AUTHOR=Meng Haohan , Li Xiaoyu , Sun Jinhua 

TITLE=Large language models prompt engineering as a method for embodied cognitive linguistic representation: a case study of political metaphors in Trump’s discourse

JOURNAL=Frontiers in Psychology

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2025.1591408

DOI=10.3389/fpsyg.2025.1591408

ISSN=1664-1078

ABSTRACT=Embodied-Cognitive Linguistics inherits and further develops the core concepts of Cognitive Linguistics, maintaining a focus on embodied cognition and conceptual metaphors. It emphasizes that language is not merely a cognitive phenomenon but also a product of human social interactions and economic conditions. From this perspective, metaphors extend beyond their simple linguistic representation and become essential structures of human cognitive expression. Political metaphors, in particular, are instrumental in shaping public ideology and emotional engagement, a phenomenon clearly demonstrated in the political speeches of Donald Trump. With rapid advancements in large language models (LLMs) technology, traditional approaches to metaphor identification are undergoing significant transformation. By leveraging the advanced text parsing and generation capabilities of LLMs, new opportunities emerge for the automatic detection and nuanced analysis of political metaphors. This study employs a critical metaphor analysis (CMA) framework, integrated with a chain-of-thought-based prompt engineering (PE) technique, utilizing the ChatGPT-4.0 Python environment to identify and examine metaphors in Trump’s speeches. The results reveal that Trump strategically employs metaphors derived from diverse source domains—such as Movement and Direction, Illness and Health and Force—to resonate emotionally with his audience. Methodologically, while LLMs demonstrate considerable strengths in analyzing political discourse, challenges remain in areas such as semantic differentiation and expression. Future research will focus on optimizing these models, conducting comparative analyses with traditional methods, and exploring their applicability in cross-cultural contexts, with the goal of providing more precise and effective tools for both natural language processing (NLP) and political linguistics research.