Multi-modal Stance Detection: New Datasets and Model

TL;DR

This paper introduces five new multi-modal stance detection datasets from Twitter and proposes TMPT, a framework that effectively combines text and image data to improve stance detection accuracy.

Contribution

The paper presents new multi-modal datasets and a novel TMPT framework that leverages target information for improved stance detection from text and images.

Findings

TMPT achieves state-of-the-art results on five datasets.

Multi-modal data improves stance detection performance.

New datasets cover diverse domains for comprehensive evaluation.

Abstract

Stance detection is a challenging task that aims to identify public opinion from social media platforms with respect to specific targets. Previous work on stance detection largely focused on pure texts. In this paper, we study multi-modal stance detection for tweets consisting of texts and images, which are prevalent in today's fast-growing social media platforms where people often post multi-modal messages. To this end, we create five new multi-modal stance detection datasets of different domains based on Twitter, in which each example consists of a text and an image. In addition, we propose a simple yet effective Targeted Multi-modal Prompt Tuning framework (TMPT), where target information is leveraged to learn multi-modal stance features from textual and visual modalities. Experimental results on our five benchmark datasets show that the proposed TMPT achieves state-of-the-art performance in multi-modal stance detection.