Research on data-driven optimization of cross-border e-commerce copywriting and artwork

In the competitive cross-border e-commerce market, copywriting and artwork are the key factors to attract consumers and improve the conversion rate. Based on the data-driven perspective, this paper thoroughly researches the optimization strategy of cross-border e-commerce copywriting and artwork. By utilizing text generation technology in natural language processing and based on the collected cross-border e-commerce marketing copywriting data, we construct a keyword topic-controlled copywriting generation model (Cross-GRU) based on cross term encoder, so as to improve the quality and efficiency of copywriting. The semantic fusion-based generative adversarial network (SF-GAN) framework is used to build the coding and decoding structural discriminator-based generative adversarial network (SF-GAN-V2) model in this paper, which is combined with the copy generated by the Cross-GRU model to correspond to the generation of cross-border e-commerce aesthetics images, so as to realize the co-optimization of copy and aesthetics images. In the optimization experiment analysis, the automatic and manual evaluation scores of this paper’s Cross-GRU model outperform other models, and the accuracy of copywriting generation is as high as 88.49% in the case of a given topic keyword. The SF-GAN-V2 model also shows good performance in the experiment, and the IS and FID index scores on this paper’s homemade dataset are better, realizing the cross-border e-commerce copywriting and artwork co-optimization.