Article Preview
TopIntroduction
With the booming development of the Internet and social networking sites, social media such as Twitter, Weibo, and WeChat are gradually changing people’s lives. More and more people are sharing their experiences on social media, posting their comments and reviews about a product or service, spreading information, and expressing their opinions or feelings about some issues. For instance, they may discuss hot social issues, comment on national policies and laws, and express joy or sadness about their own experiences. These reviews and opinions can be positive, negative, or neutral (Hammou et al., 2019; Khader et al., 2019; Naik et al; 2021). Using big data analysis techniques such as data mining to analyze data generated by social media users and predict their behavior is of great social importance (Han et al., 2019; Chen et al., 2020). Governments and administrations can analyze people’s emotional responses for hot issues or policies to understand their ideological trends and elicit the prevailing public opinion. Based on users’ reviews, companies can analyze their interests to recommend products they may want to buy or to improve the quality of products, while platforms can recommend content that users may be interested in according to their interests so that like-minded users can interact with each other (Hajiali, 2020; Lu et al., 2020; Jena et al., 2019; Alnashwan et al., 2020).
In addition, in the era of information abundance, there is an unprecedented growth in the demand for online public opinion expression, leading to a large amount of information received and transmitted every day or many tweets posted on the Internet or the spread of misinformation. Some of the information may threaten public safety, which is a great challenge to both Internet-related administrations and public and social stability (Jain et al., 2021; Fahd et al.,2021; Deniz et al., 2021). Hence, analyzing the content posted by users on Weibo and actively guiding public opinion to reduce the impact of misinformation are necessary means to maintain societal stability and security (Liu, 2020; Wang & Shin, 2019). Therefore, it is of great research value and practical significance to analyze the content posted by users on social networks and to accurately classify its sentiment.
Sentiment analysis is classifying the sentiments conveyed in documents or statements posted by users as positive, negative, or neutral. There is an enormous amount of useless data on the Internet and sentiment analysis is needed to analyze the data and extract useful information that expresses specific sentimental content (Wang et al., 2019; Seng & Ang, 2019; Correia et al., 2022). However, when analyzing sentiment in Weibo data, it is important not to focus only on the sentences themselves, but also on the information contained in images, retweets, comments and “likes” on the online platforms. In addition, the user’s personality and influence on others is also relevant to the sentiment polarity of tweets. Therefore, when the sentiment polarity of the text is not obvious, it is of great significance to take other information into account, such as users’ personalities and images, to obtain a more accurate reading of the content’s sentiment polarity (Alqarafi et al., 2019; Lappeman et al., 2021; Shi et al. 2020).
This paper proposes a sentiment classification method of social network text based on an attention mechanism and bidirectional short and long-term memory (AT-BiLSTM) model to address the problem of low accuracy of current text sentiment classification methods and difficulty in effective sentiment prediction. Compared with traditional text sentiment classification methods, the main contributions of the proposed method are as follows.