In this work we explore the usefulness and practicality of domain adaptation and multi-domain learning methods in question-answer generation. Unlike recent work in question-answer generation which focuses on processing single-domain data to create synthetic reading comprehension datasets (Du and Cardie, 2018), we propose a question-answer generation system that can adapt to datasets containing multiple domains while still achieving similar or better performance in single domains compared to a baseline. We apply our system, consisting of an answer extraction system and a question generation system, to the SQuAD and SciQ reading comprehension datasets and evaluate its efficacy in mixed- and single-domain settings. Our domain adaptation method achieves higher performance than baselines on the mixed-domain and SciQ datasets in both answer extraction and question generation.
Title
Multiple Domain Question-Answer Generation
Published
2019-05-15
Full Collection Name
Electrical Engineering & Computer Sciences Technical Reports
Other Identifiers
EECS-2019-44
Type
Text
Extent
11 p
Archive
The Engineering Library
Usage Statement
Researchers may make free and open use of the UC Berkeley Library’s digitized public domain materials. However, some materials in our online collections may be protected by U.S. copyright law (Title 17, U.S.C.). Use or reproduction of materials protected by copyright beyond that allowed by fair use (Title 17, U.S.C. § 107) requires permission from the copyright owners. The use or reproduction of some materials may also be restricted by terms of University of California gift or purchase agreements, privacy and publicity rights, or trademark law. Responsibility for determining rights status and permissibility of any use or reproduction rests exclusively with the researcher. To learn more or make inquiries, please see our permissions policies (https://www.lib.berkeley.edu/about/permissions-policies).
