Design And Implementation Of A Cross-lingual Text Summary System Based On Deep Learning

With the rapid improvement of computing power and the rapid development of artificial intelligence technology,natural language processing(NLP),as an important research field,has a high research value and a wide range of application scenarios,and has become a hot research topic.The purpose of this paper is to use the deep neural network technology,through the translation of the long text from the foreign language to the target language,and then the intelligent extraction of the key content of the long text,so as to extract the key parts of the foreign language,and achieve the purpose of quickly reading the core content of different language articles and obtaining the key information.The purpose of this paper is to design and implement a cross language text summarize system based on deep learning.By combining machine translation with text summarize,and modifying the deep neural network architecture,the main work is as follows:(1)According to the Bert model,the pre-training model is introduced.Before entering the machine translation module and the text summary module,the corpus is trained in advance to improve the accuracy of the output results;(2)In the machine translation module,we use the transformer model,and modify the self attention to adaptive attention,so that the model can automatically modify the scope of concern;(3)In the text summary module,the key information is extracted by referring to the HSSAS model,and the Bi LSTM framework is modified to be the Bi-GRU with simpler network structure and similar effect.Machine translation technology is used to translate the target text accurately,and then text summarize technology is used to extract the core content of the text to shorten the text content.The input is the long corpus of the source language and the output is the phrasal material of the target language.Finally,the optimization effect of the pre-training model is tested through the comparative experiment before and after the optimization of the design model;the different attention architectures in the machine translation model are compared and analyzed to explore the differences of No-attention,Attention,Self-Attention and Adaptive-Attention for the experimental results;the Bi-LSTM structure and Bi-GRU structure in the text summary model are compared to study the replacement model Optimize the effect.Through the pre training module,the system trains a large number of Chinese English parallel corpora,so that the accuracy of the model can be guaranteed,and the translation from English to Chinese can be carried out accurately,and then extract the key information of the translation content by using the extract text summary technology,so as to extract the core content of the article accurately.Thus,the original intention of the system-cross language text summarize is realized.