With the rapid expansion of e-commerce, consumers increasingly rely on online platforms to purchase fashion products. However, the vast selection of products often leads to choice overload, making it challenging for consumers to find products that meet their needs. To address this challenge, we propose an advanced Conversational Recommender System (CRS) that applies a Functional, Expressive, and Aesthetic (FEA) consumer needs model. Using this model as a theoretical framework, this study constructs a taxonomy of consumer needs organized into categories and subcategories, each containing multiple attributes, and uses a Large Language Model (LLM) to apply it to review data, extracting attributes that influence purchase intention. Furthermore, CRS experiments were conducted to assess the impact of consumer needs attributes on recommendation performance. Our results indicate that the FEA model-based consumer needs taxonomy effectively categorizes consumer needs, with ease, good value for money, and leg contouring emerging as the most frequently mentioned attributes. Moreover, consumer needs attributes vary across different pants types, highlighting the importance of need-aware recommendations. Experimental evaluation of the CRS demonstrates that incorporating consumer needs attributes improves the recommendation success rates and reduces the average number of turns. Through the empirical application of the FEA model in the CRS, this study demonstrates its effectiveness in improving recommendation performance and its potential in enhancing consumer satisfaction.
扫码关注我们
求助内容:
应助结果提醒方式:
