The BERT Algorithm for Multimodal Emotion Analysis of Tourism Online Products in E-Commerce Live Broadcasting Based on Big Data

This approach is expected to improve multimodal sentiment analysis of online products within the e-commerce live broadcast environment. The results demonstrate that sentiment analysis of user evaluations based on the bidirectional encoder representations from transformers (BERT) algorithm achieves an accuracy rate of up to 92%, indicating strong potential for future development. The study concludes that emotion analysis of e-commerce products using the BERT algorithm is effective. Accurate calibration of user evaluations also supports product alignment by e-commerce vendors, thereby promoting increased sales, enhanced user satisfaction, and the sustainable, healthy development of live e-commerce platforms. Furthermore, the integration of digital information systems enables e-commerce enterprises to offer more targeted and practical online products while simultaneously enhancing users' subjective satisfaction.