• The Automatic Generation of Contextual Questions and Answers for English Learners

    Sungjin Ahn, Sung Won Han, Minji Kang (see profile) , Hoyeol Kim, Donghwan Kim
    ACH 2021
    Digital humanities, Machine learning, Teaching
    Item Type:
    Conference poster
    Conf. Title:
    ACH 2021
    Conf. Org.:
    The Association for Computers and the Humanities
    Conf. Loc.:
    Conf. Date:
    July 21-23, 2021
    BERT, deep learning, GPT-2, T5, transformer, Pedagogy
    Permanent URL:
    Understanding context is essential for ESL (English as a Second Language) students to become skilled in English. While there is an abundance of extant contextual questions, they are not tailored to ESL teachers’ course objectives and reading materials. For this reason, ESL teachers must continuously create their own contextual questions. The NLP question and answer generation tasks can lift ESL teachers’ workload by creating MCQs (Multiple Choice Questions), T/F (True or False) questions, and fill-in-the-blank questions, along with answers. We deployed a model which automatically generates MC and Wh- questions with answers. We display several examples and explain the process for generating MC and Wh- questions and answers. For our research methods, we first performed text preprocessing with the CoNLL-2014 and BEA-2019 datasets, which consist of essays written by native and non-native English students. After that, we deployed GPT-2, BERT, and T5 in order to complete the question and answer generation task. The contextual question and answer generation model will contribute specifically to ESL teachers who manually create MC and Wh- questions for ESL students, as well as to the fields of education, digital humanities, and computer science. In addition, we share tutorials for this task with the public so that anyone can make use of our research.
    Last Updated:
    3 years ago
    All Rights Reserved


    Item Name: pdf ach2021.pdf
      Download View in browser
    Activity: Downloads: 146