README.md

Cite Us

I'm pleased to announce that this work is accepted by Frontiers in Psychology https://www.frontiersin.org/articles/10.3389/fpsyg.2021.644801/full!

File Structure

├── Data
│   ├── en_dup.csv
│	└── news
│	└── twitter
├── Data Analysis
│   └── PowerLaw Analysis.pdf
├── Data Collecting
│   └── snopes.py
├── LICENSE
└── README.md

Data Collecting

• snopes.py by Tianqi
  • It is used to collect data from website www.snopes.com and qc.wa.news.cn (departed)

Data Analysis

• PowerLaw Analysis includes the statistical analysis of rumor popularity data.

Pre-processed Data for Deep Learning Tasks

• https://drive.google.com/drive/folders/1jpnyHw1_XWR7lZTSb86AYWFsNC0bNEwO?usp=sharing.

Deep Learning Models in Rumor Classification

• Sentiment analysis: https://colab.research.google.com/drive/1CyGtXPSol_Ayt_WR7zsq60M2Q5kDUofd?usp=sharing#scrollTo=GzCJKcoWT1aa.
• Stance classification: https://colab.research.google.com/drive/1ItGmEvumyOesXeB2dV7w5cnjRW9fTAKL?usp=sharing.
• Veracity prediction: https://colab.research.google.com/drive/1qtt0X9a7I9vBZPoZ-PzyhDuF5L_G3thj?usp=sharing.

Data

• news

  • news.csv (4129) and subfolder of each news
  • The number of subfolder records: 3936

• twitter

  • Twitter.csv (2705) and subfolder of each twitter
  • The number of subfolder records: 1383

• en_dup.csv

  • Unprocessed data with both news and twitter records.
  • The number of records: 7179 (with duplication).
  • Part of data are collected manually by keywords searching from sources such as twitter.com.
  • Data from www.snopes.com and qc.wa.news.cn are collected by 'snopes.py'.

Acknowledgement

• We thank Tianqi, Wenshuo, Jianni, Xiaofeng, and Hanlong for rumor data collection and labeling.