FROM STARTUP GLOSSARY TO VERIFIED CORPUS EVIDENCE: A TRANSPARENT AND AUDITABLE WORKFLOW FOR STUDYING ENGLISH-DERIVED STARTUP TERMINOLOGY IN UZBEK
Authors/Creators
Description
This article sets out a transparent and auditable workflow for studying rapidly changing English-derived startup terminology in Uzbek. The workflow was developed for a doctoral project based on a specialised corpus of Uzbek startup discourse comprising 1,011 texts, 1,133,478 tokens, 67,355 lexical types and 65 source domains. Its organising principle is the separation of three evidence layers: a broad 1,524-unit English–Uzbek reference glossary, a purposively selected and manually verified analytical core of 100 concept-level naming families, and a separately analysed institutional layer. Six stages implement that principle: defining the conceptual field before counting, normalising and versioning the corpus, constructing concept-level naming families, verifying keyword-in-context concordances manually, calculating seven corpus-internal measures, and analysing institutional evidence as a distinct layer. Together these stages address four recurring risks in terminology research: treating an automatic string match as a verified attestation, treating corpus frequency as dispersion, treating orthographic adaptation as grammatical integration, and treating official use as formal codification. The article distinguishes procedural re-runnability, which the stated rules support, from full computational reproduction of the reported figures, which would additionally require the frozen corpus version, the complete family definitions and the matching scripts; that boundary is stated explicitly rather than assumed. The workflow is offered as a transferable model for other fast-moving specialised domains, particularly in agglutinative languages where exact-form counting fragments the evidence across inflected variants.