General Northern English. Exploring Regional Variation in the North of England With Machine Learning.

Area of Science:

• Computational linguistics
• Sociolinguistics
• Phonetics

Background:

• Dialect leveling is occurring in Northern England, leading to the emergence of General Northern English (GNE).
• Understanding accent variation and its reduction is crucial for sociolinguistic research.

Purpose of the Study:

• To present a novel computational approach for analyzing accent variation.
• To investigate dialect leveling in Northern England using random forest classification.
• To identify key phonetic features contributing to regional accent differences.

Main Methods:

• Utilized random forest classification on audio data from 105 speakers across five northern UK cities.
• Employed first two formant measurements of vowel systems for classification.
• Applied undersampling, bagging, and leave-one-out cross-validation to manage data challenges.

Main Results:

• Classification accuracy measured relative similarity between city accents.
• Identified specific vowels and formants as influential features in accent prediction.
• Observed significant leveling between Manchester, Leeds, and Sheffield, with persistent subtle differences.

Conclusions:

• The study confirms dialect leveling in Northern England, with General Northern English emerging.
• Identified novel sociolinguistic regional indicators based on non-salient phonetic features.
• Combined computational and traditional methods to describe General Northern English vowels.