Cognitive Modeling of Ethnocultural Units with Somatic Components in AI Systems and Semantic Deviations in Translation

B. Kamyza, Tynyshtyk Yermekova, G.A. Eskermesova

2026

Full text Open access: Hybrid Metadata (CSL-JSON)
DOI 10.55491/2411-6076-2026-3-212-229
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

The article is devoted to a pressing issue in modern computational linguistics and NLP – investigating the ability of Large Language Models (LLMs) to recognize ethnocultural codes and national worldview layers in the Kazakh language. The relevance of this study is driven by the fact that the primary limitation in current AI-based translation systems lies in their inability to correctly identify ethnocultural units, combined with a scarcity of research in this field and a growing demand for AI models adapted to translation erratology capable of accurately conveying cultural meanings. The purpose of the study is to identify the cognitive mechanisms underlying the processing of Kazakh somatic ethnographisms (zhürek zhalgau, bauyrymdau, auzyna tükirtu, etc.) across ChatGPT, Gemini, and Claude systems, as well as to classify semantic deviations (aberrations) in the digital environment. The study employs a set of methods, including cognitive modeling, linguocultural analysis, and comparative-experimental techniques. The empirical base comprises somatic ethnounits extracted from the classical literary works of A. Kekilbayev, M. Auezov, T. Alimkulov, and T. Akhtanov. To test the models' context sensitivity, two-stage (lexical and context-pragmatic) linguistic prompts were developed. The experimental analysis results demonstrated that AI systems are frequently prone to errors such as “pragmatic inversion,” “linguocultural deritualization,” and “denotative literalism” when interpreting Kazakh ethnosomatic codes. The models exhibited phenomena of replacing sacred meanings with Western ethical norms and cognitive inertia. The value of the study lies in its being the first to examine the ethnocultural maladaptation of generative applications (ChatGPT, Gemini, Claude) from a cognitive-semantic perspective using Kazakh language material. The research findings possess practical significance for the specialized fine-tuning of AI models using the Kazakh language's ethnographic database and for improving machine translation quality.