CLIP Lab Summer Interns Jump into Machine Learning
From the advent of self-driving vehicles to personal digital assistants that provide countless morsels of information and perform multiple tasks, the use and impact of machine learning and artificial intelligence (AI) in our daily lives is rapidly expanding.
Looking to get on board this scientific juggernaut, a group of four visiting undergraduates is spending 10 weeks this summer exploring new ideas in machine learning through an internship in the Computational Linguistics and Information Processing (CLIP) Laboratory.
The students, supported by a National Science Foundation Research Experiences for Undergraduates (REU) grant, are developing an interactive system that will act like a “virtual librarian.” The project is similar to the real-life scenario where a person goes to the library to find a novel, but only remembers some of the plot vaguely. The librarian would have to ask the individual a series of questions to find the correct book.
Hal Daumé III, an associate professor of computer science and director of the CLIP lab, says the key challenge the interns are facing is getting their virtual librarian to ask a human user enough questions.
Daumé is advising the students this summer with assistance from Shi Feng, a second-year computer science doctoral student.
“We see similar interactions between humans and machines in other scenarios,” Feng says of the interns’ project. “For example, query reformulation when using search engines—people modify their queries according to the feedback from the machine. We want to make this type of feedback clearer and help improve this type of interaction.”
He adds the interns are “learning very fast and making good progress.”
Carl Denton (on left in photo), a rising junior at Harvard University, says that while the research is just getting started, he is looking forward to delving into his group’s assigned task.
“We’re planning on working with books—and I think typically in these sorts of applications it can be difficult to work with book-like text at such a high volume, so it will be interesting to explore that area and see what we can do,” says Denton, a math and computer science double major.
He notes he was attracted to the internship because of how natural language processing (NLP) is “really exploding at the moment,” as well as Daumé’s reputation in the field.
“I had heard of Hal before, which is the main reason I applied to the program,” Denton says. “He’s fairly well-known for his NLP research, which I think is really interesting and impressive.”
Daumé and other researchers in CLIP have been using machine learning for years on projects involving natural language processing, probabilistic models, and data management.
Isabella Huang (second from left in photo), a rising senior at Williams College who is majoring in math, says she sees the internship as an opportunity to get hands-experience in machine learning and AI.
“I think I come from a more philosophical side, because I do see myself as more of a cognitive scientist,” she says. “So, I was mainly interested in doing this research this summer to get the practical applications side of it.”
Trista Cao (on right in photo), a rising senior at Bryn Mawr College who is double majoring in math and computer science, says she thinks the hands-on experience this summer will help the entire group as they each continue their own studies this fall.
“I don’t think any of us have been exposed to this field a lot and with this internship, maybe we’ll see how many algorithms and theories we run across here and how we can apply them to our own work and perhaps research we’ll do in the future,” she says.
Makana Castillo-Martin (second from right in photo), a senior at Reed College majoring in math, says she is enjoying working in the CLIP lab.
“I think there’s a lot of good work we’re doing here, and I’m also impressed with all of the research that goes on in the lab,” she says. “I find everything we’ve been working on so far to be interesting because I have never really worked on applied computer science before in terms of hard coding because of my major.”
Daumé says he is very impressed with the group.
“They're awesome,” he says. “Super intelligent, friendly and hard-working.”
At the end of the 10 weeks, Daumé says he hopes the students will have found their time in the CLIP lab useful.
“I hope they'll get even more excited about natural language processing and machine learning, learn some state-of-the-art deep reinforcement learning techniques, and get excited about research in general,” he says.
***
The Computational Linguistics and Information Processing Laboratory is one of 16 centers and labs in the University of Maryland Institute for Advanced Computer Studies (UMIACS).
The diverse group of faculty and students in the lab includes experts in computer science, linguistics, information studies, business, and language science.
—Story by Melissa Brachfeld