Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: May 22, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Advancing Arabic automated essay scoring through cross-encoder BERT models and interpretable explanations.

Ayman Mohamed Mostafa1, Meshrif Alruily2, Hisham Allahem3

  • 1Information Systems Department, College of Computer and Information Sciences, Jouf University, 72388, Sakaka, Saudi Arabia. amhassane@ju.edu.sa.

Scientific Reports
|May 20, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Machine Learning Framework for HbA1c Prediction: Data Enrichment, Cost Optimization, and Interpretability Through Stratified Regression and Multi-Stage Feature Selection.

Diagnostics (Basel, Switzerland)·2026
Same author

Progressive Attention-Enhanced EfficientNet-UNet for Robust Water-Body Mapping from Satellite Imagery.

Sensors (Basel, Switzerland)·2026
Same author

Integrating multiomics data using a correlation based graph attention network for subtype classification in lower grade glioma.

Discover oncology·2026
Same author

A Score-Fusion Method Based on the Sine Cosine Algorithm for Enhanced Multimodal Biometric Authentication.

Sensors (Basel, Switzerland)·2026
Same author

Explainable Cluster-Based Predictive Framework for Early Diagnosis of Autism Spectrum Disorder Using Behavioral Biomarkers.

Diagnostics (Basel, Switzerland)·2025
Same author

Efficient multi-finger vein recognition using layer-wise progressive MobileNet fine-tuning and a Dense-Head Probabilistic Siamese Network.

Scientific reports·2025

This study introduces a BERT-based framework for Arabic Automated Essay Scoring (AES), overcoming language challenges. The CAMeLBERT-MSA model achieved high accuracy, offering interpretable results for improved fairness.

Area of Science:

  • Natural Language Processing
  • Computational Linguistics
  • Artificial Intelligence

Background:

  • Arabic Automated Essay Scoring (AES) faces challenges due to diglossia, complex morphology, and limited data.
  • Existing AES models struggle with Arabic's linguistic nuances, hindering accurate and fair evaluation.

Purpose of the Study:

  • To develop and evaluate a novel BERT-based sentence-pair cross-encoder framework for Arabic AES.
  • To enhance the performance and interpretability of Arabic AES systems.
  • To address the linguistic asymmetry in Arabic NLP tasks.

Main Methods:

  • A BERT-based sentence-pair cross-encoder framework was implemented.
  • Five Arabic BERT variants (asafaya/bert-base-arabic, aubmindlab/arabertv02, UBC-NLP/MARBERT, faisalq/SaudiBERT, CAMeL-Lab/CAMeLBERT-MSA) were systematically evaluated.
Keywords:
Arabic automated essay scoring (AES)BERT cross-encodersCross-prompt generalizationExplainable AI (XAI)Integrated gradients (IG)

Related Experiment Videos

Last Updated: May 22, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

  • An Integrated Gradients (IG) explanation pipeline was integrated for interpretability.

    • Main Results:

    • The CAMeLBERT-MSA cross-encoder achieved state-of-the-art results with R2 of 98.47%, MAE of 0.07, and 98.32% accuracy (within ±0.5 points).
    • Cross-prompt evaluations showed generalization symmetry with limited prompts but asymmetry with increased prompt heterogeneity.
    • The IG pipeline generated word-level rationales, identifying key terms for score adjustments, enhancing pedagogical value.

    Conclusions:

    • The proposed BERT-based cross-encoder framework significantly advances Arabic AES performance.
    • Task-specific fine-tuning on Modern Standard Arabic (MSA) corpora is crucial for optimal results.
    • The interpretability pipeline provides valuable insights, promoting fairness and trust in AES systems.