MCFEND A Multi-source Benchmark Dataset for Fake News Detection in China

* This dataset supports online use.Click here to jump.

The MCFEND dataset is a multi-source Chinese fake news detection benchmark dataset jointly constructed by Hong Kong Baptist University, the Chinese University of Hong Kong and other institutions.

The prevalence of fake news from various online sources has had a significant impact on the public. Existing Chinese fake news detection datasets are limited to news from Weibo alone. However, fake news from multiple sources exhibits diversity in various aspects, including its content and social context. Methods trained purely on a single news source are difficult to apply to real-world scenarios. The research team's pilot experiments showed that the F1 score of the state-of-the-art method learned from the large Chinese fake news detection dataset Weibo-21 dropped significantly from 0.943 to 0.470 when the test data was changed to multi-source news data, failing to identify more than one-third of multi-source fake news.

To address this limitation, the research team constructed the first multi-source benchmark dataset for Chinese fake news detection, called MCFEND, which consists of news collected by the research team from different sources such as social platforms, messaging applications, and traditional online news media.The dataset collects 23,974 news items from diverse sources such as social platforms, instant messaging applications, and traditional online news media, all of which have been verified by 14 international authoritative fact-checking organizations. The news content covers the period from March 2015 to March 2023, including multimodal information such as text, images, and metadata, as well as social contextual data such as posts, comments, emoticons, and user profiles. MCFEND is a benchmark dataset that aims to advance Chinese fake news detection methods in real-world scenarios.