针对翻译机器人在语义翻译过程中由于翻译误差容易导致翻译结果与原始语言意图不符的问题,提出一种基于Logistic模型的语义自动校准方法.通过语音识别模块将语音信号映射为语义文本,对识别的语义文本进行处理,通过改进的广义线性回归模型(GLR)进行误差检测,并基于Lo-gistic模型对翻译结果进行特征分析,预测流畅度以及准确度,实现语义自动校准.设计了针对翻译机器人语义自动校准的对比实验,实验结果表明,与基于Seq2Seq模型的翻译机器人语义自动校准方法相比,所研究方法语义校准的准确率为98%～100%,BLEU评分为35,语义校准时间为8.5～9.4 s.
Automatic semantic calibration method of translation robot based on Logistic model
In response to the issue that translation errors in translation robots can lead to discrepancies between the translated output and the original language intent,a semantic automatic calibration method based on the Logistic model is proposed.The process involves mapping speech signals to semantic text through a speech recognition module.The recognized semantic text is processed using an improved Generalized Linear Regression(GLR)model for error detection.Subsequently,feature analysis of the translation results is conducted based on the Logistic model to predict fluency and accuracy,thereby achieving semantic automatic calibration.A comparative experiment was designed specifically for semantic automatic calibration in translation robots.The experimental results demonstrate that,compared to the semantic automatic calibration method based on the Seq2Seq model,the proposed method achieves a semantic calibration accuracy rate of 98% to 100%,a BLEU score of 35,and a semantic calibration time ranging from 8.5 to 9.4 seconds.
semantic calibrationgeneralized linear regressionLogistic modelone-dimensional mappingsimilarity
NETL
NSTL
万方数据
