Design And Implementation Of Intelligent Question Answering System Based On Document Compiling And Checking

In recent years,document writing and inspection is a very important link in today’s information society.The requirements for the standardization of document content and format have become higher and higher.In particular,governments and schools need to complete a lot of standard document writing work.As the document writers do not have a thorough understanding of the document preparation requirements during the writing process,they need to consult the document writing requirements to complete the document writing,and the efficiency of document writing is low.Therefore,this article aims at the field of document preparation,and realizes the functions of intelligent question and answer and checking the standardization of documents and replying to corresponding suggestions.Based on the requirements of informatization project writing,this paper has established an informatization project document writing knowledge base,and designed an intelligent question answering system based on document writing and checking.The system can give standardized answers to the problems encountered by users in the process of document writing through user questions,and at the same time respond to the scoring status of the submitted documents,as well as return solutions to existing problems.It greatly saves labor costs and improves the efficiency of document writing.The main results of this paper are as follows:(1)Based on the design rules of government documents,this paper derives the data model of semantic triples in unstructured text and builds a knowledge base that serves to standardize the compilation of government documents.(2)In text processing,taking the difficulty of generating embedded representations for long text of documents in application scenarios into account,this paper improves the BERT model by using parameter reduction technology and cross-layer parameter sharing technology based on the characteristics of the knowledge base text of the intelligent question answering system.At the same time,the attention mechanism is added to fuse the semantic information of the full text of the document,which enhances the semantic representation of the target text.This paper proposes to use Faiss search engine to store vectorized text with index information,use index number to match answer ID number in database,and recall high similarity answers from users.(3)Based on the BERT and Faiss algorithm,this paper implements an intelligent question and answer system for government personnel.The system is mainly implemented using the Flask Web technology framework.It has the function of preparing questions and answers on documents with standardized content,checking the document standardization and responding Write suggested features.