LARGE LANGUAGE MODELS FROM A COGNITIVE LINGUISTIC PERSPECTIVE

This article examines Large Language Models (LLMs) from the perspective of cognitive linguistics. It explores how LLMs process, represent, and generate linguistic meaning using statistical learning while comparing these mechanisms with human conceptualization and cognitive processes. Particular attention is given to conceptualization, frame semantics, conceptual metaphor, categorization, and cognitive grammar as theoretical frameworks for evaluating language models. The findings indicate that although LLMs can simulate many aspects of human language use, they do not possess embodied cognition or experiential knowledge comparable to human conceptual systems. The study highlights both the potential and the limitations of LLMs in natural language processing and demonstrates the value of cognitive linguistics for understanding and improving artificial intelligence.