Automatic question generation: a review of methodologies, datasets, evaluation metrics, and applications

Abstract

Question generation in natural language has a wide variety of applications. It can be a helpful tool for chatbots for generating interesting questions as also for automating the process of question generation from a piece of text. Most modern-day systems, which are conversational, require question generation ability for identifying the user’s needs and serving customers better. Generating questions in natural language is now, a more evolved task, which also includes generating questions for an image or video. In this review, we provide an overview of the research progress in automatic question generation. We also present a comprehensive literature review covering the classification of Question Generation systems by categorizing them into three broad use-cases, namely standalone question generation, visual question generation, and conversational question generation. We next discuss the datasets available for the same for each use-case. We further direct this review towards applications of question generation and discuss the challenges in this field of research.

Harnessing the Power of Prompt-based Techniques for Generating School-Level Questions using Large Language Models

“…According to Cao and Wang [3], these corpora are limited to the generation of simple fact-based questions. Furthermore, as stated in [20,21], the majority of these QA datasets are borrowed or crowd-sourced from open-source platforms such as Wikipedia articles, and the questions generally do not incorporate multiple sentences as their basis. There is a notable QG dataset for educational purposes called LearningQ [4], which utilizes complete articles or videos as contexts, resulting in a substantial portion of sentences within the contexts being irrelevant to the specific target question.…”

Opinerium: Subjective Question Generation Using Large Language Models

“…In early studies, two primary taxonomies based on the abstraction of answers [6], [7] were introduced. Recently, a new schema was proposed based on the form of possible answers into four categories factual, multiple sentences spanning, yes/no, and deep understanding [8]. Similarly, we divide questions into two groups objective questions where the answer is retrievable from the given text regardless of answer types, and subjective questions where a subjective answer is provided by individuals and the text only provides a particular topic to be questioned.…”

How Teachers Can Use Large Language Models and Bloom’s Taxonomy to Create Educational Quizzes

“…The goal is to generate natural-language questions that are useful and fluent. Many approaches also attempt to generate the corresponding answers, or use the answer to generate the question (Kurdi et al 2020;Mulla and Gharpure 2023). Due to their recent success in NLP, recent QG research has been dominated by the use of Transformer-based large language models (LLMs) (Kurdi et al 2020;Liu et al 2023).…”

“…These LLMs are deep learning models trained on massive corpora of data to improve their generative performance . The reason for applying this approach in QG research is in large part due to its significant performance improvements over earlier rule-based and other types of systems (Kurdi et al 2020;Steuer et al 2021;Mulla and Gharpure 2023).…”

“…Aligning with the common LLM training objective of the next-token prediction, the emerging paradigm for QG is to feed a textual input, called a prompt, to an LLM for the model to complete (Mulla and Gharpure 2023). Designing this prompt to generate a desired output can be a difficult task, which has resulted in a new research direction called prompt engineering.…”