Modelled Multivariate Overlap: A method for measuring vowel merger
This provides a more accurate method for linguists studying vowel mergers, though it is incremental as it builds on existing overlap measurement techniques.
The paper tackles the problem of quantifying vowel overlap in speech analysis by introducing a method that models acoustic dimensions and simulates distributions to compute overlap, resolving tensions between multivariate measures and control for unbalanced data. The method, evaluated on PIN-PEN merger data in four English dialects, shows that using modelled distributions with Bhattacharyya affinity substantially improves results over empirical distributions.
This paper introduces a novel method for quantifying vowel overlap. There is a tension in previous work between using multivariate measures, such as those derived from empirical distributions, and the ability to control for unbalanced data and extraneous factors, as is possible when using fitted model parameters. The method presented here resolves this tension by jointly modelling all acoustic dimensions of interest and by simulating distributions from the model to compute a measure of vowel overlap. An additional benefit of this method is that computation of uncertainty becomes straightforward. We evaluate this method on corpus speech data targeting the PIN-PEN merger in four dialects of English and find that using modelled distributions to calculate Bhattacharyya affinity substantially improves results compared to empirical distributions, while the difference between multivariate and univariate modelling is subtle.