Study On The Construction Of Knowledge Graph On Debris Flow Prevention Domain And Its Application In Q&A

Debris flow is one of the three typical geological disasters.Debris flow is often accompanied by a landslide,flood,collapse and other types of disaster chains.Debris flow erupts frequently,seriously threatening people’s life and property safety.The improvement of the awareness of debris flow disaster prevention and the improvement of intelligent information service of debris flow prevention are of great significance to the work of disaster prevention and reduction.At present,the relevant studies of debris flow are more and more abundant.However,there is a problem that the relevant data and models of debris flow disaster prevention increase rapidly while the cognitive knowledge of debris flow disaster prevention is not strongly related.Therefore,an effective data management tool is urgently needed to link the knowledge of debris flow prevention and promote the transformation of debris flow data-information-knowledge.As a graph model,knowledge graph can connect the semantic relation between entities well.In view of the above problems,this study builds knowledge graph of debris flow prevention based on the existing knowledge of debris flow prevention and develops a question-answer system based on it,so as to realize intelligent information service of debris flow disaster prevention and reduction.The main research work and achievements of this paper are as follows:(1)Construction of knowledge graph on debris flow prevention domain.The construction of debris flow prevention knowledge graph mainly includes the construction of model layer and data layer.By building a corpus of debris flow control knowledge to analyze the characteristics of the debris flow control knowledge corpus,and to dissect the semantic relationships between debris flow theoretical knowledge,disaster events,modeling methods,and disaster prevention and mitigation strategies.The ontological thought is used to conduct a unified normative description of each element from the top down,establish a comprehensive ontology library in the domain of debris flow prevention,and complete the construction of a model layer of debris flow prevention knowledge graph.Under the guidance of the model layer framework,the data layer of debris flow prevention knowledge graph was constructed from the bottom up.The data layer focuses on entity extraction and relation extraction,and proposes entity extraction method based on ALBERT-Bi LSTM-CRF and relation extraction method based on number of predicates.In addition,knowledge fusion and knowledge storage are further completed on the technology of knowledge extraction.Finally,a relatively complete knowledge graph of debris flow prevention is formed.(2)Construction of question and answer model based on knowledge graph of debris flow prevention.To address the problem of the insufficient number of debris flow prevention question sets,this paper uses the method of question enhancement to enlarge the question set and obtain the question corpus.The entity extraction method based on ALBERT-Bi LSTM-CRF is selected to complete the entity identification of debris flow prevention questions.According to the text features of the questions on debris flow prevention,four kinds of intentions are divided into debris flow theory knowledge,model method,disaster event and prevention strategy,and the subintentions of each kind of intention are integrated,and 13 kinds of user query categories are determined.The BERT method is used to judge user intention,and comparative experiments are carried out to analyze the performance of various classifiers.The experimental results show that the F1 value of BERT intent classification experiment is about 85%,which is the best effect.(3)Design and implementation of a question-and-answer system based on knowledge graph of debris flow prevention.This paper designs and develops a question and answer system based on the knowledge graph of debris flow prevention.Firstly,demand analysis is carried out on the question and answer system of debris flow prevention knowledge graph.Secondly,the function and overall architecture of the system are designed.The system functions mainly include intelligent question and answer,knowledge query of debris flow prevention,entity query,relationship query and so on.Finally,the system uses Python language and Django framework to complete the development of the system,users can use this system to achieve knowledge of debris flow prevention questions and answers.