Responsive Navbar | CodingNepal

Group Joint Engagement Prediction

< Projects
Nov 2022 ~ Present
Personal Robots Group,
MIT Media Lab

MultiPar-T: Multiparty-Transformer for Capturing Contingent Behaviors in Group Conversations

Research Topics

# engagement prediction # cross-party attention

Overview

Understanding a multiparty setting is more challenging than that of an individual setting due to the fact that (1) the model must first recognize individual behavioral cues, (2) deal with the complexity of capturing these cues from multiple streams of behavioral data from the multiple people, and most importantly (3) recognize the subtle contingent behaviors that take place amongst group members. To tackle this challenge, we propose the Multiparty-Transformer (Multipar-T), a transformer model for group behavior modelling.

Publications

Dong Won Lee, Yubin Kim, Rosalind Picard, Cynthia Breazeal, Hae Won Park, "MultiPar-T: Multiparty-Transformer for Capturing Contingent Behaviors in Group Conversations." in preparation for IJCAI 2023