Improving Learner's Comprehension Using Entailment-Based Question Generation

Abstract

In recent decades, artificial intelligence has made significant contributions to education. With the rise of online learning, most tedious tasks have been transferred from human hands to machines and programs. Both human and machine intelligence require good questioning abilities. To improve learning efficiency, asking questions on the reading material is an effective approach. However, the learners' reading comprehension may diminish if the questions are similar to the original sentence. In this study, a question generation method based on textual entailment is proposed. A question is constructed from the inferred text of the original sentence to help readers improve their reading comprehension. As a result, in order to answer the questions, learners must comprehend textual entailment. The proposed system first generates the entailment for the sentence to which the question should be generated, then generates the question for the generated entailed sentence. Furthermore, the findings are evaluated based on lexical and semantic similarity to ensure that the generated questions are semantically consistent.