Design of an English Document Automatic Summary and Translation System Combining BERT and Tree Long Short Term Memory Network

This paper investigates the optimization of a text digest and translation system based on the fusion of two advanced natural language processing models, BERT and Tree-LSTM. Traditional text processing methods are faced with the problems of imprecise information extraction, inadequate structural information processing and poor translation quality. This project will make full use of the advantages of BERT in data preprocessing and short-term memory learning to improve the automation of abstracts. BERT language is used to encode and interpret English documents to form abstracts containing the main body of the documents, and tree-like LSTM is used to analyze the text and extract the internal relations and logical structure among the texts so as to improve the accuracy of abstracts. The experimental results show that the translation quality of English texts is over 0.95 when BERT is combined with tree-like short-term memory network. By combining BERT with Tree-LSTM, the effective use of text information is realized.