How AI Describes Nirvana: Can Large Language Models Compile Explanatory Dictionaries?

This article examines the potential of large language models for automating lexicographic descriptions. The study aims to evaluate the quality of dictionary entries produced using contemporary multilingual models and to identify the limitations of their application in academic lexicography. The research material comprises two lexical groups: neologisms from the field of artificial intelligence and vocabulary of Buddhist origin. Five language models were tested, including both domestic and international developments. The methodology combines a detailed prompt designed based on the principles of academic lexicography, parameter tuning, the direct generation of lexicographic descriptions and the subsequent expert validation of outputs. The results show that, while the models produce definitions of comparable quality, they encounter significant difficulties in assigning stylistic labels and selecting illustrative examples. The most stable performance was observed in cases involving vocabulary that is deeply integrated into the Russian language system. The analysis highlights several limitations of language model applications, such as the formalised use of stylistic labels, the generation of false definitions based on phonetic similarity, and the fabrication of large quantities of illustrative examples with fictitious sources. The study concludes that, at the current stage of development, language models can only serve as auxiliary tools in lexicographic practice, requiring mandatory expert validation and strict methodological control.