MTPChat: A Multimodal Time-Aware Persona Dataset for Conversational Agents

TL;DR

MTPChat is a novel multimodal, time-aware dialogue dataset designed to enhance conversational agents' understanding of temporal and persona dynamics, introducing new tasks and an adaptive framework for improved temporal reasoning.

Contribution

We introduce MTPChat, a comprehensive multimodal, time-aware dataset with new tasks and a framework for modeling temporal and multimodal aspects in conversations.

Findings

Our dataset reveals significant challenges for current models in temporal understanding.

The proposed framework effectively captures temporal dependencies in multimodal dialogues.

Experimental results show improved performance on time-sensitive tasks.

Abstract

Understanding temporal dynamics is critical for conversational agents, enabling effective content analysis and informed decision-making. However, time-aware datasets, particularly for persona-grounded conversations, are still limited, which narrows their scope and diminishes their complexity. To address this gap, we introduce MTPChat, a multimodal, time-aware persona dialogue dataset that integrates linguistic, visual, and temporal elements within dialogue and persona memory. Leveraging MTPChat, we propose two time-sensitive tasks: Temporal Next Response Prediction (TNRP) and Temporal Grounding Memory Prediction (TGMP), both designed to assess a model's ability to understand implicit temporal cues and dynamic interactions. Additionally, we present an innovative framework featuring an adaptive temporal module to effectively integrate multimodal streams and capture temporal dependencies. Experimental results validate the challenges posed by MTPChat and demonstrate the effectiveness of our framework in multimodal time-sensitive scenarios.