Modelling the Interaction Space of Twitch: A Multimodal Framework for Corpus Structuring and Analysis

This paper proposes a model of the interaction space of Twitch, based on and adapted from the CoMeRe framework for computermediated communication. The model addresses three core challenges of Twitch as a communicative environment: its multimodality, the asymmetric access to communicative modes based on participant roles, and the influence of non-verbal events such as gameplay or interface actions. We describe the different roles involved (broadcaster, viewer, moderator, bot) and map their access to six communicative modes. This model supports both theoretical reflection on role-based discourse production and the technical development of multimodal corpora. As a concrete application, we extended S. Coats’s alignment tool by implementing role-based pseudonymisation and identifying areas for future extension, including the integration of non-verbal data layers. This work lays the foundation for corpus-based linguistic and discursive studies of live-streaming platforms and invites further inquiry into role-asymmetric and intermodal communication environments.