Multi-Layered Annotations for Polyadic conversation: Interactional labels from the GaMMA corpus

Computational models of dialogue often struggle to capture the nuanced structures of spontaneous conversation, particularly in polyadic, real-world settings. This dataset introduces a multi-layered annotation protocol designed for the GaMMA corpus, a Danish dataset of four-person conversations recorded in both quiet and noisy environments. The protocol targets key interactional phenomena, including Turn Construction Units, backchannels, floor transfer attempts, and repair sequences. Each annotation layer is grounded in Conversation Analysis while remaining machine-actionable, enabling alignment with multimodal data such as gaze and motion. Inter-annotator agreement metrics are reported across annotation tiers, and the dataset supports both fine-grained interaction analysis and the training of context-aware dialogue models.The archive contains raw ELAN annotations, aligned CSV files, and summary statistics for 44 four-person conversations across four acoustic conditions. A detailed README describing file structure, annotation tiers, and usage is included in the repository.