Research On Intelligent Question-answering System Based On Meteorological Science Knowledge Grap

Meteorology is closely related to human life,and having knowledge of the weather is of great significance to our well-being.With the advent of the big data era,several search engines compete to satisfy users’ needs for information retrieval.However,users can only search for answers using keywords and have to filter answers from a large number of search results,which makes it difficult to meet users’ needs for quick and accurate access to information.As a potential solution to this problem,question-and-answer systems have been heavily researched.The development of domain knowledge graphs provides a high-quality knowledge base for domain Q&A systems.In this paper,we focus on the knowledge graph and intelligent Q&A system in the field of meteorological science and technology.This paper combines computer deep learning algorithms to develop an intelligent Q&A robot system based on meteorological scientific knowledge mapping,drawing on the National Natural Science Foundation of China project.The primary research for this paper includes the following aspects:Building a knowledge graph in the field of meteorological science and technology is the first step.There is no publicly accessible knowledge map for the promotion of the field of meteorological science.This paper analyzes the common core ontology concepts in the field of meteorological science popularization by comparing the depth of professional knowledge to the breadth of public understanding with the aim of applying intelligent Q&A in the field of meteorological science popularization.Additionally,a knowledge graph in the field of meteorological science popularization is built through the processes of data collection and knowledge storage,which provides a reliable data base for the intelligent Q&A system.Secondly,the main knowledge graph-based intelligent question-and-answer algorithms are studied.This paper proposes an entity recognition algorithm,MGTNER,for the named entity recognition job.MGTNER gathers semantic feature information from the dataset at different granularities and produces good entity recognition results.Then the AC algorithm and similarity algorithm are used for entity linking.For the intention recognition task,relational attribute linking is performed by an improved BERT twin network-based algorithm.The model in this paper is compared with other existing models to verify its effectiveness on the relevant tasks.Intelligent question-and-answer systems are given a strong algorithmic base.Finally,we designed and implemented an intelligent meteorological robot "Xiaoxue".This paper designs a Q&A robot system that can interact with voice because conventional Q&A systems can only deal with text,which limits their use and the population.The software includes an MFC-based human-computer interaction module and a Flask framework-based backend web module.To achieve effective human-computer interaction,the robot system uses KDDI voice technology to convert user voice information into text information,combines deep learning algorithms to understand user questions,and converts text responses into voice information before delivering them to users.