Subjective Evaluation using LSA Technique

Information and communication technologies are effectively used in educational administration and teaching-learning process. There is a need to utilize them for educational evaluation. Subjective evaluation and assessment are essential part of education system. For evaluation of subjective examinations, several statistical and mathematical techniques such as Latent semantic analysis, and maximum entropy are being used. Latent semantic analysis (LSA) is one such technique used for evaluation of non-technical prose. In this paper LSA technique is applied to evaluation of technical answers. The LSA algorithm clusters the text into groups on the basis of similarity. The detailed LSA algorithm is included in the paper. A prototype based on the LSA algorithm is developed using soft-computing platform MatLab and the programming language Java. The prototype so developed has been tested by conducting class tests of students of under-graduate classes of computer science courses. The results are analyzed and observed to be satisfactory. It is concluded that LSA technique can also be applied for the evaluation of technical answers.