Abstract

Live e-commerce has revolutionized the marketing of agricultural products, but systematic analysis of unstructured live interactive comments by viewers remains underexplored. This study uses the Latent Dirichlet Allocation (LDA) model to analyze the latent topics of 770,000 comments in live broadcasts of orange sales at different merchant levels on the Douyin platform. By combining the axis-folding method with perplexity to determine the number of LDA topics, the analysis shows that: (1) product reviews emphasize product quality; (2) product prices highlight price sensitivity; (3) quasi-social interactions reflect emotional engagement; (4) user interests reflect purchase intentions; (5) e-commerce logistics focuses on delivery efficiency; and (6) the association between live broadcast promotions showing urgency prompts and conversion rates. The results empirically show that the success of agricultural product live broadcasts depends on product quality transparency, dynamic pricing, after-sales service, and emotional resonance between the host and the audience.

Download

Share

COinS