MIT EECS | CS+HASS Undergraduate Research and Innovation Scholar
High-Precision Question Answering Using Wikipedia Definitions
- Natural Language and Speech Processing
I will be working in the Infolab group to expand the capabilities of a tool called WhoAmI. The Infolab group uses WhoAmI as one tool in a system for answering questions in a highly precise way, at the live website start.csail.mit.edu. WhoAmI processes simple sentences such as the first sentence in a Wikipedia article (e.g, Bernhard Riemann was a German mathematician who X.”) and related questions (e.g, “What German mathematicians Y?”). I plan to more deeply integrate WhoAmI with the question-answering system by making it produce structured expressions rather than plaintext. I will then further investigate improving our question-answerer by studying and covering more complex sentences, or the expressions which WhoAmI generates, so that WhoAmI could be of even greater use.
“Having worked on Omnibase last year as a UROP student, I am very interested in studying more parts of the START question-answering system. I think it is important to apply my knowledge to a real, longer-term research project. By participating in SuperUROP, I hope to further my knowledge in natural language processing and also learn how to work on a successful research project. I am most excited to see natural, interesting questions answered well by our system.”