Arabic Short-Answer Similarity and Grading Dataset for Automated Assessment Research

This dataset was developed within a sociology course and was used to evaluate similarity algorithms between student answers (SA) and model answers (MA). It extends the dataset previously introduced in [1]. The data were collected from third-year secondary school students and consist of 270 student responses (27 questions per assignment × 10 student responses per question). Each response was evaluated independently by two human graders using predefined model answers. Scores range from 0 (completely incorrect) to 5 (completely correct). The graders worked independently and were unaware of each other’s evaluations. To ensure grading reliability, the final score for each response is the average of the two graders’ scores. The dataset represents Arabic textual answers in a sociology course and can be used for research in automated short-answer grading, similarity measurement, and Arabic NLP applications.Further details on the development, structure, and applications of this dataset are provided in [2]. [1] Shehab, A., et al. (2018). An automatic Arabic essay grading system based on text similarity algorithms. International Journal of Advanced Computer Science and Applications, 9(3), 263–268. [2] Lotfy, N., et al. (2023). An enhanced automatic Arabic essay scoring system based on machine learning algorithms. Computer Modeling in Engineering & Sciences, 77, 1227–1249. Any use of this dataset in published or publicly disseminated work, including journal articles, conference papers, theses, or technical reports, requires proper citation of the two references listed above.