Girju is One Researcher Who Thrives on Challenges

Beckman Institute researcher Roxana Girju was the perfect faculty member to lead a group of University of Illinois students to victory in an international competition held earlier this year.

Beckman Institute researcher Roxana Girju was the perfect faculty member to lead a group of University of Illinois students to victory in an international competition held earlier this year.

Girju is an Assistant Professor of Linguistics and her research area of computational linguistics gave her the knowledge to lead the team, the first-ever from Illinois taking part in a competition sponsored by the Association for Computational Linguistics. Perhaps just as important were two other factors.

"Yes, I think I like to be challenged - not only by other people, but by the task itself and see how far I can get."
- Roxana Girju

The diverse team of students (only one of whom was her student) came not only from Linguistics, but also represented departments such as Computer Science, Electrical and Computer Engineering, and Library and Information Science. Girju's background includes degrees in economics and computer science, and she has several interdisciplinary collaborations at Beckman.

Then there was the competitive aspect of the academic contest. Competition is something Girju seems to thrive on, whether the arena is sports or education. She was the top gymnast in her high school class growing up in gymnastics-obsessed Romania, and believes academic competition serves students well.

"It was very important for us to compete, first of all, and also to win," Girju said. "It was good because it put the university in a good spotlight, but also for the students who participated for the first time in an international competition. It really gave them more confidence in themselves and most of them want to continue in this direction.

"And what is (also) nice is the fact that they belong to different departments, which shows how interdisciplinary the task was." In addition to her multiple degrees and continually evolving academic interests, Girju speaks five languages, and studies nutrition and military history. As if those challenges weren't enough, she also takes part in several sports, including practicing gymnastics.

"Yes, I think I like to be challenged - not only by other people, but by the task itself and see how far I can get," Girju said. With faculty appointments at Beckman and at Linguistics and a Center for Advanced Study Award in 2006, Girju has gotten far at the University of Illinois. Her research focus on theoretical and computational models of text comprehension features a particular emphasis on semantic models for textual inference. It's an interest that helped at the international competition called the Semantic Evaluation Series (SemEval 2007).

The students built a "semantic parser" that worked as a classifier to identify the underlying meaning encoded by two nouns in a sentence. Girju said the ability to recognize and classify semantic relations between words in a sentence is an important building block for creating applications such as search engines. One of Girju's earliest interests in computational modeling was not in linguistics, but in finance.

"I've always been interested in building a system which would predict the stock market," Girju said. "Then I came here about seven to eight years ago for my graduate studies and got interested in computational linguistics. I think it worked because I've always been involved in interesting projects, but I've always been interested in natural languages in general."

Now Girju is involved in interdisciplinary collaborations with other Beckman researchers like Richard Sproat and Narendra Ahuja. She is a firm believer in the interdisciplinary approach.

"Having training in different disciplines was very beneficial; it helped me get a broader view of the topic and make connections much easier," Girju said. "Despite the progress that has been done in the last ten years, all these different disciplines are still isolated. Considerable progress can be done only if we look at the problem from different angles. And my opinion is that these angles are different perspectives from various disciplines. Research communities/ groups have to start building stronger collaboration bridges and I strongly believe that Beckman is the right place for this."

This article is part of the Summer 2007 Synergy Issue, a publication of the Communications Office of the Beckman Institute.