Speech Classification of British And American English Using Machine Learning

Abstract

This study focuses on the classification of British and American English accents using machine learning, addressing the growing interest in accent-based applications within speech processing systems. Speech samples were collected, comprising 414 British and 410 American English recordings, to construct a dataset representative of both accent groups. Key features were extracted from the audio data, enabling machine learning models to differentiate between the two accents effectively. Four machine learning models were evaluated: Naive Bayes, K-Nearest Neighbors (KNN), Random Forest, and Decision Tree. Among these, Naive Bayes demonstrated the highest accuracy at 84.24%, highlighting its effectiveness in capturing the distinguishing features of British and American English accents. KNN followed closely with an accuracy of 78.79%, benefiting from its proximity-based classification mechanism. Random Forest achieved an accuracy of 78.18%, leveraging ensemble learning to improve prediction stability. The Decision Tree model, while functional, demonstrated the lowest performance at 76.36%, indicating limitations of single-tree approaches in capturing nuanced differences in speech patterns. The findings underscore the potential of machine learning in accent classification and reveal significant differences in model performance based on algorithmic design. These results contribute to advancing research in automatic speech recognition, accent identification, and related applications in natural language processing. Future work could explore larger datasets, deep learning approaches, and feature optimization to further enhance classification accuracy. By effectively distinguishing between British and American accents, this research lays the foundation for improved speech-based systems and broader linguistic studies.