Research develops a classifier to accurately classify exam questions using indobert and hyperparameter fine-tuning, suggesting enhanced educational assessments.
Key Points
IndoBERT-LSTM achieved the highest accuracy of 88.75%, outperforming IndoBERT-CNN in question classification.
The study involved hyperparameter fine-tuning, yielding optimal performance with a learning rate of 5e-5 and a batch size of 64.
Automated classification using deep learning can support educators in creating higher-order thinking assessments based on bloom's taxonomy.
IndoBERT demonstrated strong results but was limited by its focus on the Indonesian language and the interpretability of predictions.