Scholarlink

<p>Integrating Natural Language Processing (NLP) techniques in education has significantly increased the productivity of the learning and teaching approaches by the students and the teachers, respectively. Current Automatic Question Generation (AQG) models tend to produce questions that remain mainly at the surface level. These models lack systematic compatibility with Revised Bloom’s Taxonomy, which classifies different cognitive knowledge levels. This research involves building a T5 or Text-to-Text Transfer Transformer model that has been fine-tuned to generate meaningful questions based on Revised Bloom’s Taxonomy. The proposed system uses a fine-tuned T5 Model, which is optimized to retrieve the text material from various teaching materials and generate appropriate questions. The model is trained using a customized dataset of action verbs derived from Revised Bloom’s Taxonomy, ensuring that the produced questions align with different cognitive levels. Tokenization process with truncation and padding, helps create flexible and varied questions while retaining the original context. The modified T5 model has attained superior performance and generates exceptional questions that align with the cognitive tiers of Bloom’s Taxonomy. Learning materials constitute the textual basis for the system to generate questions that enhance educational tasks for students and educators. Educational institutions can benefit from this technique by accelerating content-based question generation processes, hence enhancing their assessment efficiency and knowledge evaluation capacities. The proposed method technique employs a unique methodology that links AQG models to Revised Bloom’s Taxonomy, in contrast to traditional models that mainly focus on factual recall. This improves educational effectiveness and learning results by optimizing content-driven question production. The exceptional aspect of the proposed model is that it uses Revised Bloom Taxonomy action verbs, thereby enabling it to generate the type of questions that cover a range of the different levels of cognition as opposed to emphasizing more on factual recall. The practical performance of the model demonstrates its effectiveness, with an average semantic similarity score of 88.96%. It also achieves high scores in ROUGE and BLEU, indicating that the generated questions are both contextually relevant and fluent.</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 with Fine-Tuned NLP Model Based on Revised Bloom’s Taxonomy

  • 期刊论文

  • 发表日期：

    2025年10月17日

查看全文
  • R. R. Timosh,
  • Emmanuel Joy,
  • Suganthi Evangeline,
  • Jenice Aroma

摘要

Integrating Natural Language Processing (NLP) techniques in education has significantly increased the productivity of the learning and teaching approaches by the students and the teachers, respectively. Current Automatic Question Generation (AQG) models tend to produce questions that remain mainly at the surface level. These models lack systematic compatibility with Revised Bloom’s Taxonomy, which classifies different cognitive knowledge levels. This research involves building a T5 or Text-to-Text Transfer Transformer model that has been fine-tuned to generate meaningful questions based on Revised Bloom’s Taxonomy. The proposed system uses a fine-tuned T5 Model, which is optimized to retrieve the text material from various teaching materials and generate appropriate questions. The model is trained using a customized dataset of action verbs derived from Revised Bloom’s Taxonomy, ensuring that the produced questions align with different cognitive levels. Tokenization process with truncation and padding, helps create flexible and varied questions while retaining the original context. The modified T5 model has attained superior performance and generates exceptional questions that align with the cognitive tiers of Bloom’s Taxonomy. Learning materials constitute the textual basis for the system to generate questions that enhance educational tasks for students and educators. Educational institutions can benefit from this technique by accelerating content-based question generation processes, hence enhancing their assessment efficiency and knowledge evaluation capacities. The proposed method technique employs a unique methodology that links AQG models to Revised Bloom’s Taxonomy, in contrast to traditional models that mainly focus on factual recall. This improves educational effectiveness and learning results by optimizing content-driven question production. The exceptional aspect of the proposed model is that it uses Revised Bloom Taxonomy action verbs, thereby enabling it to generate the type of questions that cover a range of the different levels of cognition as opposed to emphasizing more on factual recall. The practical performance of the model demonstrates its effectiveness, with an average semantic similarity score of 88.96%. It also achieves high scores in ROUGE and BLEU, indicating that the generated questions are both contextually relevant and fluent.