Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal International Journal of Advances in Data and Information Systems
Orvalamarva, Orvalamarva
Unknown Affiliation
Computer Science & IT Electrical & Electronics Engineering

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Multi-Output Classification of Cognitive Levels and Topics in Indonesian Questions using Deep Learning and Transformers Orvalamarva, Orvalamarva; Pratiwi, Oktariani Nurul; Fakhrurroja, Hanif
International Journal of Advances in Data and Information Systems Vol. 7 No. 1 (2026): April 2026 - International Journal of Advances in Data and Information Systems
Publisher : Indonesian Scientific Journal

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Managing large-scale digital question banks struggles with manual metadata labeling, especially when identifying material topics and cognitive levels based on the Revised Bloom's Taxonomy. Current automated approaches usually treat these two attributes as separate tasks, which adds to the system's complexity and computational load. This study introduces a multi-output classification method using a shared encoder architecture with two task-specific heads to predict topics and cognitive levels simultaneously. We performed experiments on 685 Indonesian junior high science questions, covering 15 topic labels and four cognitive levels (C1–C4), with an imbalanced distribution in which lower cognitive levels accounted for more than 75% of the dataset. To handle this imbalance, we applied Focal Loss to taxonomy classification, and class weighting was used in the comparison model. A comparative study involved CNN, BiLSTM, DistilBERT, and IndoBERT. Our results demonstrate that IndoBERT delivered the best performance, with F1-macro scores of 0.78 for topics and 0.71 for cognitive levels and showed better performance in minority classes compared to standard cross-entropy-based models. These findings suggest that an integrated multi-output approach can boost the efficiency and accuracy of question labeling and offers potential for integration into Computer-Based Test systems and e-assessment platforms in real time.
Co-Authors Fakhrurroja, Hanif Oktariani Nurul Pratiwi