English Assisted Teaching System Based Stacked Convolutional Neural Network and Bidirectional Long Short-term Memory

English assisted teaching system are widely used around world and plays an essential role in global communication. However, English learning is a difficult, long process for English assisted teaching system inevitably make various unwanted sentence and writing errors. This paper, proposed Stacked Convolutional Neural Network (SCNN) and Bidirectional Long Short-term Memory (Bi-LSTM) network is performed in the classification to achieve high accuracy. pre-processing techniques such as converting text are applied to the lowercase, while stemming is included to improve data quality in the training process. The feature extraction using Bidirectional Encoder Representations from Transformers (BERT) enhance educational tools by offering more interactive and engaging learning experiences. It also helps provide more neutral and balanced language suggestion, reducing the change of bias compared to traditional system. The proposed SCNN and Bi-LSTM techniques are evaluated using the raw dataset, achieving a higher accuracy of 91.23%, precision of 91.52% and F1-score of 91.75% and recall of 91.63%, respectively. The existing techniques, Recurrent Neural Network (RNN) and BERT are used to evaluate the proposed method.