Course-based online learning video retrieval system

Audio-visual learning media, such as learning videos on YouTube, is an alternative representation of the learning media commonly used during the COVID-19 pandemic. There are many learning videos available on YouTube; thus, selecting the appropriate learning videos is a challenge in the research field of information retrieval. This study focuses on building an information retrieval system that can recommend online learning videos based on the user requirements, namely the course name and its syllabus. The proposed information retrieval system works through cosine similarity measurement to calculate the closeness between the course and its syllabus with video annotations. The video annotation consists of the title and description metadata of the video captured in real-time from YouTube using the YouTube API. The proposed information retrieval system provides the five most recommended learning videos based on the selected courses and their syllabus. The experimental results involving 40 respondents show that the online video retrieval system has a good performance proven by an accurate level of 80.5% that in line with the relevance score of 84. Furthermore, another three performance indicators also great, i.e., the values of novelty, serendipity, and increasing recommendation diversity are 78%, 77%, and 85.5, respectively