Measuring Dialect Distance Phonetically

We describe ongoing work in the experi mental evaluation of a range of methods for measuring the phonetic distance be tween the dialectal variants of pronuncia tions All are variants of Levenshtein dis tance both simple based on atomic char acters and complex based on feature vec tors The measurements using feature vec tors varied according to whether city block distance Euclidean distance or a measure using Pearson s correlation coe cient was taken as basic Variants of these using fea ture weighting by entropy reduction were systematically compared as was the rep resentation of diphthongs as one symbol or two The results were compared to well established scholarship in dialectology yielding a calibration of the method These results indicate that feature representa tions are more sensitive that city block dis tance is a good measure of phonetic over lap of feature vectors that weighting is not useful and that two phone representations of diphthongs provide a more satisfactory base for this sort of comparison