Daumé and Graduate Student Sharaf Receive Amazon Academic Research Award

Hal Daumé III, an associate professor of computer science and a member of UMIACS, and Amr Sharaf, a second-year computer science doctoral student, recently received an Amazon Academic Research Award (AARA) to support a project aimed at improving machine translation systems.

Daumé is director of the Computational Linguistics and Information Processing Laboratory, where Sharaf does much of his work.

Their project, “Neural Machine Translation from Weak User Feedback,” is focused on improving machine translation systems using natural interactions with users.

For example, Sharaf says, a user provides a sentence in the source language and receives a target translation, the user might then provide feedback either in the form of a thumbs up/down or a star rating for the generated translation. The feedback can also be implicit, such as requesting clarification in a translated conversation.

He says their goal is to improve the quality of the translation system using those types of user interactions.

“We'll focus in this project on improving machine translation system training using neural network techniques,” Sharaf says. “I feel honored that I was deemed worthy of the award and I'm really excited by its potential. The Amazon Academic Research Awards will allow us to explore the possibility of designing smarter translation systems that can learn via direct and natural interaction with users.”

Daumé and Sharaf were awarded $49,545 and an additional $15,120 in Amazon Web Services (AWS) Cloud Credits to support their research.

The AARA program funds academic research and related contributions to open source projects by top academic researchers throughout the world.