Tiktok Through AI Eyes: A Deep Learning Approach to Sentiment Analysis

Abstract

Background: The rapid growth of social media has transformed communication, with TikTok standing out among younger users for its short-form videos. Understanding user sentiment on these platforms is key to analyzing public opinion, trends, and engagement. Aim: This study explores sentiment analysis of TikTok user reviews using deep learning approaches, specifically Recurrent Neural Networks with Long Short-Term Memory (RNN-LSTM) and Deep Belief Networks (DBN). With over 144,000 reviews collected from Google Play and Apple App stores, the dataset was preprocessed using techniques such as lemmatization, tokenization, and GloVe word embeddings. The reviews were then classified into positive and negative sentiments. Both models were trained and evaluated based on metrics including accuracy, precision, recall, F1-score, and ROC-AUC. Result: Experimental results revealed that the RNN-LSTM model outperformed the DBN, achieving an accuracy of 81.99% and an AUC of 0.8874, compared to DBN's 78.53% accuracy and 0.8577 AUC. The findings demonstrate the effectiveness of deep learning—particularly LSTM—in capturing sentiment from noisy, user-generated content on platforms like TikTok. This work contributes to the growing field of AI-driven sentiment analysis and provides a foundation for future improvements through hybrid or multimodal approaches.