A Study of Emotional Tendencies’ Differences in Reply Texts Between Local and International Chatbots: An Analysis Based on the Sina News Context

Under the background of advanced digital socialization and frequent cross-cultural communication, the collision and fusion of multiple cultures on social media platforms make cross-cultural research very important. However, the ordinary translation function on social media platforms can hardly solve the communication barriers arising from the expression of specific cultural emotions, and the unique interactive function and opinion-guiding ability of chatbots with the ability to express emotions may bring new possibilities for cross-cultural communication, but most of the chatbots are based on the training of local big language models, which will present significant local cultural characteristics. Therefore, the study focuses on the emotional tendencies’ differences between local and international chatbots in their reply texts, and analyses them in the context of Sina News. By selecting the texts of Sina News and human users’ comments in December 2024, combined with the reply texts of local and international chatbots in Llama 3.1, Mistral v0.3, Qwen 2.5, and GLM-4, the study is conducted using the emotion analysis tool based on MegatronBERT. The results show that, firstly, there are differences between the emotion tendencies of chatbots and human users, and the emotions of human users’ comments are richer and more varied, and this difference and variation shows some variability depending on the type of news texts. Secondly, the emotion of chatbots reply texts is not exactly the same as that of news texts, and its emotion richness is higher than that of news texts, but chatbots from different cultures show different emotion tendencies; Thirdly, there are significant emotional tendencies differences between chatbots from different cultures, with local chatbots tending to output Joy emotion while international chatbots show higher intensity of Fear emotion. These results provide important references for the optimal design of chatbots, their use in news scenarios, and their application in cross-cultural communication.