摘要
传统的跨语种交互翻译机器人语义纠错方法通常是单向的,效率较低,导致识别错误率较高。为此,文章提出基于语音信号的跨语种交互翻译机器人语义纠错方法。在基础语音识别的基础上,通过交互标定和特征提取来修正语义错误位置,并设计语音信号翻译机器人的语义纠错模型,采用随时间反向传播(Backpropagation Through Time,BPTT)循环训练核验方式,以确保纠错的准确性。测试结果显示,经过3个阶段测试,选定的5段语音材料的纠错识别率成功控制在10%以下,表明基于语音信号的跨语种交互翻译机器人语义纠错方法高效,具有实际应用价值。
The traditional semantic error correction methods of cross-language interactive translation robots are usually one-way,and the efficiency is low,resulting in a high recognition error rate.Therefore,this paper proposes a semantic error correction method for cross-language interactive translation robot based on speech signal.On the basis of basic speech recognition,the semantic error position is corrected by interactive calibration and feature extraction.The semantic error correction model of the speech signal translation robot is designed,and Backpropagation Through Time(BPTT)cyclic training is used to ensure the accuracy of error correction.The test results show that after three stages of testing,the error correction and recognition rate of the selected five speech materials is successfully controlled below 10%,which indicates that the semantic error correction method of the cross-language interactive translation robot based on speech signals is efficient and has practical application value.
作者
付曼
FU Man(Nanchang Institute of Technology,Nanchang Jiangxi 330099,China)
出处
《信息与电脑》
2024年第5期31-33,共3页
Information & Computer
关键词
语音信号
跨语种交互
交互翻译
机器人语义
语义纠错
纠错方法
speech signal
cross-language interaction
interactive translation
robot semantics
semantic error correction
error correction method