Automated essay scoring linguistic feature: Comparative study

Automated Essay Scoring (AES) is the solution to a tedious and time consuming activity of manually scoring students' essays. AES is usually treated as a supervised machine learning problem where feature extraction plays an important role. In an attempt to investigate the importance of lexical features in AES systems, a new extended feature set is developed by combining popularly known features. The combinedfeature set contains 22 features that captures five different aspects of writing qualities. The importance of each feature in the combined feature set is tested by eliminating each feature separately. It was found that using the number of nouns in the essay slightly degrades the AES system performance. The significance of the combined feature set is compared against three state-of-the-art AES commercial systems and its performance was found comparable.