Construction And Application Of Earthquake Disaster Knowledge Graph Based On Combined Model

Earthquakes occur frequently in worldwide,directly causing very serious casualties and economic losses,scholars and technicians around the world have made a lot of research on earthquake disasters,such as the causes of earthquakes,risk analysis,emergency response capabilities,loss assessment,prevention and early warning.With the development of network technology,knowledge and data related to earthquake disasters are distributed in every corner of the network world,while people want to query these information,they cannot quickly obtain the required accurate answers,which is not conducive to people’s research and understanding of earthquake disasters,nor is it convenient to share earthquake related information.This thesis constructs a knowledge graph of earthquake disasters based on these issues,then designs and implements a knowledge retrieval system for earthquake disasters to promote the sharing of information related to earthquake disasters and assist in the prevention and control of earthquake disasters.The main work of this thesis is as follows:(1)A combined Chinese word segmentation model based on HMM and BiGRUCRF is proposed.In order to extract more accurate and high-quality earthquake disaster information,it is necessary to improve the accuracy of information segmentation.In this thesis,the basic HMM(hidden Markov model)word segmentation model and BiGRUCRF word segmentation model are combined to improve the effect.The existing corpus is used to validate the above models,and the results show that the combined model improved word segmentation accuracy by 1.23%,1.83%,and 1.59% on three datasets.(2)A combined POS(part of speech)tagging model based on HMM and BERTBiLSTM-CRF based on attention mechanism assisted classification layer is proposed.In the process of knowledge extraction for earthquake disaster information,this thesis improves the entity extraction method for earthquake disaster data,adds an auxiliary classification layer based on attention mechanism to the BERT-BiLSTM-CRF entity extraction model,and then combines it with the HMM model.The HMM model marks the simple part of speech in the sentence to be segmented,and the BERT-BiLSTM-CRF model based on the auxiliary classification layer of attention mechanism marks the complex part,Then perform regular expression extraction on the annotation results.The model was validated using corpus,and the results showed that the combined model improved its accuracy by 0.22%,0.36%,and 1.11% on three datasets,respectively,and also showed a 26% improvement in work efficiency.(3)The knowledge graph of earthquake disasters is constructed,and based on this,a retrieval system was designed and implemented.First,a large number of data about earthquake disaster are obtained from the network and other channels,and they are sorted into a corpus of earthquake disaster.Then,the combination model proposed in this paper is used for knowledge extraction of earthquake disaster information,and the extracted entities and relationships are stored in the Neo4 j graph database in the form of triplets according to the defined types.Subsequently,a retrieval model was constructed based on the knowledge graph,and a knowledge retrieval system was designed and implemented.