MIT EECS - Fano Undergraduate Research and Innovation Scholar
Utilizing Story Understanding to Improve Automated Planning
Patrick H. Winston
The vision for this project is to create an automated planning algorithm that mimics the way humans make plans by adding functionality to the Genesis program. The focus of Genesis is the Strong Story Hypothesis, which states that human intelligence is defined by the ability to gain information from stories. This project is also based on that hypothesis and seeks to augment the traditional approach to planning with Genesis ability to draw conclusions and generate insights from a set of stories. This extra information allows for the synthesis of alterative options, which helps populate the solution space and can be used to improve search heuristics. This would allow the algorithm to operate even when relatively little information is provided.
I spent the last year and a half working with Professor Ceder in the material science department writing Monte Carlo based algorithms to virtually perturb crystal structures and compare their X-ray diffraction patterns. I spent this past summer working with Oculus360, a big data/machine learning company, and had the chance to work with topic identification, feature extraction, and trend analysis algorithms and also wrote some natural language processing programs.