Natural Language Understanding for Application-Specific Domains

Sean Trott

ICSI

Tuesday, April 28, 2015
12:30 p.m., Conference Room 5A

This talk will focus on a modular system we have developed and implemented for natural language understanding in domains of moderate scale. Most of the work has been done in the domain of robotics, but there has also been work on metaphor analysis. The system uses Embodied Construction Grammar and a best-fit Analyzer (Bryant 2004) for parsing input.This analysis is then used to fuel further action, such as a robot manipulating a simulated world.

Bio:

Sean Trott graduated from UC Berkeley in 2014 with a B.A. in Cognitive Science. He focused on cognitive linguistics and computational modeling. Now, he works as a Research Assistant at the International Computer Science Institute in Berkeley, doing research on Embodied Construction Grammar and various application domains for language understanding, such as robotics.