Project Info
Incremental Open-World Reference Resolution
Tom Williams | twilliams@mines.edu
For humans to interact with robots through natural language, they must be able to understand natural language in a way that (1) is incremental, in order to respond efficiently and to allow for backchannel responses; and (2) is open world, in order to enable operation in realistic environments where not all referents are known of a priori. While there has been some limited work on each of these capabilities, this project seeks to enable both within a single system.
More Information
For more information, please consult the publications listed in the “Reference Resolution and Referring Expression Generation” section of the MIRRORLab research page: http://mirrorlab.mines.edu/research.html
Grand Engineering Challenge: Reverse-engineer the brain
Student Preparation
Qualifications
Interested students should either (a) have experience with robot language understanding algorithms, or (b) have taken AI, Prob-Stats, and Algorithms.
Time Commitment
40 hours/month
Skills/Techniques Gained
The student will gain experience with robot language understanding algorithms and integrated robot architectures.
Mentoring Plan
The selected student will meet with Dr. Williams weekly in research group meetings, and will meet with Dr. Williams and his graduate students as-needed.