Mohammad Raihanul Islam, DAC Ph.D. student in the Department of Computer Science

Graphic is from Islam’s paper on “RumorSleuth: joint detection of rumor veracity and user stance”

Classifying rumors and fake news in social media is the focus of Mohammad Raihanul Islam’s work at the Discovery Analytics Center.

“A rumor generally refers to an interesting piece of information — widely disseminated through a social network — that is not easy to substantiate,” said Islam, a Ph.D. student in computer science.

Later, it can turn out to be true, false, or remain unverified.

“The threat of rumors and fake news is very real and identification is crucial because rumors and fake news can lead to deleterious effects on users and society,” he said. “For example, spreading unverified malicious content could cause severe economic downfalls within a short period of time.”

The objective of his research, he said, is to develop a range of machine learning methods to effectively detect and characterize rumor veracity in social media.

DAC’s emphasis on applied machine learning, especially in social network analysis, is what attracted Islam to the center. Advised by Naren Ramakrishnan, he is on track to graduate this spring.

In the first part of his Ph.D. thesis, Islam worked on creating rich representation for users that can be helpful in rumor classification. He then applied this representation to classify which conversation is talking about a fake news/rumor.

Now, in the final stages of his research, Islam is focusing on creating a generative model for rumor classification using state-of-the-art deep learning models.

“At DAC I have enjoyed working on applied machine learning problems of my choosing and the opportunity of collaborating with fellow graduate students,” Islam said.

He is first author on four major publications: Inferring Multi-Dimensional Ideal Points for US Supreme Court Justices, 2016 Association for the Advancement of Artificial Intelligence (AAAI) conference; DeepDiffuse: Predicting the ‘Who’ and ‘When’ in Cascades, 2018 IEEE International Conference on Data Mining (ICDM); RumorSleuth: joint detection of rumor veracity and user stance, 2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM); and NActSeer: Predicting User Actions in Social Network using Graph Augmented Neural Network, 2019 ACM International Conference on Information and Knowledge Management (CIKM).

Islam personally presented the first two papers; the other two were presented by collaborators.