Exploring Attitude in Political Discourse with Large Language Models: The Case of News Articles on Trump’s Election Victory

Donald Trump’s election victory, as a landmark international political event, offers a critical site for observing how national ideologies are discursively constructed and communicated through news media. In particular, the reporting of such events provides a rich empirical context for examining the deployment of evaluative language and its ideological implications. This study investigates how attitude resources are realized across news discourse produced by Chinese, American, and British media, with the aim of uncovering both cross-national patterns and the ways in which evaluative meanings align with broader framing practices. Drawing on the appraisal framework from Systemic Functional Linguistics, the analysis focuses on 60 English news articles and employs RoBERTa to automatically annotate instances of attitudinal resources. In addition, the study integrates DiVoMiner, a discourse-structure-aware computational tool, to identify dominant news frames and explore their alignment with specific attitudinal patterns. The findings reveal that although national media exhibit culturally shaped differences in their use of attitude resources, there is a notable convergence in the overall discursive structure of reporting. This convergence suggests the emergence of a shared transnational framing logic in the coverage of global political events. These results not only contribute to a deeper understanding of evaluative strategies in multilingual media discourse, but also demonstrate the potential of combining systemic functional theory with large language model–based annotation to analyze the ideological underpinnings of news narratives.