Skip to content

Latest commit

 

History

History
65 lines (53 loc) · 3.2 KB

index.md

File metadata and controls

65 lines (53 loc) · 3.2 KB
layout mathjax title subtitle use-site-title meta-description
page
true
MuMiN
A Large-Scale Multilingual Multimodal Fact-Checked Misinformation Social Network Dataset
true
A Large-Scale Multilingual Multimodal Fact-Checked Misinformation Social Network Dataset

The MuMiN dataset is a challenging misinformation benchmark for automatic misinformation detection models. The dataset is structured as a heterogeneous graph and features 21,565,018 tweets and 1,986,354 users, belonging to 26,048 Twitter threads, discussing 12,914 fact-checked claims from 115 fact-checking organisations in 41 different languages, spanning a decade.

Tasks

The dataset has three different sizes and features two graph classification tasks:

  • Claim classification: Given a claim and its surrounding subgraph extracted from social media, predict whether the verdict of the claim is misinformation or factual.
  • Tweet classification: Given a source tweet (i.e., not a reply, quote tweet or retweet) to be fact-checked and its surrounding subgraph extracted from social media, predict whether the tweet discusses a claim whose verdict is misinformation or factual.

Getting Started

See Getting Started for a quickstart as well as an in-depth tutorial, including the building and training of multiple misinformation classifiers on MuMiN.

Tutorial

We have created a tutorial which takes you through the dataset as well as shows how one could create several kinds of misinformation classifiers on the dataset. The tutorial can be found here.

Leaderboard

See the leaderboard for a list of the best performing models. For new submissions, please email [email protected].

Citation

If you use this dataset, you can cite it as follows:

Dan Saattrup Nielsen and Ryan McConville. MuMiN: A large-scale multilingual multimodal fact-checked misinformation social network dataset. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR). ACM, 2022 (to appear). bib

or

Dan Saattrup Nielsen and Ryan McConville. "MuMiN: A Large-Scale Multilingual Multimodal Fact-Checked Misinformation Social Network Dataset." arXiv preprint arXiv:2202.11684 (2022). bib

Other MuMiN repos

You can also check out the following MuMiN-related Github repositories:

  • MuMiN-build, containing the source code for our Python package, mumin, which allows easy compilation and export of the dataset.
  • MuMiN-baseline, containing the source code for all our baseline experiments, the scores of which are in the leaderboard.
  • MuMiN-trawl, containing our data collection and linking system, which can be used to extend the MuMiN dataset.