Autonomous Educational Testing System Using Unsupervised Feature Learning

With the increase of ubiquitous data all over the internet, intelligent classroom systems that integrate traditional learning techniques with modern e-learning tools have become quite popular and necessary today. Although a substantial amount of work has been done in the field of e-learning, specifically in automation of objective question and answer evaluation, personalized learning, adaptive evaluation systems, the field of qualitative analysis of a student’s subjective paragraph answers remains unexplored to a large extent. The traditional board, chalk, talk based classroom scenario involves a teacher setting question papers based on the concepts taught, checks the answers written by students manually and thus evaluates the students’ performance. However, setting question papers remains a time consuming process with the teacher having to bother about question quality, level of difficulty and redundancy. In addition the process of manually correcting students’ answers is a cumbersome and tedious task especially where the class size is large. In this paper, we put forth the design, analysis and implementation details along with some experimental outputs to build a system that integrates all the above mentioned tasks with minimal teacher involvement that not only automates the traditional classroom scenario but also overcomes its inherent shortcomings and fallacies.