Knowledge Graph Construction Based On Earth System Science Data(ESSD) Journal Articles

The problems of massive data clustering,untimely services,and ungrounded applications are prominent with the development of digital era,which seriously restrict the application potential of data.Based on the innovation of remote sensing satellite technology and the rapid development of Earth observation capability,Earth system science data have exponentially exploded in growth.And the improvement of observation data resolution and the development of other science and technology have provided more data products for observation data.Therefore,the current Earth system science data have very significant characteristics of large data volume,complex spatialtemporal relationships,diverse features,and rich semantics.At present,the Earth system science data services based on remote sensing images are mainly through database technology and metadata-based query and retrieval methods,which lack the ability to describe and understand semantic information and make it difficult to accurately discover data that meet users’ needs,also,the huge amount of data information also increases the difficulty for users to collect comprehensive data.Therefore,this study forms a knowledge network of Earth System Science Data(ESSD)by building a knowledge graph of the datasets proposed by articles in the ESSD journal to realize intelligent querying of data and mining the features of the datasets proposed by articles in the journal.The research proposes an Attention-based Probability Calculation model(APNER)which could use for entity recognition tasks.The model acquires the semantic representation of words using the encoding part of the BART pre-training model,learns the word-to-word association within the sentence based on the self-attention mechanism,and performs similarity calculation with the output of the decoding part of the BART pre-training model to predict the sequence of entities and entity labels.The model outperforms baseline models in the experiments in extracting entities from the abstracts of ESSD journal articles.After entity identification,this study generates entity nodes and relational edges of the knowledge graph based on the ontology design of the knowledge graph and presents the visualization results of the knowledge graph on the Neo4 j platform.Finally,the study conducted two case mining based on the constructed knowledge graph.The first case analyzed the research regions of the datasets proposed by the ESSD journal article,and mined the hot research regions of the journal article as well as the data situation and features of a certain research region;the second case mined the carbon-related research to view the popular topics of carbon research,the current research status and the associated datasets of different research topics.In conclusion,the study completes the construction of a knowledge graph based on journal article information and demonstrates the significance of the Earth system science data knowledge graph for data query and mining of research features.