Research On Knowledge Gragh Construction Method For Coastal Wetland Ecological Environment Management

On the Internet,a large amount of wetland knowledge exists in semi-structured and unstructured texts,from which important information is often hidden.With the rapid growth of information resources on platforms,it takes more and more time and effort to find the information you need from the vast amount of text.In this context,learning how to extract wetland information from unstructured text is an essential part of improving the efficiency of wetland ecological management.The knowledge graph is a means of visualizing data.coastal wetland knowledge graph can visually display the distribution status,species characteristics,and scientific research status of each species in coastal wetlands.In this paper,we studied the construction method of the coastal wetland knowledge graph and the main work includes the following aspects:(1)Methodology for ontology construction of coastal wetlands.In this paper,we design a coastal wetland ontology construction method.The method has two main features: ⅰ)we conducted a keyword co-occurrence analysis of the literature in the field of coastal wetlands using Cite Space to identify the scope of the coastal wetlands field ontology;and ⅱ)an ontology structure containing temporal information was proposed to more comprehensively represent coastal wetland knowledge by introducing nodes that record time for wetland attributes.The coastal wetland ontology we finally conduct contains six aspects:types,resources,landscapes,ecological hazards,record time,and literatures,which provides model support for the knowledge graph construction.(2)Methodology for knowledge extraction in coastal wetlands.In terms of data collection,the textual information on major coastal wetlands in China was crawled using official websites such as the Encyclopedia database,the China Forestry Network,and the China Knowledge Network as the main data sources.Structured data containing wetland attributes and wetland research literature attributes,as well as unstructured text data,were obtained.For the structured data,manual processing and filtering were performed to transform the data into a triple.For the unstructured coastal wetland knowledge,a deep learning-based joint entity-relationship extraction model was used to transform the unstructured text into a structured triple.For unstructured coastal wetland knowledge,a ME-R-T-BIO annotation strategy is proposed based on textual features and ontology structures containing temporal information.A Bi LSTM-CRF model is used to train the annotated corpus.The experimental F1 value was 84.51%,higher than 0.8.This indicates that the model performs well on the coastal wetland corpus and can be used to automate the extraction of coastal wetland knowledge based on this model.(3)Methodology for knowledge storage and visualization of coastal wetlands.The triples are stored in the Neo4 j graph database to realize the construction of the coastal wetland knowledge map completely.The application of knowledge retrieval and relational query provides coastal wetland managers and people with a more intuitive coastal wetland status,which is of great significance to the management of coastal wetland ecology and environment as well as conservation and publicity work.