Abstract
Automatic question generation (AQG) has many diverse applications in educational contexts. To bring these benefits to as many students as possible, it is prudent to expand AQG capabilities in as many languages as possible. However, English remains the dominant language in AQG research, and the required natural language processing tools for other languages are often under-resourced relative to English, which can make developing AQG pipelines difficult or impractical altogether. An approach called parallel construction has been developed to leverage existing English AQG systems for AQG in other languages. The benefits of this parallel construction approach are described, and examples of questions generated from Spanish and Brazilian Portuguese textbooks using the parallel construction method are presented and discussed.
