Italian Sign Language: Searching for common ground between the sociosemiotic perspective and computational annotation

This paper presents a protocol for annotating Italian Sign Language (LIS) that merges cognitive and socio-semiotic principles with the technical demands of sign language processing for automatic sign recognition. The protocol is applied to a preliminary LIS dataset and integrates insights from both formalist and functionalist frameworks in sign language processing, aiming to leverage their respective strengths to create a comprehensive and accessible documentation of LIS. Unlike traditional gloss translations, which often introduce ambiguity by not directly associating signs with their meanings and relying on verbal language categories, the proposed annotation model incorporates a multi-layered approach that includes vocal language labels in written Italian and English, information on the Unit of Meaning performed combined with the implementation Typannot, a language-specific system sign languages annotation. This hybrid approach ensures that the annotated data is both human- and computer-readable, enhancing accessibility for both signers and non-signers. The presented multi-layered annotation model not only mitigates ambiguity but also provides a richer, more precise annotation. Initial findings suggest that this protocol can enhance the clarity and usability of annotated data, positioning it as a valuable resource for both linguistic research and technological applications.