Deep Learning-Based Topic Modeling for Apex Legends User Reviews

With the rapid evolution of the gaming industry, understanding user preferences is crucial for improving game quality. This study tackles the challenge of extracting insightful topics from extensive and unstructured user reviews for the game Apex Legends. Traditional topic modeling methods like Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF) often fail to capture the nuanced semantic context of user feedback. To overcome these limitations, we apply deep learning approaches, specifically Top2Vec and BERTopic, which excel in uncovering hidden user preferences and identifying areas for improvement. The research involves collecting and preprocessing Apex Legends reviews, followed by topic modeling with Top2Vec and BERTopic methods. BERTopic, enhanced with all-MiniLM-L6-v2 embeddings and optimized hyperparameters, significantly outperforms traditional approaches in topic coherence and interpretability. Our analysis uncovers substantial user dissatisfaction with technical issues, while also highlighting positive feedback on gameplay and characters. These insights provide valuable guidance for developers aiming to refine the gaming experience in Apex Legends.