This study investigates how computational linguistics terminology performs communicative and pragmatic functions in English and Uzbek scientific discourse. It adopts an exploratory qualitative comparative design and analyzes 21 English-Uzbek term entries representing core concepts in natural language processing, machine translation, corpus linguistics, and contemporary language-model research. The terms are classified according to their principal transfer strategy—international borrowing, calque or hybrid formation, and descriptive or competing translation—and are then interpreted in relation to precision, economy, audience accessibility, conceptual transparency, and compatibility with international research. The analysis shows that the contrast between the two languages is not adequately described as a simple opposition between concise English terms and explanatory Uzbek terms. Instead, Uzbek terminology operates through a layered system in which internationalisms such as token, parser, and transformer coexist with transparent formations such as tabiiy tilni qayta ishlash and so‘z turkumlarini belgilash. First-mention glossing, parallel use of English forms, and hybrid constructions function as pragmatic resources for managing differences in specialist knowledge. The paper argues that future standardization should be evidence-based, concept-oriented, and sensitive to actual disciplinary usage. It also proposes practical principles for building a stable Uzbek terminology resource for computational linguistics.