In Turkish, the “r” is pronounced as an alveolar tap as in the word araba ˈɑ.ɾə.bə “car.” Impressionistic descriptions of the Turkish tap say that word-initial taps may be fricated as in the word resim ɾ̌e.ˈsim “painting” and that word-final taps are devoiced and fricated as in the word bir biɾ̌ ̥ “one.” Little previous research has been done to explore these acoustic characteristics and what may predict this frication. This pilot study uses the Turkish dataset from the Mozilla Common Voice corpus to examine if tap duration and segment position are predictors of frication and if the first three mel frequency cepstral coefficients are discriminative for frication on alveolar taps. The data will be analyzed using generalized additive mixed modeling, the results of which will be discussed with relation to the distribution of frication on the tap in Turkish.
Miklas et al. (Tue,) studied this question.