analyze task vector conflicts and merge subgroup-specific models. It subsequently identifies critical intersectional axes and applies targeted correction vectors to the global model to optimize fairness. Evaluated on the Fair-Speech dataset, the proposed approach reduces the overall word error rate (WER) from 7.38% to 5.13%, significantly enhancing both speech recognition accuracy and system fairness across diverse demographic populations.

Abstract

Automatic Speech Recognition (ASR) systems often show uneven performance across demographic groups, and errors can be especially difficult to address for speakers belonging to multiple demographic groups. This work studies demographic-aware model merging for fair Speech-LLM-based ASR. Starting from a SLAM-ASR-based model, we fine-tune only the connector on demographic-specific subsets and merge the resulting subgroup-adapted connectors into a global model. We then identify critical cross-axis demographic pairs using subgroup WER and task-vector conflict, and apply intersection-specific correction vectors to the global merged model. Experiments on Fair-Speech show that global demographic merging improves overall WER over the base model, while intersection correction provides additional gains for several merging strategies. In particular, TIES with WER-based correction achieves the best overall WER, reducing it from 7.38\% to 5.13\%. Subgroup and disparity analyses further show that the proposed approach improves performance across demographic axes, while highlighting that lower average WER does not always imply reduced subgroup disparity.

Problem

• Automatic Speech Recognition
• Fairness
• Demographic Groups
• Intersectionality
• Model Merging

Innovation

• Model Merging
• Fair ASR
• Intersectionality
• Task Vector Conflict
• Speech-LLM