A Causative-specific Error Taxonomy for Uzbek–english Machine Translation

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

A translation score alone does not tell us whether causative meaning has survived. This article presents a procedure for evaluating Uzbek causative verbs translated into English. It combines an operational definition and inclusion rules with a targeted test set, two layers of error codes, and blind assessment against a short statement of the required meaning. Each target receives one causative code (C0–C7); general errors elsewhere in the sentence are recorded separately (G1–G7). Coding 112 outputs from four systems exposed several weaknesses in the procedure. Sixteen decisions (14%) were ties, all in sentences with uncertain class labels. Ten of the 27 sentences fitted none of the predefined classes. A reference set drafted after the outputs had been collected also closely resembled one system’s translations. These findings guide the proposed revisions.