A Sentiment-Based Approach to Predict Learners’ Perceptions Towards YouTube Educational Videos

Given the increasing proliferation of educational resources on YouTube, this social media platform has become an optimal space to which students usually resort to improve their learning in various subjects. It has also turned into an online community in which all users can voice their opinion about videos they have watched. However, getting the perceptions of learners on these learning materials is not always an easy task. Our objective in this paper is, therefore, to propose a machine learning approach that can classify learners’ feedback as being positive or negative. For this purpose, textual data was retrieved from YouTube educational videos, pre-processed and then classified using different machine learning algorithms. Results of the present study demonstrated that the Support Vector Machines algorithm performs better than other models. In fact, with an accuracy score of 92.82% on combinations of unigrams and bigrams and 92.67% for associations of unigrams and trigrams, this supervised machine learning approach outperforms other implemented models, namely Naïve Bayes, Random Forest and Logistic Regression. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.