A cross-lingual COVID-19 fake news dataset

Overview

CrossFake

An English-Chinese COVID-19 fake&real news dataset from the ICDMW 2021 paper below:
Cross-lingual COVID-19 Fake News Detection.
Jiangshu Du, Yingtong Dou, Congying Xia, Limeng Cui, Jing Ma, Philip S. Yu.

Introduction

The COVID-19 pandemic poses a significant threat to global public health. Meanwhile, there is massive misinformation associated with the pandemic, which advocates unfounded or unscientific claims. Even major social media and news outlets have made an extra effort in debunking COVID-19 misinformation, most of the fact-checking information is in English, whereas some unmoderated COVID-19 misinformation is still circulating in other languages, threatening the health of less informed people in immigrant communities and developing countries (The Vox, New York Times).

In the above paper, we make the first attempt to detect COVID-19 misinformation in a low-resource language (Chinese) only using the fact-checked news in a high-resource language (English).

This repo contains a Chinese-English real & fake news dataset according to existing English fact-checking information. Details on this dataset are described in Dataset Detail.

The highlights of our dataset are as follows:

  • Bilingual news pieces for the same event (fact).
  • Multiple Chinese news pieces for the same event (fact).
  • Comprehensive metadata for each news (see below).

Dataset Detail

The table below shows the number of annotated news in each language:

Lang. Fake Real Total
ENG 55 82 137
CHN 101 118 219

The metadata of our dataset can be found at CrossFake_metadata.xlsx, which includes two sheets (news_fake and news_real). Given the news id, you can find the corresponding news body text in the body_text directory. The meanings of each column of the metadata are shown below:

  • Column A (id):

    News id. Chinese real & fake news is annotated according to existing English fact-checking information. Thus, each piece of English news may correspond to multiple pieces of Chinese news from different sources. For example, in the news_fake sheet, the ids 1_1 and 1_2 indicate one piece of English news, corresponding to two pieces of Chinese news.

  • Column B (fact_check_url):

    The fact-checking source of the corresponding English news.

  • Column C (type):

    The news type. Post and Article represent the news is from a social media post or an online article, respectively. Note that we also annotated some clickbait news whose title and body text present contradictory information.

  • Column D (source):

    The news source. Personal and Professional represent the news is from a personal account or professional source (WHO, NIH, etc.), respectively.

  • Column E (mixed?):

    Whether the news include mixed content? If a news body text only has the content related to the checked fact, the piece of news is annotated as not mixed. Accordingly, the news whose content includes events/facts besides the checked fact is regarded as mixed news.

  • Column F (platform):

    The platform where the news is published.

  • Column G (news_url):

    The news source URL. Note that some of the links are invalid due to the deletion/removal of the news. We have archived the accessible news (see Column H) during we curate the dataset.

  • Column H (archive):

    The archived news link. To permanently store the original news, we archived the news source URL.

  • Column I (newstitle):

    The news title.

  • Column J (publish_date):

    The news publishing date.

  • Columns K to R have the same meanings as Columns C to J, but they indicate the information of Chinese news.

Case Study

Besides the findings and conclusions presented in our paper. We have extra interesting findings during collecting the data:

  1. Mixed Fact. For some fake news, their corresponding Chinese news articles presented them in the form of a news digest with other news events. It brings an extra hurdle to fact-check those news pieces since only partial content of the news contains misinformation. A typical example is news_id 8_3 in the news_fake sheet. You can check out other news whose mixed? annotated as Yes.

  2. Misused Fact. For news_real id 9_2, we find a Chinese social post leveraging the fact that "coronavirus can live for up to 4 hours on copper" to promote their copper-made pot. In this case, even the title and most of the news content seem legit, but the connection between "the copper kills coronavirus" and "copper pot is good" is still questionable.

  3. Fake News Type. During we annotate the Chinese news based on the fact-checked English news. We find that most of the fact-checked fake news from Politifact have no corresponding Chinese news. Those news pieces usually are local news in the United States.

  4. Cross-lingual Fact-checking. For the news_real id 9_1, we find a Chinese news piece from a professional news outlet published five days earlier than the fact-checked English Facebook post. It suggests that we could leverage fact information from another language to help fact-check the news. Note that most of the Chinese news in our datasets are published later than the source English news since most of the checked news events are originated in English media.

Future Directions

Given the current dataset, some future research directions include:

  • The writing style/sentiment/stance differences between fake news and real news.
  • The writing style/sentiment/stance differences between professional news outlets and personal accounts.
  • The information distortion/loss from English news to Chinese news.
  • The temporal patterns of cross-lingual news migration.
  • The title patterns of different news.

Citation

If you use our code, please cite the paper below:

@inproceedings{du2021cross,
  title={Cross-lingual COVID-19 Fake News Detection},
  author={Du, Jiangshu and Dou, Yingtong and Xia, Congying and Cui, Limeng and Ma, Jing and Yu, Philip S},
  booktitle={Proceedings of the 21st IEEE International Conference on Data Mining Workshops (ICDMW'21)},
  year={2021}
}
Owner
Yingtong Dou
Ph.D. @ UIC. Graph Mining; Fraud Detection; Secure Machine Learning
Yingtong Dou
Here we present the implementation in TensorFlow of our work about liver lesion segmentation accepted in the Machine Learning 4 Health Workshop

Detection-aided liver lesion segmentation Here we present the implementation in TensorFlow of our work about liver lesion segmentation accepted in the

Image Processing Group - BarcelonaTECH - UPC 96 Oct 26, 2022
TensorFlow port of PyTorch Image Models (timm) - image models with pretrained weights.

TensorFlow-Image-Models Introduction Usage Models Profiling License Introduction TensorfFlow-Image-Models (tfimm) is a collection of image models with

Martins Bruveris 227 Dec 20, 2022
Music Source Separation; Train & Eval & Inference piplines and pretrained models we used for 2021 ISMIR MDX Challenge.

Music Source Separation with Channel-wise Subband Phase Aware ResUnet (CWS-PResUNet) Introduction This repo contains the pretrained Music Source Separ

Lau 100 Dec 25, 2022
Rate-limit-semaphore - Semaphore implementation with rate limit restriction for async-style (any core)

Rate Limit Semaphore Rate limit semaphore for async-style (any core) There are t

Yan Kurbatov 4 Jun 21, 2022
nnFormer: Interleaved Transformer for Volumetric Segmentation Code for paper "nnFormer: Interleaved Transformer for Volumetric Segmentation "

nnFormer: Interleaved Transformer for Volumetric Segmentation Code for paper "nnFormer: Interleaved Transformer for Volumetric Segmentation ". Please

jsguo 610 Dec 28, 2022
PyTorch Implementation of Exploring Explicit Domain Supervision for Latent Space Disentanglement in Unpaired Image-to-Image Translation.

DosGAN-PyTorch PyTorch Implementation of Exploring Explicit Domain Supervision for Latent Space Disentanglement in Unpaired Image-to-Image Translation

40 Nov 30, 2022
[CVPR 2022 Oral] MixFormer: End-to-End Tracking with Iterative Mixed Attention

MixFormer The official implementation of the CVPR 2022 paper MixFormer: End-to-End Tracking with Iterative Mixed Attention [Models and Raw results] (G

Multimedia Computing Group, Nanjing University 235 Jan 03, 2023
A port of muP to JAX/Haiku

MUP for Haiku This is a (very preliminary) port of Yang and Hu et al.'s μP repo to Haiku and JAX. It's not feature complete, and I'm very open to sugg

18 Dec 30, 2022
Large-scale open domain KNOwledge grounded conVERsation system based on PaddlePaddle

Knover Knover is a toolkit for knowledge grounded dialogue generation based on PaddlePaddle. Knover allows researchers and developers to carry out eff

607 Dec 31, 2022
Code for "Learning Structural Edits via Incremental Tree Transformations" (ICLR'21)

Learning Structural Edits via Incremental Tree Transformations Code for "Learning Structural Edits via Incremental Tree Transformations" (ICLR'21) 1.

NeuLab 40 Dec 23, 2022
An efficient PyTorch implementation of the winning entry of the 2017 VQA Challenge.

Bottom-Up and Top-Down Attention for Visual Question Answering An efficient PyTorch implementation of the winning entry of the 2017 VQA Challenge. The

Hengyuan Hu 731 Jan 03, 2023
Official implementation of "Implicit Neural Representations with Periodic Activation Functions"

Implicit Neural Representations with Periodic Activation Functions Project Page | Paper | Data Vincent Sitzmann*, Julien N. P. Martel*, Alexander W. B

Vincent Sitzmann 1.4k Jan 06, 2023
🤗 Transformers: State-of-the-art Natural Language Processing for Pytorch, TensorFlow, and JAX.

English | 简体中文 | 繁體中文 | 한국어 State-of-the-art Natural Language Processing for Jax, PyTorch and TensorFlow 🤗 Transformers provides thousands of pretrai

Hugging Face 77.4k Jan 05, 2023
Invert and perturb GAN images for test-time ensembling

GAN Ensembling Project Page | Paper | Bibtex Ensembling with Deep Generative Views. Lucy Chai, Jun-Yan Zhu, Eli Shechtman, Phillip Isola, Richard Zhan

Lucy Chai 93 Dec 08, 2022
Designing a Minimal Retrieve-and-Read System for Open-Domain Question Answering (NAACL 2021)

Designing a Minimal Retrieve-and-Read System for Open-Domain Question Answering Abstract In open-domain question answering (QA), retrieve-and-read mec

Clova AI Research 34 Apr 13, 2022
Tools for investing in Python

InvestOps Original repository on GitHub Original author is Magnus Erik Hvass Pedersen Introduction This is a Python package with simple and effective

24 Nov 26, 2022
Pytorch implementation of the AAAI 2022 paper "Cross-Domain Empirical Risk Minimization for Unbiased Long-tailed Classification"

[AAAI22] Cross-Domain Empirical Risk Minimization for Unbiased Long-tailed Classification We point out the overlooked unbiasedness in long-tailed clas

PatatiPatata 28 Oct 18, 2022
This repository contains the source code and data for reproducing results of Deep Continuous Clustering paper

Deep Continuous Clustering Introduction This is a Pytorch implementation of the DCC algorithms presented in the following paper (paper): Sohil Atul Sh

Sohil Shah 197 Nov 29, 2022
Code, Data and Demo for Paper: Controllable Generation from Pre-trained Language Models via Inverse Prompting

InversePrompting Paper: Controllable Generation from Pre-trained Language Models via Inverse Prompting Code: The code is provided in the "chinese_ip"

THUDM 101 Dec 16, 2022
Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm

DeCLIP Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm. Our paper is available in arxiv Updates ** Ou

Sense-GVT 470 Dec 30, 2022