Skip to main content

Graduate Programs

Student Profiles

A photo of the student.

Linfeng Song

Hometown: Yingkou, China
Degree: PhD, 5th year
Area: Graph problems in Natural Language Processing
Advisor: Dan Gildea

Why did you pick UR for your graduate work?

UR has had a very good worldwide reputation, it's a top university in the U.S. I knew the professors working on natural language processing, such as James Allen, Len Schubert, and Dan Gildea, before applying here. I think that working with them is a great honor to me.

URCS is a great place to do a PhD because of the close-knit community. Each PhD student and faculty advisor are connected to one another and it is a great benefit to one’s studies and research. Also, the department’s commitment to building our foundation as computer scientists is highly valuable. Another advantage is the in-depth courses of this department. My academic work can benefit from both the AI courses, such as machine learning, and natural language processing, and from those that are not in my direct area of research. I believe my academic life can go much further with the solid foundation built upon these courses.

How did you choose the area of research you have studied? Did you know what you wanted to study before you came? did that change once you were here, and if it did, why?

Before joining the University of Rochester, I'd already started a master program in natural language processing. I was so excited about this area of research. I am so lucky to work with Professor Dan Gildea. I never needed to change my research area.

Human beings use sentences to communicate, while computers can better understand structural data, such as tables and graphs. In general, a natural language understanding (NLU) system maps input sentences into graph structures. One line of my research is on how map between these representations, including the problem of how an NLU system can effectively consume the graph structures. Since these structures do not fit neatly into the modern NLP system architectures, how to encode them is an important topic. Another line of my research is on how to improve modern question answering (QA) systems.

What have you enjoyed most about your work here at Rochester?

There are actually a couple of things, and it's very hard to pick one. If I have to pick one, it's the fundamental education that I have had. The education is very comprehensive, which includes not only the course work and the research but also how to be a good and successful person. I learned these things from my advisor, Dan Gildea, other faculty members, and other graduate students too. They set really good examples for me.

What are your future plans? Do you plan to go into academia or industry?

I'll be a senior research scientist in an industrial lab, so it's kind of in between. I'm very excited about doing research, and hopefully, I can get an academic job in my future.