Machine Vision-based English Translation Grammar Error Detection Using Temporal Convolutional Network-Recurrent Neural Network

Machine vision-based English translation grammar error detection leverages advanced methods to evaluate and correct errors in translated text. It employs computer vision methods to interpret and increase textual content which enhances translation accuracy and fluency. However, it struggles with context understanding and subtle grammatical nuances because of limited semantic analysis abilities which leads to low accuracy. This research proposes Temporal Convolution Network-Recurrent Neural Network (TCN-RNN) to detect the grammar error in translated text. Initially, the Lang 8 dataset is used to determine the TCN-RNN performance. The stemming and tokenization are applied to remove the unwanted words in translated text. Then, the word2vec approach is performed to extract the pre-processed features effectively. At last, TCN-RNN detect the errors accurately which combines TCN's long-range dependencies whereas RNN manages sequential data which leads to enhanced accuracy in error detection. The proposed TCN-RNN achieves a better accuracy of 97.34% compared to existing methods like TCN-Convolution Neural Network (TCN-CNN) and TCN-Long-Short-Term Memory (TCNN-LSTM) respectively.