A Sociolinguistic Analysis of Automatic Speech Recognition Bias in Newcastle English

2026-03-25 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Advanced, quick

Summary

A study investigated Automatic Speech Recognition (ASR) bias in Newcastle English, a regional dialect from North-East England, using spontaneous speech from the Diachronic Electronic Corpus of Tyneside English (DECTE). Researchers evaluated a commercial ASR system, analyzing over 3,000 transcription errors classified by linguistic domain and social variables like gender, age, and socioeconomic status. An acoustic case study also examined how phonetic variation contributes to misrecognition. Results showed that phonological variation caused most errors, specifically dialect-specific vowel quality, glottalisation, local vocabulary, and non-standard grammar. Error rates were higher for men and speakers at the extremes of the age spectrum, indicating socially patterned ASR errors.

Key takeaway

For NLP Engineers developing or evaluating ASR systems, understanding that errors are socially patterned and linked to dialectal variation is crucial. You should prioritize incorporating sociolinguistic expertise into system development and evaluation, ensuring training data explicitly addresses regional accents and community-based speech to build more equitable and accurate ASR technologies.

Key insights

ASR errors are not random but socially patterned, driven by dialectal variation and demographic factors.

Principles

• Dialectal variation challenges ASR performance.
• Sociolinguistic factors influence ASR error rates.

Method

Evaluated a commercial ASR system on spontaneous dialectal speech, classifying over 3,000 transcription errors by linguistic domain and social variables, complemented by an acoustic case study.

In practice

• Focus ASR training on dialect-specific features.
• Include diverse community-based speech data.
• Analyze errors by social group.

Topics

• Automatic Speech Recognition
• ASR Bias
• Sociolinguistic Analysis
• Dialectal Variation
• Newcastle English

Best for: NLP Engineer, Research Scientist, AI Researcher, AI Scientist, AI Ethicist

Related on AIssential

• How Dialect Variation Challenges Natural Language Processing Systems — NLP systems struggle with dialects due to biased training data, leading to technical and social inequities.
• GRUFF: LLM Pronoun Fidelity, Reasoning, and Biases in German — The GRUFF dataset reveals German LLMs struggle with neopronouns and distractor robustness, despite strong grammatical gender agreement.
• Can I guess where you are from? Modeling dialectal morphosyntactic similarities in Brazilian Portuguese — Clustering morphosyntactic features can reveal regional dialectal patterns in Brazilian Portuguese.
• Disparities In Negation Understanding Across Languages In Vision-Language Models — VLMs exhibit affirmation bias, with negation understanding varying significantly across typologically diverse languages.
• Fully Automated Identification of Lexical Alignment and Preference-Stage Shifts in Large Language Models — Automated metrics can identify and quantify lexical overuse and preference learning shifts in LLMs without manual curation.
• Analysing Differences in Persuasive Language in LLM-Generated Text: Uncovering Stereotypical Gender Patterns — LLMs exhibit pervasive gender-stereotypical persuasive language biases across models and languages, reflecting societal norms.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.