Overview of ImageArg-2023: The First Shared Task in Multimodal Argument Mining

TL;DR

This paper overviews the ImageArg-2023 shared task, the first multimodal argument mining challenge, which involved classifying argument stance and image persuasiveness in social media posts, with notable participation and performance results.

Contribution

It introduces the first shared task on multimodal argument mining, providing a benchmark for argument stance and persuasiveness classification involving images and text.

Findings

High top F1-score of 0.8647 for argument stance classification

Moderate F1-score of 0.5561 for image persuasiveness classification

Participation from 9 teams across 6 countries

Abstract

This paper presents an overview of the ImageArg shared task, the first multimodal Argument Mining shared task co-located with the 10th Workshop on Argument Mining at EMNLP 2023. The shared task comprises two classification subtasks - (1) Subtask-A: Argument Stance Classification; (2) Subtask-B: Image Persuasiveness Classification. The former determines the stance of a tweet containing an image and a piece of text toward a controversial topic (e.g., gun control and abortion). The latter determines whether the image makes the tweet text more persuasive. The shared task received 31 submissions for Subtask-A and 21 submissions for Subtask-B from 9 different teams across 6 countries. The top submission in Subtask-A achieved an F1-score of 0.8647 while the best submission in Subtask-B achieved an F1-score of 0.5561.