Design And Implementation Of Intelligent Question Answering System For Bridge Accidents Based On Knowledge Graph

In recent years,the development of transportation construction in China has been changing rapidly.The study of bridge accidents is of great significance for the development of transportation construction in China,but currently there are two main problems in the field of bridge accidents research.One is that the data types in the field of bridge accidents are heterogeneous from multiple sources,making it difficult to effectively integrate.Secondly,there is a lack of efficient retrieval platforms for data related to bridge accidents both domestically and internationally.This thesis conducts research on the above two issues.As a network like knowledge base,knowledge graph has rich nodes and relationships,and due to its network like knowledge structure,it can effectively integrate structured,semi-structured,and unstructured data.Therefore,constructing a knowledge graph in the field of bridge accidents can effectively solve the first problem.As the most widely used traditional retrieval platform,search engines can help people quickly obtain information,but the answers they return still require manual secondary screening to obtain the desired results.Intelligent question answering systems can be combined with emerging artificial intelligence technologies to better analyze user questions,thereby fully identifying user intentions and filtering answers,gradually becoming a new trend in human internet interaction.Therefore,building an intelligent Q&A system in the field of bridge accidents can effectively solve the second problem.In summary,this study combines knowledge graph with intelligent question answering systems.Firstly,the modeling of the core ontology in the field of bridge accidents was completed and a knowledge graph of the field of bridge accidents was constructed based on it.After confirming the entity and relationship attributes in the field of bridge accidents,the modeling of the core ontology was completed,followed by the construction of a knowledge graph in the field of bridge accidents.Then,a flowchart of the intelligent question and answer algorithm in the field of bridge accidents was proposed.This flowchart combines semantic parsing based methods with rule matching based methods.Specifically,the BERT-Bi LSTM-CRF combination model is used to complete the named entity recognition task,and entity mapping steps are added to improve the accuracy of entity recognition;The SBERT combination model was used to complete the task of relational attribute mapping,and attribute string matching steps were added to accelerate the overall Q&A process.Finally,using the popular B/S architecture,the knowledge graph and intelligent question answering algorithm were applied to the field of bridge accidents,and an intelligent question answering system in the field of bridge accidents was designed and implemented.It provides convenience for users to quickly and accurately obtain relevant data in the field of bridge accidents.Based on the above discussion,this thesis first introduces the relevant theories and technical analysis of this study.This includes theoretical foundations related to the construction of knowledge graphs,crawler technology,knowledge storage,as well as named entity recognition and relational attribute mapping technologies related to intelligent question answering algorithms.It also includes technologies such as the Flask framework used in the development of intelligent question answering systems and Echartxs components used in the visualization of knowledge graphs at the front end of the system.Then the process of building the domain knowledge knowledge map of bridge accidents is described in detail,including the core ontology research,knowledge storage and knowledge extraction of bridge accidents.Subsequently,the workflow of the intelligent question answering algorithm designed by our research institute was proposed.This study combines rule matching based methods with semantic parsing based methods.In terms of semantic parsing,two sub tasks have been identified: named entity recognition and relationship attribute mapping.Firstly,a named entity recognition algorithm based on the BERT-Bi LSTM-CRF model was designed for entity recognition.In the relationship attribute mapping section,first match the user’s intention based on string matching.If the matching fails,then use the SBERT model for similarity calculation,which can greatly shorten the system’s response time.In terms of rule matching,corresponding problem templates and Cypher statement query templates were constructed from the perspective of bridge management,and then the graph database was queried,ultimately returning the results.Based on the research of intelligent question answering algorithm,an intelligent question answering system based on the map of domain knowledge of bridge accidents is designed and implemented.The system is mainly divided into four modules:permission management,map display,map management and map query.This thesis focuses on the research of bridge accidents.First,the core ontology in the field of bridge accidents is studied.Then,a knowledge map based on the field of bridge accidents is constructed.Then,the related intelligent question answering algorithms are introduced.Finally,an intelligent question answering system based on the domain knowledge map of bridge accidents is constructed.This system can help users quickly and accurately retrieve the accident information in the field of bridge accidents,Injecting new vitality into the intelligent and digital development of the bridge industry.