A neural network-based quality assessment model for English-to-Chinese text translation

Addressing the urgent need for cross-language text translation quality assessment, this paper proposes a neural network-based model for evaluating English-Chinese translation quality.Current widely adopted automated evaluation methods exhibit significant limitations in handling specialised terminology and nuanced semantics, particularly when addressing culture-specific concepts.The neural network model constructed in this study integrates deep semantic representation with contextual correlation analysis, achieving remarkable results on the Chinese-English test set of the public WMT 2020 metrics shared task dataset.It achieved a core correlation metric (Pearson's r) of 0.682, along with a multi-dimensional classification evaluation (macro-F1) of 0.689 and a ranking quality metric (normalised discounted cumulative gain @10) of 0.927, comprehensively outperforming mainstream baseline models.This model provides a reliable technical tool for cross-language text quality control.