WildFireCan-MMD: A Multimodal Dataset for Classification of User-Generated Content During Wildfires in Canada

Braeden Sherritt, Isar Nejadgholi, Efstratios Aivaliotis, Khaled Mslmani, Marzieh Amini · 2025

Read WildFireCan-MMD: A Multimodal Dataset for Classification of User-Generated Content During Wildfires in Canada by Braeden Sherritt, Isar Nejadgholi, Efstratios Aivaliotis, Khaled Mslmani, Marzieh Amini (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-13231.txt.

Description

Braeden Sherritt and four co-authors present a 2025 arXiv paper with a new data set on wildfire posts from social media. During Canadian wildfires, residents post fast updates mixing text and images, but sorting them by theme is slow by hand. Existing data sets barely cover Canada, and disaster needs differ by region.

The team gathered posts from recent Canadian fires and labeled them across twelve themes. They then raced three kinds of classifiers: vision-language models used cold, simple baselines, and models trained on their labeled data. The trained models won clearly, with the best reaching a strong score well above the rest. Applying that model to a large unlabeled pool revealed trends across the fire events. The authors argue local, labeled data beats general tools for disaster response, and they share the set for future work.

Catalog record

CREATORBraeden Sherritt, Isar Nejadgholi, Efstratios Aivaliotis, Khaled Mslmani, Marzieh Amini
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-13231

Browse the full library · About the Tower of Babel