A Master's thesis at the College of Computer Science and Information Technology, University of Basrah, explored the topic of "Off-Topic for Automatic Arabic Assessment."
The thesis, submitted by student Hanadi Hamed Abdulrahman, aimed to develop an intelligent system for automatically correcting written Arabic responses. The system focused on detecting off-topic deviations and improving the accuracy of grading, relying on deep learning techniques and language converter models, specifically the AraBERT v2 model, which is pre-trained on Arabic texts.
The thesis included developing a hybrid framework that combines contextual semantic representations extracted from AraBERT v2 with a set of additional analytical indicators. These indicators included measuring semantic similarity using Cosine Similarity, calculating the lexical coverage of key ideas, and aligning the structural framework of ideas. These features were then integrated into a Random Forest-based regression model to predict the final score. The system demonstrated a strong correlation with human assessment, enhancing its potential for supporting electronic assessment systems and improving the reliability of automated grading in the Arabic educational environment.
