An Improved Automated Question Answering System from Lecture Videos

Learning is important for students for individual as well as nation development. However, due to the ever increasing student base and lack of sufficient number of mentors all over the world, one-to-one student-teacher interaction in real life has become a tough task. Question Answering (QA) systems reduce the requirement for physical interaction between students and teachers. QA systems allow students to post their queries and get answers for the same. The paper discusses implementation of an automated QA system that uses a knowledge base constructed from the video lectures recorded in the classrooms as well as the ones that are available online. Two approaches to generate the transcript from the video lectures have been implemented: (1) with use of Online Speech Recognition APIs (Application Programming Interfaces) and (2) with use of offline method. CMU sphinx has been automatically trained using predefined dataset in offline method. Transcripts from both approaches, i.e., online and offline are then compared on the basis on semantic similarity and results are presented.