Automatic Answer Assessment in LMS Using Latent Semantic Analysis

E-Learning uses systems like Learning Management Systems (LMSs), to support and enhance teaching-learning process. When we compared some of the popular LMSs, a lack of efficient mechanisms to assess descriptive answers like long answers and essays, was observed. Also, it is tedious for the teachers to manually evaluate for large number of students. Hence, we propose an automatic system to assess descriptive answers of students and provide teachers with immediate feedback. This is achieved by first, comparing student's answer with teacher's ideal answer set, using latent semantic analysis (LSA). Then, estimating the order of previous and upcoming words in an answer using positional indexing, based on the keyword list added by the teacher. A final score is then generated depending on LSA, correct keyword usage and also on spell check. Cohen's kappa coefficient of human rater-tool agreement showed a good strength when the system was integrating into an existing LMS.