Sociophonetic Analysis of ASR Performance: Investigating Socially Patterned Bias in Newcastle English

Forum topic · 小凯 · 2026-03-27

Summary

This paper by Kevin Tang (arXiv:2603.24549) examines bias in Automatic Speech Recognition (ASR) systems through a sociophonetic and sociolinguistic analysis of Newcastle English. While ASR systems are widely deployed in everyday communication, education, healthcare, and industry, their performance remains uneven across speakers, particularly when dialectal variation diverges from the mainstream accents represented in training data. The study uses Newcastle English—a well-documented regional dialect of British English—as a case study to investigate how socially patterned phonetic variation affects recognition accuracy. By connecting sociolinguistic theory with ASR evaluation, the work highlights how dialect and accent bias can lead to systematic performance disparities, with implications for fairness, accessibility, and the development of more inclusive speech technology.

Paper Overview

Field: NLP
Author: Kevin Tang
Published: 2026-03-25
arXiv: 2603.24549

Abstract

Automatic Speech Recognition (ASR) systems are widely used in everyday communication, education, healthcare, and industry, yet their performance remains uneven across speakers, particularly when dialectal variation diverges from the mainstream accents represented in training data. This study investigates ASR bias through a sociolinguistic analysis of Newcastle English.

Context

ASR systems are increasingly embedded in daily life and critical services, but speakers of regional or non-mainstream dialects often experience lower recognition accuracy. This work applies a sociophonetic lens to quantify and explain such disparities, using Newcastle English as the case study, and underscores the need for accent-aware evaluation and more inclusive training data in speech technology.

*Auto-collected on 2026-03-27.*