Comparative Study of Deep Learning Models for Sentiment Classification in YouTube Comments
Sadargari Viharika, N. Hiranmayee, B. Srilatha, Sirisetti Gopi Chandu, U. Nagendra · 2025
In the last ten years, social media enabling users to interact by sharing their personal experiences, educational content, and emotions has skyrocketed to unprecedented levels. Comments, images, reactions and stickers have become common currency on social media platforms. The world became acquainted with Facebook, YouTube, Twitter and Instagram which are now globally utilized as multifunctional communication and interaction platforms by people spanning numerous age categories, something that became pronounced after 2010. In Myanmar, YouTube has gained remarkable popularity, particularly among younger users who use it to upload personal videos, short films and music clips as well as comment on them in emotionally charged terms of joyous, hopeful, sad, and even angry sentiments. Text classification has become one of the most important tasks in Natural Language Processing (NLP) and one of its subdivisions, Sentiment Analysis, deals with the extraction and classification of emotions contained in comments as positive, negative or neutral. This research focuses on YouTube comments sentiment analysis in English through the Bi-Lstm (Bidirectional Long Short-Term Memory) deep learning model vis-a-vis the previously existing Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU) models which sentiment classify in Myanmar Language. The effectiveness of the Bi-Lstm model is assessed using accuracy, precision, recall, F-score, and other associated metrics, and is compared with existing deep learning models in relation to classification accuracy.