kelaxon / covid19-health-rumor

Annotated dataset for the paper "Know it to Defeat it: Exploring Health Rumor Characteristics and Debunking Efforts on Chinese Social Media during COVID-19 Crisis"

Health Rumors and Debunking (Fact-checking) Efforts on Weibo During COVID-19

This repository contains annotated data associated with the ICWSM 2022 paper:

W. Yang, S. Wang, Z. Peng, C. Shi, X. Ma, D. Yang. Know it to Defeat it: Exploring Health Rumor Characteristics and Debunking Efforts on Chinese Social Media during COVID-19 Crisis. International AAAI Conference on Web and Social Media (ICWSM). 2022. [Preprinted pdf]

This dataset consists of health rumors circulated on China's internet during early COVID-19, as well as posts on Sina Weibo (China's largest microblogging website) intended to refute or debunk these rumors. Unlike conspiracy theories, health rumors are about healthcare and medicine and do not involve a primary actor (e.g., the U.S. military). In our paper, we categorize health rumors into two categories: wish (which inspires hope) and dread (which causes fear). We examine their content and propagation characteristics, as well as the efforts from different groups of users to debunk them.

Data Format

data/health_rumors.csv contains 408 health rumors that we manually identified as having appeared on China's Internet during the early four months of COVID-19 crisis.

• rumor_id: unique identifier for the health rumor;
• publish_date: published date of the fact-checking article;
• rumor_type: type of the health rumor that includes wish or dread;
• article_title: title of the fact-checking article;
• article_content: content of the fact-checking article;
• search_expression: query string we used to retrieve Weibo posts related to the health rumor from a large dataset;
• source: source cited in the fact-checking article;
• article_url: url of the article.
• Sample:

rumor_id,publish_date,rumor_type,article_title,article_content,search_expression,source,article_url
80016,2020-01-21,wish,放烟花爆竹可以消毒，预防瘟疫,农业上确实有在大棚中燃烧硫磺杀灭害虫细菌的做法...,(烟花|爆竹)+(消毒|瘟疫),医学博士、副主任医师、中华医学会科普分会青年委员,https://vp.fact.qq.com/article?id=a54f4a260301565af3454048724350f5
81628,2020-02-10,wish,谣言：挂烫机可以杀死衣服上的新型冠状病毒,有消息称，使用挂烫机可以杀死衣服上的新型冠状病毒鉴定结果：谣言权威解读：56°C且持续30分钟是用高温杀死新冠状病毒的两个必备条件，也就是说得对衣服的每一个部位都持续烫30分钟以上才有效。,挂烫+(新冠|病毒),头条辟谣,http://toutiao.com/group/6791398324525072910/
82020,2020-02-12,dread,水果、蔬菜表面会附着新型冠状病毒？,,(水果|蔬菜)+(附着|沾染)+病毒,国务院联防联控机制新闻发布会,http://www.piyao.org.cn/2020-02/12/c_1210469941.htm

data/debunking_posts.csv contains 238,554 debunking posts published on Weibo between January and May 2020. Some of the posts debunked multiple rumors at once.

• Sample:

url,rumor_id,rumor_type
https://weibo.cn/comment/IqDc47lFX,80016,wish
https://weibo.cn/comment/IsG23ChSN,81628,wish
https://weibo.cn/comment/ItrKKoMos,82020,dread

Data Collection

The process of data collection includes:

1. We first crawled 11 popular Chinese fact-checking sites for 5,958 articles published from January 2 to April 8, 2020. The crawler was run between January 29 and April 18.

	Source	Count	Website
1	腾讯较真	563	https://vp.fact.qq.com
2	丁香医生	284	https://ncov.dxy.cn/ncovh5/view/pneumonia_rumors
3	头条辟谣	1551	https://toutiao.com/c/user/62596297771/#
4	微博辟谣	1869	https://weibo.com/1866405545
5	捉谣记	85	https://weibo.com/6590980486
6	阿里健康	56	https://mdeer.taobao.com/healthhome
7	**Yahoo新闻	124	https://tw.news.yahoo.com/topic/2019-nCoV-rumor
8	**互联网联合辟谣	663	https://www.piyao.org.cn/2020yqpy
9	北京辟谣平台	423	https://www.qianlong.com/yiqing2020
10	湖南省辟谣平台（辟谣侠盟）	236	https://moment.rednet.cn/topic/pc/index.html?topicId=62464&siteId=6
11	科普**	104	https://piyao.kepuchina.cn

* sources 1-7 of the fact-checking articles are commercial companies, while sources 8-11 are government-related organizations.

A total of 408 health rumors were manually filtered from these fact-checking articles.

2. We then filtered posts related to these health rumors from a large Weibo dataset based on the following process. Please see the paper for details.

3. Lastly, we use regular expressions (utils/regular_expression.txt) to identify debunking posts that express denials to health rumors and manually filter them for several rounds to ensure precision. A total of 238,554 (201,491 unique) debunking posts were identified.

Citation

If you are using the dataset or the search schemes, please cite the following in your work:

@misc{yang2021know,
      title={Know it to Defeat it: Exploring Health Rumor Characteristics and Debunking Efforts on Chinese Social Media during COVID-19 Crisis}, 
      author={Wenjie Yang and Sitong Wang and Zhenhui Peng and Chuhan Shi and Xiaojuan Ma and Diyi Yang},
      year={2021},
      eprint={2109.12372},
      archivePrefix={arXiv},
      primaryClass={cs.SI}
}