CORPUS AS A MIRROR OF COLLECTIVE COGNITIVE EXPERIENCE IN THE CONTEXT OF LANGUAGE MODELS DEVELOPMENT (2020–2025)

Abstract

In the context of developing large language models (LLMs) between 2020 and 2025, the article provides a thorough examination of the evolution of corpus linguistics in Ukraine through the lens of the linguo-cognitive dimension. The research investigates the development of of natural language processing tools, linguistic annotation techniques, and Ukrainian corpus resources that serve as the foundation for training contemporary language models. The emergence of domestic artificial intelligence architectures, particularly MamayLM and Lapa LLM, and the dynamic transformation of corpus data are highlighted as ways that they reflect the national communicative space and contribute to the development of a digital linguistic worldview of Ukrainian society. The idea that the linguistic corpus is an objective “mirror” of collective cognitive experience that accumulates mental structures is grounded by the author. The article identifies key linguistic challenges of the Ukrainian language for natural language processing systems, including high inflectional complexity, morphological variability, and flexible syntax. These features show up as manifestations of particular cognitive mechanisms of speech organization rather than just as technical challenges for NLP models. In order to help language models better understand the context and pragmatics of Ukrainian speech, the study examines contemporary corpus labelling techniques, such as multi-level annotation of grammatical, semantic, and emotional text parameters. It is argued that the creation of national corpora and large language models is an essential factor in ensuring the cognitive security and digital sovereignty of Ukraine, as it facilitates the representation of unique cultural and linguistic experiences in the global information space and forms the foundation for the further development of Ukrainian artificial intelligence technologies.