Essay Questions nowadays are one of the essential tools to evaluate students at all educational levels because it measures high-level skills such as linking ideas and using complex semantic structure. Essay assessment is time-consuming, challenging, and requires long focus time to understand, find mistakes, and grade. These questions are mainly characterized as a subjective evaluation since there is usually a wide range of answers. Designing and developing efficient and accurate machine-learning models will help both students enhance their writing skills and teachers evaluate essays.
In this thesis, I examine the manual process of grading essays and develop a system that can grade them automatically for English essays. The process contains feature extraction, applying different machine learning techniques, and comparing the results of different models. I rely on the Natural Language Toolkit (NLTK) to extract features from the data set. In terms of evaluating the models, I used mean square error (MSE) and root mean square error (RMSE) as of measurement metrics.
Ultimately, a further study could enhance the model by expanding the training set and applying more machine learning techniques to reduce the gap between the system and the human assessment.