IEEE BigMM 2020

Grand Challenge

Spot Fake: Multimodal fake news detection

CALL FOR PARTICIPATION

MOTIVATION

The IEEE International Conference on Multimedia Big Data (BigMM) 2020 is hosting a grand data challenge that aims to bring together researchers around the globe to solve a problem that is of great societal need. The increase in proliferation of fake content in the environment is disrupting the lives of human beings in many ways and this calls for an immediate solution. One of the themes of this year data challenge is in the area of detecting misinformation in news articles.

TASK DESCRIPTION

Given a dataset of news articles containing text and multiple images, the task is to perform a fake news detection task leveraging the information obtained from text and multiple images. The aim should be to propose a model that works well on unseen real-time dataset too.

DATASET DETAILS

  • The FakeNewsNet repository will consists of two datasets i.e. Gossipcop and Politifact.

  • The gossipcop contains news samples from the entertainment domain whereas the politifact dataset contains news samples from the political domain.

  • Each news sample consists of title, text (content of the article), top image (the first image that is encountered while reading the article), other images, source and label. The dimensions of the images will be variable.

  • Every news sample has one or more images associated with it. Each dataset has samples belonging to two classes: real (1) and fake (0).

  • The total number of training samples for Gossipcop and Politifact are 15305 and 549 samples respectively.

Note:

  1. Training dataset can be downloaded from here.

  2. Participants need to download images by themselves. Reference code for the same is attached here.

  3. Evaluation for politifact and gossipcop dataset would be done separately.

EVALUATION

A test-set of 3827 samples for gossipcop domain and 138 for politifact domain will be used to evaluate the performance of the model. The evaluation metric to be used are: Accuracy, Precision, Recall and F1 score.

SUBMISSION INSTRUCTIONS

The instructions on how to submit the solution will be released along with the test data.

Note: Top submissions would be invited to submit system description/ model papers describing the proposed methodology, improvements and limitations in the proceedings of IEEE BigMM 2020.

IMPORTANT DATES

Competition Timeframe:

  • April 18, 2020: Competition begins.

  • July 20, 2020: Public release of test data along with demo submission file.

  • July 22, 2020 PST: Deadline to make the submissions.

  • Submission Portal: https://forms.gle/YLNzeavF1jDTCbz57

Conference Preparation

  • July 26, 2020: Top submissions would be invited to open source the code and models for evaluation.

  • July 30, 2020: Invitations sent to top submissions for submitting the system description papers in the proceedings of IEEE BigMM 2020.

  • TBA: Deadline for submitting the system description papers.

Note: SUBMISSION INSTRUCTIONS RELEASED !!!

  • The test file can be found here.

  • The demo solution file format is present here.

  • The submission has to be made here.

Note: The submissions have to done seperately for both the datasets. The submission deadline is 22 July 2020 PST.

ORGANIZING COMMITTEE (PhD Students)

  1. Kai Shu (Arizona State University USA)- kai.shu@asu.edu

  1. Shivangi Singhal (IIIT-Delhi)- shivangis@iiitd.ac.in

ORGANIZING COMMITTEE

  1. Girish Keshav Palshikar (TCS Research and Innovation)- gk.palshikar@tcs.com

  1. Ponnurangam Kumaraguru (IIIT-Delhi)- pk@iiitd.ac.in

  1. Pradeep Atrey (University of Albany USA)- patrey@albany.edu

  1. Rajiv Ratn Shah (IIIT-Delhi)- rajivratn@iiitd.ac.in

  1. Tanmoy Chakraborty (IIIT-Delhi)- tanmoy@iiitd.ac.in

For any query, contact the following: Shivangi Singhal (IIIT-Delhi)- shivangis@iiitd.ac.in