Stacking Neural Network Models for Automatic Short Answer Scoring

Abstract Automatic short answer scoring is one of the text classification problems to assess students’ answers during exams automatically. Several challenges can arise in making an automatic system, which quantity and quality data. The data labeling process not easy because it requires a human annotator who expert their field. Further, imbalance also challenge number labels for correct always much less than wrong answers. In this paper, we propose use stacking model based on neural network XGBoost with sentence embedding feature. We upsampling method handle classes hyperparameters optimization algorithm find robust Ukara 1.0 Challenge dataset our best obtained F1-score 0.821 exceeding previous work at same dataset.