Approaching variation in the Phonologie du Français Contemporain project: the segmental level

Current studies in phonological and phonetic variation make an ever increasing use of large oral corpora such as the one developed in the framework of the Phonologie du français contemporain project. The exploitation of these large-scale corpora has fostered the development of powerful techniques for the processing and analysis of speech data. These include tools for the automatic segmentation, phonetic labelling, and extraction of phonetic parameters from speech recordings, as well as a wide variety of probabilistic data analysis methods. Perceptual experiments have also been conducted to better characterize how phonological/phonetic variation is dealt with in large groups of listeners by means of experimental designs that allow the results to be generalized to yet larger populations. In this chapter, we give a brief overview on these tools and methods, using cross-disciplinary work recently carried out on the distribution and phonetic realization of mid vowels in different varieties of French as a primary example. Probabilistic data analysis methods and experimental studies of speech processing both open new and exciting avenues for research on segmental variation