In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
will defend his proposal
Improving False Information Detection on Social Media
With social media growing dominant, false information, such as questionable claims and reports, diffuses fast. False information on social media can be debunked by examining linguistic features and external evidence. Despite effective performance achieved on false information detection, existing works require assumptions that not favor real application. For example, these detection methods demand a large amount of annotated data and related evidence, underestimating the difficulty of evidence linking and the cost of manual annotation. Therefore, we suggest two realistic settings that relax manual efforts: (i) claims are paired with evidence but only some true claims are labeled and more claims remain unlabeled; (ii) we know the truthfulness of claims but the evidence is mixed with similar but unrelated documents. We propose deep learning based methods to detect false information for each setting, respectively. We further propose to investigate paragraph-level linguistic features for questionable news reports when evidence is not available. Our experiments show that the developed models: (i) effectively detect satirical fake news and reveal what linguistic features are important at the paragraph level; (ii) effectively detect false claims when no labeled false claim is presented for training; (iii) require further research to pick the right evidence for false information detection.
Date: Thursday, April 4, 2019
Time: 1:30 - 3:00 PM
Place: PGH 501D
Advisors: Dr. Arjun Mukherjee
Faculty, students, and the general public are invited.