Automated Essay Scoring System with Grammar Score Analysis

An Automated Essay Scoring System is a system that deals with grading Hand written essays without any human intervention. Most of the research done in this field involves direct mapping of an essay’s numerical representation, using word embedding, to its golden score, without any specific trait scoring. So, this research aims to use latest contextual Text Embedding i.e., BERT Embedding for numerical representation of Essay and granulate the scoring of essay into two modules: structure scoring module which deals with scoring essay on the basis of its structure and grammar scoring module, which deals with scoring essay on the basis of its grammatical correctness. To evaluate the proposed model, Quadratic Weighted Kappa Score is used. In this implementation, a QWK score of 0.75 for Structure score and 0.70 for Grammar Score has been obtained. This research and its specific trait scoring can be used as base for future implementation with more detailed feature specific scoring tasks and improve the scope of grammar scoring by considering more grammatical cases.