Ellie is a 4th year Ph.D. student at the University of Pennsylvania, advised by Chris Callison-Burch. She received her B.A. in economics from Johns Hopkins University and her B.M. in saxophone performance from the Peabody Conservatory in 2012. Ellie’s research aims at teaching computers to understand human language.
As humans, the majority of our social experience, especially online, is played out in the form of text or speech. Without the ability to comprehend what humans say and write, computers can access only a tiny fraction of this experience. Automatic natural language understanding can allow computers to organize the vast amount of information and to communicate with the people who want to access it.
Although humans do it without thinking, understanding natural language is very difficult for computers. Much of this difficulty is due to the fact that language exhibits both ambiguity‚– we can use the same words to express many different meanings‚– and redundancy‚– we have many different ways of expressing the same meaning. The focus of Ellie’s dissertation is enabling computers to understand paraphrases through data-driven methods. Her work allows machines to recognize when two phrases mean the same thing (e.g. “the head of the company” means “the company’s leader”) and when they do not (e.g. “the head of the femur” does not mean “the femur’s leader”). When phrases can be treated as paraphrases, Ellie’s work focuses on what information is gained or lost by using one phrase over another. For example, we might paraphrase a general word with something more specific (“leader” -> “CEO”), or we might paraphrase a formal word with something more casual (“leader” -> “top guy”). Ellie is currently working on models of how linguistic context and composition effect the meaning of a paraphrase, and on applying those models to help computers reason better about human language.