The objective of this research is to carry out a tone and semantic sentiment analysis of network comments on new media platforms by leveraging the BERT model. With the burgeoning popularity of social media, network comments, rich in emotional and tonal features, have emerged as a significant part of the online culture. Accurate interpretation and analysis of these comments' sentiment and semantic meanings are paramount to grasping online public opinion and user psychology. In this study, the BERT model, lauded for its bidirectional encoding and contextual understanding capabilities, is selected to scrutinize the sentiment and tone of network comments on new media platforms. Through a process of pre-training and fine-tuning, the sentiment attitudes and polarity of comments are accurately identified along with their conveyed tonal features, such as joy, anger, and sarcasm. Conducting an accurate tone and semantic sentiment analysis of network comments on new media platforms facilitates a profound understanding of user preferences and trends in public opinion. This can assist in optimizing content recommendations, enhancing user experiences, and increasing the operational effectiveness of new media platforms. The outcomes of this research will bear significant implications for studies and applications in online culture, offering invaluable references and guidance in related domains.