Dynamics of Collective Group Affect: Group-level Annotations and the Multimodal Modeling of Convergence and Divergence

This repository contains the codebase for the paper:

Navin Raj Prabhu, Maria Tsfasman, Catharine Oertel, Timo Gerkmann, Nale Lehmann-Willenbrock, "Dynamics of Collective Group Affect: Group-level Annotations and the Multimodal Modeling of Convergence and Divergence", Under Review, Submitted to IEEE Transactions on Affective Computing, Sep, 2024. arxiv

The repository contains the following:

Data and analysis

1. Data Reader and Classes
2. Annotation Analysis
3. Inter-annotator agreement
4. Ground-truth consensus

Modeling

1. Individual-level feature extractors
   1. Audio (prosody)
   2. Video
2. Dyad-level feature extractors
   1. Synchrony (correlation, lagged correlation)
   2. Convergence (global, symmetric and asymmetric)
3. Convergence and Divergence analysis
4. Group Affect recognition baselines