Using Pose Estimation for Reference Tracking in Turkish Sign Language

In this study, native and late deaf adult signers of Turkish Sign Language were asked to narrate the cartoon clips they watched. The expressions used by the signers were analyzed using the method of pose estimation to calculate (i) the spatial and angular changes of the body and (ii) the emotional intensity scores for each referent. The average spatial and angular changes in the participants' linguistic structures were found to significantly differ depending on the iconicity of the sign. Furthermore, our findings indicate the late signers may use facial expressions or non-manual signs more for affective communication, while native signers may exhibit more grammatical use of non-manual markers. This supports the notion in the literature that learning Turkish Sign Language may lead to certain pragmatic differences.