Learning Analytics on YouTube Educational Videos: Exploring Sentiment Analysis Methods and Topic Clustering

Abstract

The popularity of social media is continuously growing, as it endeavors to bridge the gap in communication between individuals. YouTube, one of the most well-known social media platforms with millions of users, stands out due to its remarkable ability to facilitate communication through the exchange of video content. Despite its primary purpose being entertainment, YouTube also offers individuals the valuable opportunity to learn from its vast array of educational content. The primary objective of this study is to explore the sentiments of YouTube learners by analyzing their comments on educational YouTube videos. A total of 167,987 comments were extracted and processed from educational YouTube channels through the YouTube Data API and Google Sheets. Lexicon-based sentiment analysis was conducted using two different methods, VADER and TextBlob, with the aim of detecting the prevailing sentiment. The sentiment analysis results revealed that the dominant sentiment expressed in the comments was neutral, followed by positive sentiment, while negative sentiment was the least common. VADER and TextBlob algorithms produced comparable results. Nevertheless, TextBlob yielded higher scores in both positive and negative sentiments, whereas VADER detected a greater number of neutral statements. Furthermore, the Latent Dirichlet Allocation (LDA) topic clustering outcomes shed light on various video attributes that potentially influence viewers’ experiences. These attributes included animation, music, and the conveyed messages within the videos. These findings make a significant contribution to ongoing research efforts aimed at understanding the educational advantages of YouTube and discerning viewers’ preferences regarding video components and educational topics.