AccentClustering

This project proposes a method to group accents based on phonetic similarities using Hierarchical clustering. Audio from different accents is compared using Dynamic Time Warping similarity measure, performed at the mono-phone and tri-phone level. A dendrogram is generated for a visual representation of clusters, which also gives an idea on the number of clusters to be used. Agglomerative clustering is performed and clusters are measured empirically and using the Davies Bouldin and Silhouette metrics. Audio recordings are taken from the Mozilla-Common Voice dataset