Construction Of Knowledge Graph And Research On Disaster Information Dissemination Are Conducted For The Urban Comprehensive Pipe Gallery

The urban underground comprehensive pipe gallery is a type of public municipal infrastructure that can solve the difficulties of refined construction and management in the city,such as "road zipper" and "air spider web".However,during the maintenance and operation stage,the comprehensive pipe gallery has many safety risks and sources.When emergencies occur in the urban underground comprehensive pipe gallery,visual analysis based on a knowledge graph of dangerous fault sources can greatly improve the decision-making ability of emergency management personnel.At the same time,sudden disaster events in the urban underground comprehensive pipe gallery are usually difficult to predict and have strong destructive power.Based on the characteristics of the concealed,complex,and chain-like disasters in the comprehensive pipe gallery,how to achieve efficient and accurate dissemination of disaster information is an important issue for social security.A reasonable information dissemination model can improve event transmission efficiency and transmission accuracy,effectively helping people to avoid disasters and helping decision-makers to quickly take response measures.This paper mainly focuses on the construction of knowledge graph in the field of comprehensive pipe gallery and the dissemination of disaster information.(1)Aiming at the problem of insufficient semantic information and overlapping entity relationships,a relational attention mechanism-based joint extraction model called RAJE(Joint extraction model based on relational attention mechanism)is proposed for knowledge extraction in the field of urban underground integrated pipe gallery.RAJE integrates active learning,data augmentation,BERT(Bidirectional Encoder Representations from Transformers),relation-specific attention networks,and sequence labeling schemes to extract all entities and relationships simultaneously.The effectiveness of the model in solving nested entity and relationship problems is validated through experiments.Firstly,a value-based sampling and annotation method based on active learning is used to select more valuable training data,and data augmentation is applied to the annotated dataset.Then,an end-to-end joint extraction method is used for entity and relationship extraction tasks.The joint extraction method includes a BERT model for obtaining sentence representations,and a relation-specific attention network is used to capture relation-based sentences.Finally,sequence labeling is performed on the sentence representations to obtain entity pairs.The proposed model is evaluated using a collected dataset of integrated pipe gallery hazard and engineering data and a public dataset.(2)The knowledge storage and visualization based on Neo4 j are implemented.Utilizing the knowledge storage and visualization functions of the Neo4 j graph database,the extracted triple information is stored in the database.The complete knowledge graph of danger and fault sources is constructed based on the setting of entity nodes and relationship type edges in the field of urban underground comprehensive pipe galleries.Through querying,the danger and fault sources and related node information can be quickly and accurately located.The effectiveness of the knowledge graph in the field of urban underground comprehensive pipe galleries is verified through case studies.(3)We propose a social sensor network(SSNs)-based information propagation model for disaster events.In the event of a disaster in the comprehensive pipe gallery,it is necessary to quickly disseminate information about the disaster.Based on the inherent autonomy of human individuals,we propose a social sensor network information propagation model that studies the impact of individual characteristics,social characteristics,and group information propagation patterns on social sensor networks.Specifically,we first construct a human sensor model based on the inherent social and psychological attributes of human autonomy.Then,by considering different transmission media and human interaction preferences,we propose various information propagation models such as one-to-one,one-to-many,and peer-to-peer.The Net Logo platform is used to simulate the information propagation environment in disaster events.The performance of the social sensor information propagation model is evaluated using an application evaluation matrix.