Automatic Question Extractor and Answer Detector in a Recorded Webinar

Online learning becomes the primary method of education due to the novel coronavirus (COVID 19). This research paper describes the automatic extraction of domain-related questions in a lecture video, identifies the answers given by the lecturer for both voice-based questions and chat questions. The paper also presents a method to identify whether a lecturer gave a valid answer for the chat-based questions in the later section of the video. Additionally, this paper describes the approach to identify the most accurate solutions from the custom search engine-based responses.

I. Introduction

During the pandemic era of the novel coronavirus (COVID-19), the Sri Lankan education system and the entire globe's education system have moved to online platforms. The demand for video conferencing apps has surged during this period, such as Google Meet, Skype, Zoom, Microsoft Teams, etc. All the state and non-state universities moved to this new normal [1]. For example, all the lecturers in the Sri Lanka Institute of Information Technology (SLIIT), one of the top private universities in Sri Lanka, conducts webinars with students through the platforms mentioned above, and upload recorded lecture videos to the Eduscope video portal at lecturecapture.sliit.lk. Most students prefer to watch video lectures before the exam again to refresh their memory. Instead of just watching the video, most students in the survey requested that it be better if the system can automatically identify the Q&A sessions. The “Automatic question-answer extractor in recorded webinars” was implemented to address that problem. After recording a webinar, the system will automatically identify domain-related questions of students and lecturer answers with additional suggested answers. It is a great benefit for students for exam preparation.