DCTM: Dilated Convolutional Transformer Model for Multimodal Engagement Estimation in Conversation

TL;DR

This paper introduces a dilated convolutional Transformer model for estimating conversational engagement, achieving significant improvements over baselines by effectively fusing multimodal data.

Contribution

The study presents a novel dilated convolutional Transformer architecture and demonstrates its effectiveness in multimodal engagement estimation in conversations.

Findings

7% improvement on test set over baselines

Simple concatenation with self-attention fusion performs best

Effective multimodal fusion enhances engagement estimation

Abstract

Conversational engagement estimation is posed as a regression problem, entailing the identification of the favorable attention and involvement of the participants in the conversation. This task arises as a crucial pursuit to gain insights into human's interaction dynamics and behavior patterns within a conversation. In this research, we introduce a dilated convolutional Transformer for modeling and estimating human engagement in the MULTIMEDIATE 2023 competition. Our proposed system surpasses the baseline models, exhibiting a noteworthy $7$\% improvement on test set and $4$\% on validation set. Moreover, we employ different modality fusion mechanism and show that for this type of data, a simple concatenated method with self-attention fusion gains the best performance.