Research on the Construction of a Chinese–English–Uzbek Agricultural Terminology Database

Abstract

The widespread terminological asymmetry in China–Uzbekistan agricultural cooperation raises the costs of qualification verification, technical training, and contract negotiation. This paper explains the construction pathway for a Chinese–English–Uzbek agricultural terminology database: using a Chinese–English agricultural dictionary as the base text, 15,158 trilingual entries were empirically annotated, and 275 newly added terms were supplemented, forming a dual-source corpus of “stock resources plus incremental resources.” The terminology database uses entry numbers as the primary key and contains eight fields: Chinese headword, Pinyin annotation, English equivalent, Uzbek translation, subject classification, original page number, status marker, and cross-reference entry. It supports three-way retrieval in Chinese, English, and Uzbek. The database construction process is divided into six stages: layout parsing, column-by-column transcription, main-table alignment, discrepancy checking, translation-name generation, and output release. Artificial intelligence significantly reduces manual costs in transcription, batch comparison, and preliminary translation drafting. Quality control relies on a four-stage process of “machine drafting—rule constraints—expert review—version consolidation,” with emphasis on preventing four types of risks: misrecognition of visually similar characters, coexistence of multiple translations, semantic drift, and instability in the translation of policy terminology. This paper summarizes four roles and six practices of artificial intelligence in terminology construction and compares the applicable boundaries of general large language models, execution-oriented agents, and terminology-specific systems.