Automated Essay Scoring

The way in which computers can be applicable to essay scoring to simplify the process of grading a large amount of student written works has been of growing research interest since the end of the 20th century. Automated essay scoring (AES) is defined as the computer technology that evaluates and scores written prose. Increasing attention to AES has resulted from various factors, among which are cost, standards, and technology, as the advance of IT promises to ease and reduce the cost of the process of grading educational achievements in some spheres. The purpose of this study was to implement natural language processing tools and machine learning for the creation of an automated essay scoring algorithm for the Russian language. It looked in particular at 213 human graded essays written by the last year high school students for the Unified State Exam, that have been automatically collected and processed in order to retrieve language features, which represent the given score, such as the usage of various parts of speech or lack of tautology. Several machine learning algorithms were implemented, including Linear Regression and such classifiers as Decision Trees, Naïve Bayes, kNN and ensemble methods, AdaBoost and Gradient Boosting. The research achieved its goal to discover the algorithm for the automated essay grading. The ensemble model Gradient Boosting was chosen as the most efficient with its f-score of 85%. The results would only be the first step in the exploration of the AES for the Russian language. The features extracted for this research would assist the further study in this field as a basis.