WEBVTT 00:01.000 --> 00:06.500 This study makes use of 2 hours of dyadic conversational data from the Swedish Sign Language Corpus, 00:06.600 --> 00:09.000 in order to answer two main research questions. 00:09.100 --> 00:16.500 Using the computer vision model Mediapipe as a tool for analyzing sign language data directly from video, is it possible to: 00:16.600 --> 00:26.800 1. Determine who is signing out of the two signers in a dyad? 2. Determine which of each signer's two hands is their dominant one overall? 00:27.900 --> 00:34.000 The methodology involves looking at the distance traveled by the signers' hands as well as the height of the hands in signing space. 00:35.000 --> 00:41.000 The results show that there is a high degree of accuracy in determining the main signer in different segments of conversation. 00:41.000 --> 00:49.000 Determining the signers' hand dominance works best if one first classifies segments into active signing sequences only, 00:49.100 --> 00:52.400 to remove noise from backchanneling, fidgeting and grooming.