AUTOMATED ASSESSMENT OF ESSAYS AND SHORT-TEXT ANSWERS

Abstract

Automatic essay evaluation is a challenging research field. Several technologies have gained considerable achievement in this field. In this project, a prototype of Automatic Essay Grader is implemented with Latent Semantic Analysis (LSA) technology, an effective technology for extracting and representing contextual usage meaning of words by applying statistical computations to a corpus of text. LSA is a powerful tool because of its characteristic corpus based representation of words, the information theoretic weighting, the use of the cosine to calculate distances between texts, and also Singular Value Decomposition (SVD) despite of its limitation in distinguishing words order. Upon the prototype of the essay grader software, several improvements on the algorithm were made and the web implementation was successfully realized. Test based on the corpus of IT management shows that the software performance is comparable to other essay evaluating applications.