CRA Awards Recognize Four Illinois CS Undergrads For Research


Illinois CS students Jonah Casebeer and Ben Zhou were finalists, while Harshay Shah and Edward Huang were recognized with honorable mentions from the CRA.

Written by

Two Illinois Computer Science students were finalists for the Computing Research Association's 2019 Outstanding Undergraduate Researcher Award and two more earned honorable mentions, one of the department’s best-ever performances in the competition.

Illinois CS students Jonah Casebeer and Ben Zhou were finalists, while Harshay Shah and Edward Huang were recognized with honorable mentions from the CRA, the association announced in late December.

The Undergraduate Researcher Award program recognizes undergraduate students in North American colleges and universities who show outstanding research potential in an area of computing research. A committee of researchers from eight universities selected four winners and four runners up, as well as the finalists and honorable mentions.

All four Illinois CS entrants were honored, the strongest performance by the department since 2015, when Urvashi Khandelwal was among the winners. In addition, Assistant Professor Edgar Solomonik (BS CS ’10) was a finalist in 2010.

More about this year’s Illinois CS undergraduates recognized by the CRA:


Associate Professor Paris Smaragdis, who nominated Jonah Casebeer, said the student has worked on a series of challenging research projects in the year-plus they have worked together, some involving problems that stump even very strong graduate students.

Jonah Casebeer
Jonah Casebeer

Smaragdis cited one problem in particular, an audio classification system in which he wanted Casebeer to replace standard front-end processing with a neural net.

“Two weeks later he comes with a working implementation; within a couple of months he had almost matched the state-of-the-art results,” Smaragdis wrote in his letter nominating Casebeer. “Mind you, that state-of-the art came from groups with multiple researchers working on this for years; in fact many such groups had systems that performed worse than what Jonah did.”

Casebeer, who plans to graduate in 2019, has five published peer-reviewed papers and more in the works.


Professor Dan Roth said Ben Zhou was a rarity when he joined Roth’s Cognitive Computation Group: a sophomore.

Ben Zhou
Ben Zhou

I would rarely take such a young student to my group,” Roth wrote in his nomination letter for Zhou. “(But) I view Ben as a PhD student already, and so do my other students. He is considered the authority in my group on a number of important topics such as context-sensitive word embeddings (Elmo embeddings), dataless classification, and relation extraction.”

Roth, who is now the Eduardo D. Glandt Distinguished Professor at the University of Pennsylvania but remains an adjunct professor at Illinois CS, praised Zhou in particular for his work on a paper for the 2018 Conference on Empirical Methods in Natural Language Processing.

Roth suggested Zhou work on incidental supervision -- training machine learning tasks with minimal, indirect, and weak supervision signals. To do so, Zhou studied it in the context of recognizing fine semantic types of entities and concepts, that problem of identifying through context the semantic types an entity can take – that the University of Illinois, for example, is in coarse type an organization and in fine type an educational institution, and that Illinois can be a river in one context and a tennis team in another.

“Ben’s work is the first work on semantic typing that has the flexibility to transfer across text genres and domains, and to generalize to new type taxonomies,” Roth wrote. “It requires no annotated data and can flexibly identify newly defined types.”

Zhou plans to graduate in 2019 and has authored several other papers, too.


Associate Professor Hari Sundaram began working with Harshay Shah in 2016, and said two questions drove the research that Sundaram cited in nominating Shah: “Why do social networks possess the structure that we observe? How do individuals with limited information help create them?”

Harshay Shah
Harshay Shah

“We know that people tend to make friends with those who are popular, but also to those who are like themselves. We know of mechanisms that explain each of these two phenomena separately, but not jointly,” Sundaram said. “Prior scholarship tends to ignore that fact that individuals have limited information when making these decisions. A decision-making framework by individuals acting with limited information that accounts for both phenomena simultaneously would connect the macroscopic properties that we observe in real-world networks, with local decisions by individuals.”

Harshay proposed an Attributed Random Walk (ARW) model, Sundaram said, and was able to explain macroscopic network structure while accounting for multiple sociological phenomena.

“Harshay has been instrumental to the development ARW,” Sundaram wrote. “We’ve started to make extensions to the model and Harshay is now mentoring an undergrad to work alongside him. In summary—Harshay has all the makings of a star graduate student.”

Harshay plans to graduate in 2019.


Edward Huang was nominated by Nigel Bosch, a postdoctoral researcher at the National Center for Supercomputing Applications, for his work on two projects.

Edward Huang
Edward Huang

The first examined students’ improvement in an online STEM class, focused on how much students from underrepresented groups improved over the course of the class. Huang impressed with both his programming ability as he dealt with data that originated in a learning management system from the early 1990s and in his careful treatment of the data, according to Bosch.

The second project was also focused on online classes, trying to automatically detect metacognitive language in text, then measure the amount of such language (phrases such as “I think the answer is” or “I am actually not really sure”) that exists and the degree of confidence these phrases reflect.

“Eddie’s method breaks down each piece of text into parts of phrases and filters them based on individual components of metacognitive phrases (that he identified), such as the negations and adjectives found in these phrases,” Bosch wrote in nominating Huang. “Without Eddie’s work, examining the metacognitive language underrepresented students utilize would still be nothing more than an idea.”

Huang plans to graduate in 2019.


Share this story

This story was published January 10, 2019.