Text Sentiment Analysis Based on Binary Images

Dawei Xu, Yue Lv, Min Wang, Fan Huang, Jiaxin Zhang

Abstract: Privacy protection is an issue of great concern in today's information society. Ensuring the privacy and security of personal and sensitive information is critical. In the field of privacy protection, text sentiment analysis plays an important role. By applying text sentiment analysis technology, potential privacy leak risks and threats can be discovered and identified promptly, improving the detection and response capabilities of privacy events, and further strengthening the privacy protection of individuals and organizations. In recent years, with the continuous increase and complexity of text data, traditional machine learning, and deep learning methods may face some challenges in text sentiment analysis. This study not only focuses on the methods and techniques of text sentiment analysis but also pays special attention to its importance in the field of privacy protection. This article achieves the purpose of sentiment analysis by using text data to construct binary images of different dimensions and using multiple models for classification, thereby ensuring that private information is not leaked or abused. Finally, the research results are compared and evaluated with traditional methods to verify the practical application value of the proposed method in the field of privacy protection. The study conducted experiments on the Weibo_senti_100k and online_shopping_10_cats (Online shop) datasets and demonstrated that the method proposed in the study performed well, with an accuracy of 95.24% and 95.94% on the two datasets, respectively.