A Hybrid Deep Learning Approach for Spam Comment Detection on Vietnamese Healthcare YouTube Channels

Abstract

With the explosion of digital content and the rapid growth of social media platforms like YouTube, the problem of comment spam has become increasingly complex. Spam comments negatively impact user experience and diminish the quality of online interactions. Especially in the healthcare domain, inaccurate information from spam comments can lead to serious consequences regarding consumer trust and the quality of healthcare services. This research presents an approach to classifying spam comments based on a hybrid deep learning model, incorporating PhoBERT, TextCNN, and BiLSTM, to process and classify YouTube comments related to the healthcare field. The proposed method not only allows for improving accuracy in comment classification but also helps to optimize content management costs. By filtering out inappropriate comments, the proposed model helps to enhance user trust, thereby contributing to improving economic efficiency by increasing opportunities for valuable interactions between healthcare providers and users.