Investigating the predictive power of lexical, syntactic, and cohesive features on Chinese-English translation quality: The case of Test for English Majors grade 8 (TEM8)

Abstract

This study investigates the language features that may predict translation quality in TEM8 Chinese-English tasks and the magnitude of their predictive power. Drawing upon a self-compiled corpus comprising of 1,000 translation samples from 2022 and 2023 test administrations, 38 language feature indices spanning lexical, syntactic, and cohesive dimensions were extracted using Coh-Metrix (Graesser et al., 2004) and the NeoSCA (a refined iteration of the Second Language Syntactic Complexity Analyzer, namely L2SCA) (Tan, 2022). Quantitative analysis identified seven significant language feature predictors: four lexical and three syntactic. Notably, no cohesive features demonstrated statistically significant predictive capacity. Subsequent regression analysis revealed that five of these predictors collectively explained 17.1% of the variance in translation quality scores. These findings furnish empirical evidence for translation assessment and pedagogy, offering actionable insights for candidates and teachers to pinpoint specific linguistics domains warranting enhancement in TEM8 translation performance.