Aided Reading System Based On BERT Model

With the development of society,information shows explosive growth.Meanwhile,due to the annual rise of human cost,the ability to quickly obtain text information from a large number of texts and help summarize texts has become an urgent demand for relevant staff.With the continuous development of deep learning technology,natural language processing technology has also been constantly developing,especially the use of natural language processing abstract and text rewriting has become a research focus.This paper designed and implemented a bert-based assisted reading system,which can effectively assist users to obtain important information of the original text and generate a summary example.This paper makes a detailed analysis to the actual requirements.There are many challenges in the implementation of the system.This paper first uses a more fine-grained sentence splitting to make the minimum unit of the abstract lower,so that the results are more accurate.This paper use BERT as the text representation model.Firstly,the sentence representation is generated by using fine-tuned BERT,and an improved CNN--DGCNN is used to judge whether it is a summary sentence.NEZHA is then used to rewrite the summary as a rewrite summary of the source text;In this paper,the hybrid strategy of Elasticsearch+ model fuzzy generation is used as the strategy of fuzzy retrieval.Finally,Spring Boot,Mysql,Elasticsearch and other development technologies are used to complete the development of the system.The system is divided into five core modules.The abstract generation module is used to obtain abstracts of text uploaded by users.Data management module for management and maintenance of background data;The text rewriting module is used to rewrite the summary into a coherent summary.Automatic label system is used to label text intelligently for easy viewing by users;The data retrieval module is responsible for mixing keywords and returning the most suitable search results to the front end.This system can help users quickly get the core sentence of the target text and give the corresponding tag of the text,and give a rewriting summary.Reduce the burden of reading a large number of articles and secondary creation.The system constructs the front-end interface in the form of web pages,which is suitable for the actual use scenarios of users and provides a humanized interactive scheme,which can significantly improve the user’s work efficiency.