MOSAIC : a Corpus of Small-Group Interactions During a Collaborative Task

Amine Benamara, Celine Clavel, Brian Ravenet, Nicolas Sabouret, Mathilde Sassier–Roublin, Julien Saunier


Abstract
This paper presents MOSAIC (Multimodal Observations of Social Affect, Intimacy, and Cohesion), a multimodal interaction corpus of video and audio recordings of 17 groups of 4 participants (68 participants in total) playing a collaborative board game. The aim of this corpus collection is to support the design of socially interactive agents. We describe the experimental protocol of this corpus collection, the perceptive questionnaires completed by participants, the automatic annotation process of game specific elements and non-verbal behaviors, and the manual verbal annotations collected. We provide a preliminary characterization of this corpus with descriptive results for some of the perceptive scales used in this study. We illustrate the possibilities offered by this corpus on the question of interpersonal social relations and group dynamics.
Anthology ID:
2026.cas-1.14
Volume:
Proceedings of Computational Affective Science (CAS) @ LREC 2026
Month:
May
Year:
2026
Address:
Palma, Mallorca (Spain)
Editors:
Christopher Bagdon, Krishnapriya Vishnubhotla, Kristen A. Lindquist, Lyle Ungar, Roman Klinger, Saif M. Mohammad
Venues:
CAS | WS
SIG:
Publisher:
ELRA Language Resources Association (ELRA)
Note:
Pages:
164–174
Language:
External URL:
https://lrec.elra.info/lrec2026-ws-cas-14
DOI:
10.63317/59gt7h7mibpf
Bibkey:
Cite (ACL):
Amine Benamara, Celine Clavel, Brian Ravenet, Nicolas Sabouret, Mathilde Sassier–Roublin, and Julien Saunier. 2026. MOSAIC : a Corpus of Small-Group Interactions During a Collaborative Task. In Proceedings of Computational Affective Science (CAS) @ LREC 2026, pages 164–174, Palma, Mallorca (Spain). ELRA Language Resources Association (ELRA).
Cite (Informal):
MOSAIC : a Corpus of Small-Group Interactions During a Collaborative Task (Benamara et al., CAS 2026)
Copy Citation: