AGent: A Novel Pipeline for Automatically Creating Unanswerable Questions
Introduction
AGent: A Novel Pipeline for Automatically Creating Unanswerable Questions
Son Quoc Tran, Gia-Huy Do, Phong Nguyen-Thuan Do, Matt Kretchmar, Xinya Du
Computer Science Department, Denison University, Granville, Ohio
The UIT NLP Group, Vietnam National University, Ho Chi Minh City
University of Texas at Dallas
AGent
Agent pipeline has three steps:
1. Matching questions with new contexts.
2. Identifying hard unanswerable questions.
3. Filtering out answerable questions.
Citation and Contact
If you found this repository helpful, please cite:
@misc{tran2023agent,
title={AGent: A Novel Pipeline for Automatically Creating Unanswerable Questions},
author={Son Quoc Tran and Gia-Huy Do and Phong Nguyen-Thuan Do and Matt Kretchmar and Xinya Du},
year={2023},
eprint={2309.05103},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
Please contact Son Quoc Tran at sqt2@cornell.edu
if you have any questions.