基于神经网络的智能外语翻译机器人语义纠错系统  

Semantic error correction system of intelligent foreign language translation robot based on Neural Network

作　　者：李星[1] LI Xing(Xianyang Normal University,Xianyang Shaanxi 712000,China)

机构地区：[1]咸阳师范学院,陕西咸阳712000

出　　处：《自动化与仪器仪表》2023年第1期246-250,共5页Automation & Instrumentation

基　　金：咸阳师范学院“青年骨干教师”培养项目(XSYGG201904);咸阳师范学院科研计划项目《村上春树文学中的中国“事”与“情”》(XSYK21002)。

摘　　要：针对传统语法错误纠正系统存在并行化程度低的问题,提出以神经语法错误纠正基线模型为基础模型,在基于循环神经网络编码器-解码器基础上对其进行改进,并构建一个基于自注意力机制的语法错误纠正模型—Transformer,通过此模型提升语法纠正效果。实验结果表明,Transformer模型在对冠词、名词、介词、形容词等语法错误进行纠错时,其纠错召回率明显高于传统的MLConv模型,且本模型的计算并行化程度更高。由此说明,基于自注意力机制的语法错误纠正模型性能更为优越,构建的Transformer系统在语法错误纠正中具有可行性。In view of the problem of low parallelization of the traditional syntactic error correction system,the proposed neural syntactic error correction baseline model is based on the recurrent neural network encoder-decoder,and builds a syntactic error correction model based on the self-attention mechanism,Transformer,through which to improve the grammar correction effect.Experimental results show that the error-correction recall of Transformer model is significantly higher than the traditional MLConv model,when correcting grammatical errors like words,nouns,prepositions and adjectives,and the computational parallelization of this model is higher.This shows that the grammar error correction model based on the self-attention mechanism is superior,and the constructed Transformer system is feasible in the grammar error correction.

关 键 词：语法错误纠正 循环神经网络 编码器-解码器 Transformer系统 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]