轻量型的计算机辅助翻译系统设计

Design of a lightweight computer-aided translation system

针对翻译系统运行资源消耗多、响应慢、翻译准确率低等问题,设计一个轻量型的计算机辅助翻译系统。基于Transformer模型和混合迁移模型,提出将相似多语言和领域融合的混合迁移学习模型相结合,得到相似多语言领域融合的混合迁移学习模型。将融合模型应用到基于B/S构架的低资源翻译系统中,以提升语言翻译质量,降低系统开销。实验结果表明,相较于单一的相似多语言和领域融合模型,本模型的BLEU值分别提升了4.4和2.1,模型泛化能力和翻译质量显著提升。实际应用发现,系统可在短时间内实现快速翻译,可实现低资源多语料的准确翻译,具备一定的有效性。

A lightweight computer-aided translation system is designed for the problem of translation system, low translation accuracy.Based on Transformer model and mixed migration model, the hybrid transfer learning model of similar multilingual and domain fusion is proposed to obtain a hybrid transfer learning model with similar multilingual domain fusion.The fusion model is applied to the low resource translation system based on B/S architecture to improve the quality of low resource language translation.The experimental results show that compared with a single similar multilingual and domain fusion model, the BLEU value of this model was improved by 4.4 and 2.1, respectively, and the model generalization ability and translation quality were significantly improved.Practical application finds that the system can achieve rapid translation in a short time, and realize the accurate translation of low-resource multi-corpus, and have a certain effectiveness.