ABSUM: ABstractive SUMmarization of Lecture Videos

Nowadays, most of the educational institutes record lecture videos for online courses and allow students to watch those lecture videos. Most of the videos have a lengthier duration with very little indication of the subtopics. These videos also consume large amounts of backup memory space over a period of time. For many students, one of the biggest challenges of watching lengthier lecture videos is the struggle with focusing on the screen for long periods of time. With this online learning method, most of the time students get easily distracted by social media and other platforms. The proposed work Abstractive Summarization of Lecture Videos (ABSUM) aims to summarize the lecture videos. In lecture videos, instructors used to write down the most important keywords, phrases, and mathematical formulae in the chalkboard as well as used to teach orally. Only the most important parts of the video should be contained in the created summary. With the number of lecture videos increasing at a rapid rate day by day, the automatic video summarization will be beneficial for anyone who wants to save time and learn more in less time. In this proposed system, every frame is binarized with the help of FCN model, and each frame is reconstructed by CC grouping to form key frames. From the identified key frames, content from the board is extracted. The text from audio is weighted with the text from the key frames and summarized.