Article Preview
TopIntroduction
With the rapid development of Internet + in recent years, Internet information has shown explosive growth. For example, users can easily express their feelings about life and share their joys and sorrows in real time through social platforms such as Weibo, blogs, and Twitter (Zhou et al., 2019). In the era of Web 2.0, numerous Internet products have enabled people to have a greater voice and influence in the network. For example, after some daily consumption, people can comment and score merchants and dishes on life service platforms such as Meituan, Dianping, and Alipay Koubei. These comments and scores directly affect which restaurants other people eat in and what dishes they order (Day & Lin, 2017; Kim et al., 2019; Ruiz-Mafe et al., 2020; Simpson et al., 2011; Xu, 2020; Xu et al, 2019). Among such comments, emotion is the best way to reflect users’ attitudes, thoughts, and judgments. Therefore, emotion analysis task emerged, that is using an intelligent computing method to identify the emotional tendencies expressed by users in a paragraph of text (Rúa-Hidalgo et al, 2021; Yu, 2014).
Emotion analysis also known as viewpoint mining (Meng et al., 2021; Murgia et al., 2018; Tosun & Sezgin, 2021; Wu et al, 2019). Emotion analysis is also widely used in industry for predicting market changes based on news comments and emotions in blogs. The government can understand people’s needs and feelings through the views they express on the Internet, and thus focus on people’s livelihood. Emotion analysis is an indispensable part of the existing recommendation system. The system can understand the user’s preferences by analyzing their emotional changes in comments and can recommend more appropriate content to the user (Christodoulides et al., 2021; Park & Han, 2018). Businesses can understand the advantages and disadvantages of their products and consumers’ opinions on the products according to this information, improve the shortcomings of the products, and promote the products to more suitable target groups. Consumers also want to buy goods that are suitable for them, and they use other users’ reviews to make purchasing decisions. Therefore, comment information is also the focus of consumers. However, e-commerce platforms include a large number of reviews on widely distributed product. It is difficult for consumers and sellers to browse all the review information, and it is even more difficult to find valuable information from the huge amount of reviews. Inappropriate and incomplete comments can lead to regretful shopping decisions by consumers; this directly affects the attitudes of future consumers towards products and market evaluation of products (Han et al., 2018; Kujur & Singh, 2018; Wang et al, 2021).
A growing number of deep learning models have been applied to natural language processing tasks, especially the emergence of word vector technology, which enables text to be represented as low-dimensional and continuous features. The advantages of automatic learning and feature extraction by deep learning can not only overcome the disadvantages of complex feature engineering brought by traditional machine learning to a certain extent, but also achieve better results (Dhaoui et al., 2017; Haavisto & Sandberg, 2015).
The rest of the paper is organized as follows: The second section provides the literature review; the third section describes the text sentiment analysis based on the deep learning model; the fourth section presents the experiment and its analysis; lastly, the fifth section provides the conclusion.