Automated Essay Scoring Using Machine Learning

• Published in: 2022 4th International Conference on Cybernetics and Intelligent System (ICORIS) pp. 1 - 5
• Main Authors: Kusuma, Jason Sebastian, Halim, Kevin, Pranoto, Edgard Jonathan Putra, Kanigoro, Bayu, Irwansyah, Edy
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08.10.2022
• Subjects: AES; Automated Essay Scoring; Bit error rate; Deep learning; Feature Engineering; Measurement; Neural networks; Semantics; Systematics; Vocabulary
• DOI: 10.1109/ICORIS56080.2022.10031338

Essays are frequently employed in the educational system to gauge students' comprehension of particular subjects. However, marking essays requires a lot of time and work and could be prejudiced. In order to save time, lessen human effort, and eliminate biased scoring, automated essay scoring tries to automate scoring. Due to its lack of transparency, limited language support, and requirement for tagged data for the target prompt, which is not always available, AES is still not frequently utilized. This study's goal is to examine automated essay scoring methods. The PRISMA Flow Diagram is used in this study to conduct a systematic literature review. Studies that were released between 2016 and 2021 were found. Information pertinent to the research topics is taken from these studies and then processed to provide a response. Datasets, methods, and models are found in the publications. The performance score of models utilizing the same dataset is then used to compare them. According to the study, AES uses feature engineering and deep learning as its two core methodologies. More scholars are currently researching the deep-learning methodology. CNN, LSTM, and BERT are a few examples of neural network models used in the deep learning method. Most studies use the average QWK and the ASAP dataset as performance metrics. SBLSTMA (Siamese Bidirectional LSTM Neural Network Architecture) and BERT + handcrafted-features, both with 0.801 average QWK, are the models with the highest performance score on the ASAP datasets.