Design and Application of Automatic English Translation Grammar Error Detection System based on BERT Machine Vision

Abstract

Given the traditional handwritten English fonts, the accuracy of grammar error detection is unsatisfactory. This attribute leads to poor grammar error correction. Based on the optimized BERT machine vision model, an automatic English translation grammar error detection system is proposed in this paper. First, the basic architecture of the Transformer model and BERT model is considered, and a mixed attention module is discussed into the Transformer model to capture the features of the target in space and channel dimensions and realize the modeling of the context dependence of the target features. The feature maps are sampled by multiple parallel only if then cavity convolutions with different void rates to obtain multi-scale features and enhance local feature representation. Then, the input words of the BERT model are weighted by TFIDF to improve the feature extraction ability of the BERT model and construct the TF-BERT model. A database query rewriting model based on BERT and Transformer is proposed. The construction details of the model are described from the aspects of encoding processing, table embedding, and decoding processing respectively. Based on the principles of English translation, we extract grammatical features and build a grammar error detection method. TF-BERT model is selected as the basic framework. Combined with the hybrid attention mechanism, an automatic error correction model of English grammar is constructed. Finally, it is found that the loss value of the traditional system is as high as 0.7411, and the accuracy rate is 75%, while the loss value of the English grammar error detection system proposed in this paper is 0.2639, and the accuracy rate is 100%, which is 25% higher than that of the traditional system, and the performance is remarkable.