Automatic assessment of the content of essays based on course materials

Assessing free-text responses, like essays, is a demanding task even for a human. To reduce the costs of essay grading, methods to automate the assessment process have been developed. Approaches based on surface features and content have been introduced. We present a method for evaluating the content of essays based on latent semantic analysis (LSA), which was originally developed for information retrieval and has successfully been applied for assessment of essays in several applications. We use LSA to compare the conceptual similarity between the essays and selected text passages from the course material covering the essay assignment-specific subject matter. In our experiment the correlation between the scores given by the system and human grader varied from 0.78 to 0.82.