Terminology-enhanced Neural Machine Translation in Field of Electrical Engineering
ZHANG Juwei
LIU Kaiwen
CHEN Yuan
Abstract：Significant progress has been made in neural machine translation models for general domains,but challenges remain in specialized fields,especially under low-resource conditions.These challenges are particularly evident in the effective utilization of terminology.In response to the limited utilization of terminology information in English-Chinese translation models for the electrical engineering field under low-resource conditions,terminology vocabulary was incorporated as prior knowledge.Terminology information is extracted using the original structure of the Transformer model.Based on this,the information is integrated into the encoder's top-layer output through copy-convolution and doubling methods.This approach enhances the translation performance of the Transformer model with minimal addition of learnable parameters.Experimental results show a 1.11％improvement in BLEU score,with a 6.86％increase in training time,balancing translation quality with computational cost.This approach provides a new perspective on fully leveraging terminology information in low-resource settings.
Keywords：neural machine translationterminology informationelectrical engineeringconvolutionlow resource
Publication Date：2024-12-25
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：11( 43-53 )