Using large language models to verify semantic shift parameters
Philology. Theory & Practice · 22.05.2026 · 19 (5) · 1987-1993
DOI: 10.30853/phil20260253 ↗Kurz gefasst
Fachartikel in der Zeitschrift Philology. Theory & Practice, online erschienen am 22.05.2026. Verfasst von Nikita Sergeevich Timofeev. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The aim of this study is to assess the capacity of large language models (LLMs) to recognize diachronic semantic shifts and identify the underlying semantic mechanisms. This paper examines the capabilities of the GPT-5.2 and Claude Opus 4.5 models for this task, utilizing data from the Oxford English Dictionary. A comparative analysis was conducted to evaluate the models’ effectiveness in recognizing semantic shifts arising from various parameters, including changes in evaluativeness, alterations in conceptual scope, metaphor, and metonymy. The scientific novelty of the research lies in determining the limits of applicability of publicly available LLMs to differentiate between these systemic parameters of semantic change without prior fine-tuning. The results establish that the tested models successfully verify transformations of a word’s denotative component when driven by changes in lexical distribution. Nevertheless, the identification of shifts at the level of connotative semantics not anchored in explicit lexical markers remains an open problem.
Zitieren
APA
Timofeev, N. S. (2026). Using large language models to verify semantic shift parameters. Philology. Theory & Practice, 19(5), 1987-1993. https://doi.org/10.30853/phil20260253
Vancouver
Timofeev NS. Using large language models to verify semantic shift parameters. Philology. Theory & Practice. 2026;19(5):1987-1993. doi:10.30853/phil20260253
BibTeX
@article{Timofeev2026,
author = {Timofeev, Nikita Sergeevich},
title = {Using large language models to verify semantic shift parameters},
journal = {Philology. Theory & Practice},
year = {2026},
volume = {19},
number = {5},
pages = {1987--1993},
doi = {10.30853/phil20260253}
}
EUVIMED