DAC Ph.D. student, Elaheh Raisi

Elaheh Raisi’s enthusiasm for math dates back to high school. So it was not surprising when Raisi chose applied mathematics as her major at the Amirkabir University of Technology -Tehran Polytechnic.

“I realized early on that mathematics is essential for many practical sciences,” said Raisi. “My aim was to gain a strong knowledge of mathematics that I could use in problem solving.”

During her freshman year Raisi concentrated on mathematics and programming-related courses but after taking some computer science classes, she developed an interest in artificial intelligence. She earned a master’s degree in artificial intelligence at the Science and Research branch of the Islamic Azad University.

Raisi chose to pursue a Ph.D. in computer science at Virginia Tech because of “the abundance of advanced research resources and facilities in the field of data mining, a supporting environment at a prestigious university, and outstanding professors.” Now, a fifth-year Ph.D. candidate, Raisi is advised by Professor Bert Huang and works on cyberbullying detection on social media in his Machine Learning Laboratory.

To address the computational challenges associated with designing automated, data-driven machine learning approaches for harassment-based cyberbullying detection Raisi and Huang have developed a weakly supervised framework, which is specialized for cyberbullying detection,” she said.

The framework consists of two learning algorithms to improve predictive performance by taking into account not only language, but also social structures. One learner identifies bullying incidents by examining the language content in the message; another learner considers social structure to discover bullying. Intuitively, each learner is using different body of information. The learning algorithm tries to make them eventually agree whether social interactions are bullying.

“Our research is geared toward a very important topic in any online automated harassment detection: fairness against particular targeted groups including race, gender, religion, and sexual orientations,” said Raisi. “Our goal is to decrease the sensitivity of models to language describing particular social groups.”

For their research, Raisi and Huang won a best paper award at the IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM) 2017.

They also won a best paper award at the Learning with Limited Labeled data (LLD) workshop at NIPS, 2017 for including deep learning methodologies (word and node embedding) into their framework.