机器翻译中的低资源知识模型建构研究  

Low-resource Knowledge Model Construction for Improving Machine Translation Quality

作　　者：戴光荣 黄栋樑 DAI Guangrong;HUANG Dongliang(School of Interpreting and Translation Studies,Guangdong University of Foreign Studies,Guangzhou 510420,China)

机构地区：[1]广东外语外贸大学,广东省广州市510420

出　　处：《外语教学与研究》2025年第1期106-117,共12页Foreign Language Teaching and Research

基　　金：国家社科基金项目“神经网络机器翻译质量提升研究”(22BYY042)的阶段性成果。

摘　　要：低资源语言翻译质量低下一直是数字化时代机器翻译研究领域的痛点。目前,相关算法、模型在高资源语言翻译中已取得显著成绩,甚至超过人类表现;而与低资源语言相关的语言模型内部、外部知识不一致,导致机器的文本阅读能力、理解能力和翻译能力表现不佳。本文分析低资源语言中的知识本体和知识内涵,探讨低资源知识对于提高低资源语言机器翻译质量的效用。在此基础上,本文以低资源知识谱系、知识群和知识库为基准,搭建并完善低资源知识模型,进一步平衡机器翻译中低资源语言模型所需的内外部知识,进而提升低资源语言机器翻译的译文质量。In the digital era,poor machine translation quality of low-resource languages(LRL)has consistently been a significant challenge in the field of machine translation research.Currently,related works on algorithm models in high-resource languages have made remarkable achievements and even surpassed human performances.However,the external and internal knowledge systems of LRL models are so inconsistent that the performance of machine reading,understanding,and translation remains low.This study interprets the ontology and connotation of knowledge in LRL and discusses the function of low-resource knowledge to improve the machine translation quality.Based on this analysis,the Low-resource Knowledge Model is constructed using knowledge genealogy,knowledge groups,and knowledge database as its foundations,aiming to balance the external and internal knowledge required for LRL models in machine translation and improve the translation quality of LRL.

关 键 词：机器翻译质量 低资源语言 低资源知识 低资源知识模型 

分 类 号：H059[语言文字—语言学] TP312[自动化与计算机技术—计算机软件与理论]