|Title||Inducing phonetic distances from dialect variation|
|Publication Type||Journal Article|
|Year of Publication||2011|
|Authors||Wieling, M, Margaretha, E, Nerbonne, J|
|Journal||Computational Linguistics in the Netherlands Journal|
In this study we attempt to derive phonetic distances from alternative dialectal pronunciations used in different geographical varieties. We use two dialect atlases each containing the phonetic transcriptions of the same set of words at hundreds of sites. We collect the sound correspondences through alignment with the Levenshtein distance algorithm, and then apply an information-theoretic measure, pointwise mutual information, assigning smaller segment distances to segments which frequently correspond. We iterate alignment and information-theoretic distance assignment until both stabilize and we evaluate the quality of the phonetic distances obtained by comparing them to acoustic vowel distances. For both Dutch and German, we find strong correlations between the induced phonetic distances and the acoustic distances, illustrating the usefulness of the method in deriving valid phonetic distances from dialectal pronunciations.