Classifying and Clustering Dialects of North American English

This paper presents the results of experiments in which machine learning techniques were applied to the problem of determining regional dialect boundaries. Specifically, decision trees classification and k-means clustering were applied to a corpus of phonetic measurements taken from a large survey of North American English vowels. Pairwise classification and clustering experiments were done for all combinations of ten dialect regions determined by dialectologists. The results show which of these dialect regions are most distinct and similar, suggesting which of the distinctions that are usually used by linguists are the most meaningful. Furthermore, the classification trees are analyzed to show which vowel formants are most informative for each dialect region.