Study of Zhejiang Tangerine E-Commerce Reviews Based on Natural Language Processing

In recent years, the global economy has experienced significant shifts, leading to a trend of consumption downgrading. Amid economic pressures and uncertainties, consumers are increasingly turning to cost-effective shopping methods. The COVID-19 lockdowns further accelerated the growth of e-commerce platforms, presenting both opportunities and challenges for sales. Electronic commerce has played a crucial role in enhancing the sales of agricultural products with regional characteristics in China, thereby opening new channels for farmers. This article utilizes tangerines, particularly popular in Zhejiang Province, as a case study to explore e-commerce reviews and assist merchants in delivering more satisfactory products. The analysis of tangerine reviews revealed that customers primarily focused on the taste, service, quality, and price. By applying the latent Dirichlet allocation (LDA) topic model, comments were categorized into four themes: ‘quality’, ‘service’, ‘price’, and ‘flavor’, with key terms identified for each theme. Through sentiment analysis using SnowNLP and bidirectional encoder representations from transformers (BERT), it was found that online shoppers generally expressed positive sentiment toward tangerines. However, there was also some negative feedback. These findings are of paramount importance for businesses aiming to meet consumer demands. The study acknowledges certain limitations including the reliability of data mining and the accuracy of Chinese corpus analysis. Future research could benefit from employing more precise language models to enhance the analysis, ultimately improving the consumer shopping experience and aiding businesses in service improvement.