Construction Of Equipment Personalized Question Answering System Based On Web Data

With the arrival of the big data era,question answering systems,as a means to efficiently obtain effective information,have attracted widespread attention from all walks of life.In the field of military equipment,a large number of equipment types,models,and parameter information are very useful to command and combat personnel,but this information has not been effectively used.The main reason is that the number of query results given by the existing search engines is too large,and the relevant staff cannot quickly find the information they need.Aiming at the problems of traditional search engines in the field of equipment,this thesis builds a personalized question and answer(Q&A)system that fits users based on military equipment data.The system is constructed in a modular way,mainly including data acquisition module,personalization module,question and answer module,etc.Among them,the data acquisition module is mainly realized through the crawler function;the personalization module is realized by a variety of recommendation algorithms,which can be adjusted and optimized for specific application scenarios;The Q&A module constructs a variety of query methods,and uses a personalized module to optimize the Q&A results.The finally constructed personalized Q&A system solves the user's personalized query needs.The main work is as follows:First,constructed a data acquisition module based on crawler network.Based on the research of the real equipment header information and related military website equipment information,a set of asynchronous web crawler system based on the WebMagic crawler framework was constructed by using the combination of MongoDB database and MySQL database.,which can provide data support for all parts of the personalized Q&A system.Compared with the traditional way of obtaining data based on Http protocol,this data obtaining module has the characteristics of fast speed and high concurrency.Second,constructed a personalized module based on recommendation algorithm.This module is built on the basis of a recommendation algorithm based on memory,content and model,First,combined with the knowledge graph,a path-based method for finding reliable neighborhood users for new users is proposed,which effectively solves the problem of cold start for new users in the system;Then combined with the scoring prediction model,a method for predicting and scoring the initial recommendation list equipment is proposed,which solves the limitation of the lack of global evaluation indicators for the initial recommendation list generated by the traditional algorithm;Finally,when the dynamic recommendation list is generated,the user history weight factor,scoring characteristics and other measurement standards are added,which solves the limitation of the dynamic recommendation list generated by the traditional algorithm that is not personalized enough.The final experimental results show that the method proposed in this thesis has certain advantages over the traditional methods in various evaluation indicators.Third,constructed a question and answer module.By analyzing the actual needs of different user groups,two types of query methods: based on Q&A templates and based on keywords,are designed.In view of the limitations of the many answer results and insufficient personalization in traditional query methods,the personalized module is used to optimize the Q&A results.The experimental results show that the answers to the question by this answer module are relatively simple and sufficiently personalized.Fourth,built a personalized Q&A system based on the above three modules.The backend logic is based on Java's Spring Boot framework,and the frontend interface is based on html+css+js.The final Q&A system can not only provide users with natural language query services,but also provide personalized services for different users.