Domain-Robust Multimodal Misinformation Detection for Newly Emerged Events - RMGS
DescriptionMisinformation is the false information that spreads regardless of whether there is intent to mislead the public, which consists of fake news, rumour, disinformation, etc. As the advent of online social media lacks serious verification of posts and netizens are usually unable to discriminate between fake and real news, misinformation has proliferated in recent years, threading all aspects of individuals and society. There exists several techniques powered by advanced AI technologies to tackle the threat of misinformation. Although existing methods have shown some desired performance to identify misinformation, they require a huge amount of labeled data for AI model training, which is unrealistic in practice because collecting a massive volume of news and posts can be cumbersome and the data are highly relied on past historical events. As such, they will fail to generalize to unseen news events. In addition, increasing multimodal content (i.e., posts with images) make this task even more challenging.In this project, we aim to design a domain robust automatic system to detect multimodal misinformation of newly emerged events, which can help to create a cleaner internet environment and to protect users from misinformation.
|Effective start/end date||1/09/22 → …|