Article Preview
TopIntroduction
In these years, both the speed and size of broadband networks have increased. The public has entered the information age. The advancement of science and technology has greatly facilitated human life. Numerous social platforms, including Weibo and Wechat, e-commerce platforms, including JD.com and Tmall.com, and information platforms, including Toutiao and Zhihu, have emerged, attracting large numbers of Internet users (Demotte, P. et al., 2020; Yuan, J. H. et al., 2020; Dellal-Hedjazi, B., & Alimazighi, Z., 2020). On the Internet, individuals can engage in activities such as expressing their opinions, describing their experiences, and discussing their lives. Such analysis can aid businesses in comprehending the needs of the public and direct government agencies to detect public opinion. Consequently, it is commonly believed that such information is of great value (Etaiwi, W. et al., 2021; Lim, W. L. et al., 2021; Habimana, O. et al., 2020).
Sentiment analysis employs technologies to extract from the text the subjective attitude (positive, neutral, or negative) of reviewers toward a product or topic (Saleh, H. et al., 2022; Ayyub, K. et al., 2022). Today, sentiment analysis technology is frequently used to mine the public opinion of news events and the consumption risk of products to support the management decisions of operators and the selection of consumers (Suriah, G. K. et al., 2022; Cheng, Q. Y. et al., 2020; Dang, N. C. et al., 2020).
Traditional methods of sentiment analysis rely heavily on statistical machine learning techniques based on the frequency of words and parts of speech. I employ naive Bayesian classifiers. The traditional machine learning method is superior to the neural network model when the data size is small. Nevertheless, conventional methods rely on artificial elements. The work is intensive and inefficient, and the performance of the model will depend heavily on the prior knowledge of experts (Elfaik, H., & Nfaoui, E., 2021; Abdullaha, E. F., & Alasadib, S. A., 2021; Yafoz, A., 2022). D-L is a more robust technology that has become increasingly popular in natural language processing (NLP) in recent years. Multilayer neural network architectures can automatically mine complex and deep semantic representations from text and extract text features for sentiment analysis (Chandrasekaran, G. et al., 2022). Most D-L-based sentiment analysis models take word vectors as input and build a neural network-based downstream model to learn the syntax and grammar of text. Using artificial intelligence technology and related technologies in NLP to mine and analyze the sentimental tendencies of comment texts is a popular research topic at present. Extensive research is conducted on the NLP-related technologies to solve the problems encountered with online comment texts (Zhou, J. et al., 2020; Yadav, A. et al., 2020).
The second section introduces related work, the third section introduces the text emotion model algorithm, the fourth section describes the subexperiment and analysis, and the fifth section provides a summary.