Auto-evaluation of subjective answers is one of the most debatable problem in E-learning and many have worked on this problem which has transformed education in a significant way. However most of AESA techniques are based on words based similarity . By introducing bidirectional techniques in these we can solve this problem and compare student and teachers answers on basis of context rather than on basis of words. Our most of the evaluation is done using Bert model technique and Bi-LSTM. Introducing this software in e-learning platforms can bring evolution in education.