Update 2019-10-16

Congratulations on completing Phase I! The organizers decided that top 52 teams of track 1 and all teams made submissions of track 2 and 3 can enter the Phase II. Please download the test dataset, 'task1_test_stage.csv', 'task2_test_stage.csv' and 'task3_test_stage.csv' for Phase II. 

Please note that all submissions to Phase II before 0:00 utc, Oct 16 is invalid and will be deleted. The Test Dataset of the Phase I cannot be used in Phase II. All teams are required to work independently.

Introduction

False information online is threatening the security of the Internet, and is growing rapidly in terms of scale, speed of transmission, and fraud types. In 2018 and 2019, the journal Science pointed out that during the 2016 US presidential election, each voter was exposed to four pieces of fake news per day on average; and the transmission rate of false news was 6-20 times faster than true news. Researches show that false news on the Internet has even affected the results of the Brexit vote and the 2016 US presidential election. The DeepFake (changing faces in videos and images) and DeepNude (automatically generating nude photos) technology that came out at the end of 2018 brought panic to the governments. International consulting firm Gartner predicts that by 2020, false news on the Internet will be in full swing, and the artificial intelligence based false news generation technology will far exceed the ability of false detection.

The false news detection competition was jointly organized by the Institute of Computing Technology, Chinese Academy of Sciences and the Beijing Artificial Intelligence Research Institute (BAAI). It aims to promote the development of false news detection technology on the Internet and create a clear and clean online space.

Task

The competition consists of three tasks.

Task 1. False News Text Detection: Text is the main carrier of news information, and the study of news texts helps to effectively identify false news. This task provides text of various pieces of news events, and asks the participants to determine whether the text describes a piece of true or false news.

Task 2. False News Image Detection (url): False news images refer to the pictures in false news. These pictures in false news often contain rich information that helps to judge whether the news is false. This task provides images and asks the participants to determine whether it’s false or true (i.e. from false or true news). 

Task 3. False News Multi-Modal Detection (url): With the development of multimedia technology, social media news usually contains multi-modal information such as text and pictures. Different modals enhance and complement each other. Therefore, the organizers set the third task, which provides multi-modal content of news, including text, images and user profiles, etc., and asks participants to judge whether the news is false or true.