Multilingual social media sentiment analysis with deep learning—Taking the cross-cultural communication of Chinese Hanfu culture on YouTube as an example

The development of social networking services (SNS) has eliminated spatial constraints, blurring the boundaries of social relations between physical and digital spaces. As a result, the content and information on social media are exploding. Users mainly upload, watch and share videos on YouTube. As a global social media platform, it has users from more than 80 countries and regions, and the number of its active users exceeds two billion. Users can effectively carry out brand promotion, cultural communication, and other communication activities on YouTube platform. This is the introductory text. This paper describes the current situation of sentiment analysis on mainstream social media. It explores the implementation of multilingual text sentiment analysis based on deep learning. Taking the cross-cultural dissemination of Chinese Hanfu culture on YouTube as an example, this study adopts. NMT and CNN_Text_Word2vec model to extract emotional features, combined with likes weights to perform sentiment analysis of social media texts in multiple contexts. As suggested by the experimental results, users’ emotions toward Chinese Hanfu cultural videos on YouTube exhibit a bimodal distribution trend, with strong positive and strong negative emotional feedback far exceeding weak positive and weak negative emotional feedback. In this paper, combined with the weight of likes on YouTube, users’ emotions toward a certain topic are investigated from multiple perspectives, which is supposed to reflect users’ true emotions and provide decision-makers with valuable decision-making suggestions.