Asymmetry in Transition Times in American Sign Language

To enhance the legibility and naturalness of sign language generation by avatars, this study analyzes the impact of linguistic function on the transition time between signs in American Sign Language. Using annotation data from the Boston University NCSLGR Corpus, we categorize the signs into Lexical, Gesture, Classifier, Index, and Fingerspelling. We compare the transition times between categories in terms of category bigrams, the categorical representation of two consecutive signs. We take a broad approach that does not rely on sentences or sentence structure. Though we need more data for definitive conclusions, we observe that the median transition time between category bigrams is statistically different based on which categories are in the bigram. We observe that lexical items and lexical-to-lexical bigrams occur much more frequently than the other categories and category bigrams. With further study and evidence, this data will inform the transition time for our sign language avatar.