Local authorities in the UK are rolling out AI-powered phone lines that will learn regional slang to improve communication with residents regarding bin collection. These automated systems, increasingly used by councils, often struggle to interpret local accents and colloquial terms.
A new study led by researchers at the University of Sheffield, in collaboration with ICS.AI, aims to enhance AI’s ability to understand regional dialects. This is the first study of its kind in the UK, focusing on making the technology more inclusive.
Dr. Chris Montgomery, a senior lecturer in dialectology at Sheffield, has reviewed over 50 years of research on accent and dialect variation. He notes that people calling about bin collections or council tax may face difficulties if they speak quickly, use strong accents, or employ terms of endearment like “chuck,” “pet,” “chug,” or “duck.”
For instance, the word “chuck” in northern England can mean both a term of endearment and the verb “to throw,” potentially confusing AI. “It might struggle to interpret particular words, grammatical constructions, and strongly regional accents,” Montgomery explained.
The research follows issues faced by ICS.AI’s voice agent Darcie, launched with Derby City Council, which had trouble with the Midlands accent. A BBC Radio Derby presenter found that her use of “duck” was not properly understood. Montgomery adds that strong Scottish accents are harder for AI to process compared to standard southern ones.
The next phase aims to ensure AI reflects the UK’s linguistic diversity, preventing anyone from being disadvantaged by their speech. A recent survey found that 52% of UK residents worry AI may fail to comprehend accents, with figures rising to 71% in Scotland.