M-CATNAT: A Multimodal dataset to analyze French tweets during natural disasters

This repository contains a multimodal tweet annotation dataset aimed at enhancing disaster response efforts and social media analysis. Below is a breakdown of the files included in this repository: Files 1. "annotations.csv" This CSV file serves as the main dataset, containing annotations for each tweet instance. Each row represents a unique tweet, with the following columns: tweet_id: The unique identifier for the tweet. text_label: The annotation label for the tweet text. image_label: The annotation label for the tweet image. multimodal_label: The annotation label for the the multimodal instance . Notes: In cases where a tweet contains multiple images, the tweet IDs are followed by numbers separated by underscores. For instance, "tweet_id_1" represents the first image in the tweet URLs list obtained from the Twitter API. 2. "annotation_guide_v2.html" This document presents the second version of our annotation guide. It offers instructions and examples to resolve any potential disagreements among annotators, ensuring consistency in labeling the dataset. Dataset Description The dataset consists of multimodal tweets related to natural disasters, sourced from Twitter. Each tweet is annotated based on predefined criteria outlined in the annotation guide.