Automated Essay Grading System using NLP Techniques

Abstract

The purpose of the study is to develop an automated essay grading system (AES) which can grade students essays based on various factors. Our proposed system performs grading of essays based on two features. Simple features consist of finding syntactic errors such as spelling mistakes, grammatical errors, punctuations and sentence proportions. Complex features consist of finding semantic errors through discourse analysis, thematic analysis and detection of undesirable style of writing. Many existing AES systems fail to consider the semantic parts of the essay which is addressed in this study. Calculation of score would be done based on what is specified in rubrics. The proposed system is evaluated using datasets from kaggle. The accuracy of model and obtained results show an agreement with teachers’ grading. This gives us an indication that the model can be deployed for assessment of students’ essay, thereby leading to reduction in time, efforts and cost for evaluating an essay.