Question Generation via Overgenerating Transformations and Ranking
Abstract
We describe an extensible approach to generating questions for the purpose of reading comprehension assessment and practice. Our framework for question generation composes general-purpose rules to transform declarative sentences into questions, is modular in that existing NLP tools can be leveraged, and includes a statistical component for scoring questions based on features of the input, output, and transformations performed. In an evaluation in which humans rated questions according to several criteria, we found that our implementation achieves 43.3% acceptability for the top 10 ranked questions, and generates approximately 6.8 acceptable questions per 250 words of source text.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2009
- Accession Number
- ADA531042
Entities
People
- Michael Heilman
- Noah A. Smith
Organizations
- Carnegie Mellon University