Scholarlink

<p>Automatic Question Generation (AQG) is a transformative tool that significantly enhances educational practices by enabling the creation of adaptive assessments, personalized learning experiences, and continuous feedback systems, while also enriching multiple NLP tasks such as training conversational agents, improving language model robustness, and facilitating dynamic content generation yet its application to the Arabic language remains largely uncharted due to inherent linguistic complexities and limited dataset availability. Where the education process has evolved, but there is still room for improvement in areas like question generation and question answering. This paper presents a new approach using the state-of-the-art transformer model for automated question-generating tasks in Arabic. Meanwhile, our model is not limited to just one subject matter; rather, it is made to generate various types of questions in various fields of study, increasing its effect and value. Utilizing a newly compiled annotated corpus, the model generates questions based on paragraphs and answers using two important datasets, SQuAD and MS-MARCO. The model’s efficacy and potential to improve Arabic language natural language processing (NLP) applications and educational processes are validated through a comprehensive evaluation using metrics like BLEU-4, METEOR, and F1-score. The model achieved 25.72, 45.11, and 51.90 on the MS-MARCO dataset and 23.42, 44.08, and 51.32 on the SQuAD dataset.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automated question generation for Arabic language

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Rahaf A. Alhazaymeh,
  • Mostafa Z. Ali

摘要

Automatic Question Generation (AQG) is a transformative tool that significantly enhances educational practices by enabling the creation of adaptive assessments, personalized learning experiences, and continuous feedback systems, while also enriching multiple NLP tasks such as training conversational agents, improving language model robustness, and facilitating dynamic content generation yet its application to the Arabic language remains largely uncharted due to inherent linguistic complexities and limited dataset availability. Where the education process has evolved, but there is still room for improvement in areas like question generation and question answering. This paper presents a new approach using the state-of-the-art transformer model for automated question-generating tasks in Arabic. Meanwhile, our model is not limited to just one subject matter; rather, it is made to generate various types of questions in various fields of study, increasing its effect and value. Utilizing a newly compiled annotated corpus, the model generates questions based on paragraphs and answers using two important datasets, SQuAD and MS-MARCO. The model’s efficacy and potential to improve Arabic language natural language processing (NLP) applications and educational processes are validated through a comprehensive evaluation using metrics like BLEU-4, METEOR, and F1-score. The model achieved 25.72, 45.11, and 51.90 on the MS-MARCO dataset and 23.42, 44.08, and 51.32 on the SQuAD dataset.