Efficient Model for Searching and Detecting Semantically Similar Question in Discussion Forums of e-learning Platforms

Udemy and Nptel are online learning platform where learners get access to wide variety of courses. These platforms have 2 issues which we aim to solve. First, Currently in Udemy the questions asked in discussion forum are not divided based on course videos, i.e. there is common discussion forum where question with respect to any content in the course can be asked. This is problem because, when a new question is asked all the question in the discussion forum have to be compared with the new question for similarity. This can take up significant amount of computational time. To avoid this we have applied Random Forest classifier to find out the category of a newly asked question (with respect to Udemy it would mean finding which video the question relates to), if category is found only previously asked questions in the particular category are compared for similarity rather than the previously asked questions in the whole database. Second, Currently NPTEL is not course coordinator friendly. This is because lots of times there is redundancy in the questions asked, this is burden on the course coordinator because he/she has to answer to each and every version of the same question. This has been tackled by using techniques like GBDT, SVM to detect semantic similarity of questions. The Random Forest classifier achieved accuracy of nearly 73% and GBDT achieved accuracy of nearly 64%.