Article Preview
TopIntroduction
With changes in people's lifestyles and the widespread use of the internet, online shopping has become increasingly accepted and ingrained in daily routines (Lissitsa and Kol, 2016). The scale of users engaging in online shopping has gradually expanded. It is evident that as relevant technologies and supporting services mature and improve, the scale of e-commerce has grown significantly. As the popularity of online shopping continues to rise, the volume of product reviews also experiences continuous growth. For consumers, referencing these reviews can assist them in making informed decisions (Khan et al., 2023b). Browsing through these reviews allows users to gain a more comprehensive understanding of products and even learn about real user experiences in advance. This, in turn, enables more effective decision-making, leading to the purchase of ideal products. For businesses, analyzing these reviews and associated data provides a deeper insight into user needs and the expected user experience (Lewis, 1983; Grönroos, 1982). This understanding helps businesses better grasp their product positioning and areas for improvement, allowing for targeted product enhancements and ensuring the success of their products. However, automating the process of crawling consumer reviews and presenting consumer focal points in a meaningful way remains a significant challenge (Zhou et al., 2007).
Text mining is a branch of data mining, a concept formally introduced by Feldman in 1995 (Feldman and Hirsh, 1996). It leverages computer technology to extract implicit and high-value information from semi-structured or unstructured text that is of interest to users (Hotho et al., 2005). By extracting structured information from text and conducting research, text mining achieves highly automated analysis of textual data, making it applicable in various scenarios (Zhao and Chen, 2022). Hei-Chia Wang et al. constructed an integrated summarization system based on text mining algorithms (Wang et al., 2020; Khan et al., 2022). Lei C et al. utilized stacked variational autoencoder technology to extract features from system text, proposing an effective SVAE text feature extraction model (Che et al., 2020). Sandra Maria Correia Loureiro et al. presented a new text mining approach from the consumer's perspective, using aggregated dictionaries based on consumer brand authenticity and brand involvement, offering new insights for brand development (Rosado-Pinto et al., 2020). These methods reflect the development of text mining technology, which is highly significant for machine learning of text semantics. However, these methods have not been further connected with consumer sentiment analysis.