Knowledge graph construction method for island management and island data management system
By constructing an island knowledge map, the problem of temporal correlation and semantic relationships in island data management is solved, and the understanding of spatiotemporal correlation and dynamic changes of island data is realized, and the discovery of scientific decision-making and correlation between data is improved. ability.
Patent Information
- Application Number
- CN202411570594.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2024-12-06
AI Technical Summary
Existing island data management methods are difficult to capture the spatiotemporal correlation and semantic relationships between data, resulting in the intrinsic connections of data that cannot be effectively revealed and utilized, limiting the efficient utilization and scientific management of island resources.
By building a knowledge graph for island management, obtaining and preprocessing multi-dimensional data of islands, establishing entity categories and relationship categories, using deep learning models to extract entities and relationships, generating island knowledge graphs, and combining island geographic maps to achieve data Expression of spatial and temporal correlation and semantic relationship.
It realizes the correlation between island data in time and space, enhances the understanding of dynamic changes in islands, provides a solid data support platform, improves the scientificity and forward-looking decision-making, and significantly improves the discovery of correlation between data. ability.
Smart Images

Figure CN119089997A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of island management, and in particular to a knowledge graph construction method for island management and an island data management system. Background Art
[0002] Island-related information covers a wide range of dimensions, including spatial data, temporal data, and attribute data. The massiveness, complexity, diversity, multi-scale, and cross-period characteristics of the data make its data management challenging. In the current practice of island data management, the commonly used data storage strategy tends to store different categories of island information in isolation. Although this model guarantees the professional management of data to a certain extent, it invisibly builds barriers between data. The data is stored in fragments, resulting in the neglect of its inherent correlation. The lack of an effective mechanism to capture and express the spatiotemporal correlation between data has become a major bottleneck restricting the efficient use and scientific management of island resources.
[0003] The data management methods currently used in the island field are as follows: In a Chinese patent document with publication number CN103049549A, a method and system for managing island data is disclosed. The method divides the database into a main database and a sub-database. The main database classifies and stores all data classifications of each island, and calls different types of data to organize them into scenes and display them on one layer. The sub-database is used to store thematic data, and each data is stored with spatial coordinate information.
[0004] In the Chinese patent document with publication number CN117349463A, a method and system for managing island data is disclosed. Different types of data are displayed on the same three-dimensional map, and matching classified data are labeled based on coordinates. An island data management system with an autonomous platform is established that can dynamically perform data visualization extraction, query and display.
[0005] At present, the management and storage of island data mainly rely on the block-based independent storage mode based on geographic maps. Although this method realizes the management of island data, it limits the interaction and association between data.
[0006] At present, the number of knowledge graph studies in the marine field is relatively limited. These studies mainly rely on widely collected online text data to construct thematic knowledge graphs, such as knowledge graphs for the transportation of dangerous goods at sea, knowledge graphs for marine traditional Chinese medicine, and knowledge graphs for marine expert management. Existing knowledge graph construction methods mostly use deep learning and other methods to extract entities and relationships from texts, which makes it difficult to capture the spatial relationship between entities and to establish the association between geographic entities and geographic maps, limiting the ability of knowledge graphs to fully reflect the multi-dimensional and complex characteristics of island ecology, economy, society, etc. Therefore, these knowledge graphs are more inclined to be used for popular science purposes of islands, and cannot be used for data management of islands.
[0007] Most of the existing island data management methods focus on using spatial data as a basic display platform to intuitively present the diverse characteristics and status of islands through the superposition of multi-level information. Although this approach can intuitively reflect the distribution of data in geographic space, its limitation is that it is difficult to reveal and utilize the deep-level correlation and semantic relationship within the data. Therefore, despite the rich information, the intricate connections and potential knowledge value between these isolated information units are often not effectively mined.
[0008] At present, the island knowledge graph lacks effective construction and design of temporal and spatial information, which directly hinders the capture and understanding of dynamic information such as long-term evolution of islands, seasonal changes, and immediate event responses. Summary of the invention
[0009] In view of the defects in the prior art, the purpose of the present invention is to provide a knowledge graph construction method for island management and an island data management system.
[0010] A knowledge graph construction method for island data management provided by the present invention includes: Step S1: obtaining the original data of the island and preprocessing the original data; Step S2: Establish entity categories and entity relationship categories that reflect the temporal and spatial relevance of island data, and construct the island knowledge graph model layer; Step S3: extracting entities and building relationships from the processed raw data according to the island knowledge graph model layer to obtain the island knowledge graph data layer; Step S4: Generate an island knowledge graph based on the island knowledge graph model layer and the island knowledge graph data layer.
[0011] Preferably, in step S1, obtaining multi-dimensional original data of the island includes: extracting multi-dimensional literature data related to the island from the island database, obtaining journals and papers related to the island from the academic resource library, obtaining policy information of the island, and obtaining the geographical space spots of the island.
[0012] Preferably, the entity categories in step S2 include concepts, geographic entities and general entities; The concept is an artificial definition of the division of things. The original data of the island is classified, and the data category obtained is the concept; the geographic entity is a natural or artificial feature with a specific spatial location and size; the general entity includes other entities except the geographic entity.
[0013] Preferably, the entity relationship categories include spatial relationships, temporal relationships, and semantic relationships; The spatial relationship is the position and distance relationship between geographical entities, the temporal relationship is the time ordering relationship of time series, and the semantic relationship is the relationship between two entities in the linguistic sense.
[0014] Preferably, the extraction of entities in step S3 includes: for data satisfying specific rules, using regularized expressions to extract entities; When extracting island literature data, we determine whether it is a relevant island based on the keywords, title and abstract of the literature: For the keywords of the literature, long sorted string search is used, and the island names are sorted in descending order according to the string length. The island name is searched by index to see if it is included in the keyword. If it is in the string, the keyword is removed and the subsequent island name search is performed; For the extraction of entities in titles and abstracts, the Bert+BiLSTM+CRF model is used to train the corpus data and perform entity extraction.
[0015] Preferably, the extraction of entities in the title and abstract includes: Step A1: By setting prompt words, the LLM model is used to simulate the generation of paper abstracts and titles. For the simulated and real paper titles and abstracts, annotated datasets are established according to the BIOES standard; Step A2: Build a Bert+BiLSTM+CRF model, divide the labeled data set into training set, test set, and validation set, and train the Bert+BiLSTM+CRF model; Step A3: Use the trained model to annotate the paper title and abstract. According to the annotation sequence output by the CRF layer, extract and construct entities from the text, and combine consecutive "B", "I" and "E" characters into one entity.
[0016] Preferably, the relationship building in step S3 includes: For the information already existing in the original data table, the corresponding data is constructed into triples according to the list name; In the existing relationship, by customizing the division of attributes, a derivative relationship between corresponding entities and attributes is established; Establish spatial relationships between geographic entities based on geographic patches.
[0017] Preferably, the spatial relationship between the geographic entities is calculated using GIS spatial intersection analysis and GIS proximity analysis methods; if the two are separated, they are represented by distance and direction; if the two are tangent, they are represented by "tangent"; if the two intersect, they are represented by "located at".
[0018] According to the present invention, a data management system for islands includes an island knowledge graph and an island geographic map. The island knowledge graph is obtained by using the knowledge graph construction method for island data management. Layer attributes and number attributes are set for entities in the knowledge graph, and the island knowledge graph and the island geographic map are associated according to the layer attributes and the number attributes.
[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. The knowledge graph for island management constructed by the present invention innovatively introduces the concepts of time dimension and geographic entity into the design of the knowledge graph, realizes the correlation of island data in time and space, and enhances the understanding of the dynamic changes of islands.
[0020] 2. The concept of time dimension cited in the present invention effectively ensures the accuracy and logic of time information, and provides support for historical tracing of island activities, development trend analysis, and future scenario simulation.
[0021] 3. The present invention refers to the concept of geographic entities, designs unique codes for the geographic entities involved in the island map, and develops and integrates with the map system based on this code to realize functions such as map display, associated query and spatial positioning of the island knowledge graph, connects the knowledge graph with the real map spots, and builds the relationship between geographic entities.
[0022] 4. The present invention can provide a solid data support platform for multiple key links of marine development projects, such as progress monitoring, resource utilization, etc., by systematically integrating multi-source data, thereby enhancing the scientificity and foresight of decision-making.
[0023] 5. By constructing a knowledge graph, the present invention significantly improves the ability to discover the correlation between island-related information, realizes deep interconnection between data, and overcomes the limitations of traditional data storage methods in revealing the internal connections of island data. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 Construct a flow chart for the knowledge graph for island management in the present invention; Figure 2 This is a classification diagram of the island information hierarchy in the present invention; Figure 3 It is the knowledge graph model layer of the island management data in the present invention; Figure 4 It is a schematic diagram of the spatial relationship ambiguity in the present invention; Figure 5 This is a schematic diagram of the knowledge graph data layer in the present invention; Figure 6 It is the data platform association between the knowledge graph and the geographic map in the present invention. DETAILED DESCRIPTION
[0025] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.
[0026] The present invention proposes a knowledge graph construction method for island management, which aims to construct a knowledge graph that can show the spatiotemporal correlation of island information by efficiently integrating multi-source data resources. The specific implementation steps of this method are: first, systematically collect and organize existing data resources, covering various types of information such as text materials and geographic spatial patterns, and classify and preprocess the information. Then, design the ontology framework of the island knowledge graph, including clarifying entity categories (such as geographic entities, conceptual entities, etc.) and their attribute definitions, as well as the definition of relationship categories (such as spatial relationships, semantic relationships, etc.). According to the designed ontology model layer, the corresponding entities and their relationships are extracted from the data to construct the data layer. Finally, the constructed knowledge graph is combined with the database system of island management to realize real-time data updating and query interaction, and provide accurate information services for decision support, resource planning, environmental protection, etc.
[0027] The present invention discloses a knowledge graph construction method for island data management, comprising: Island data collection and preprocessing: Obtain a large amount of original island data through different channels; specifically, including but not limited to: extracting multi-dimensional information related to islands from the marine platform database, including but not limited to basic archival records of islands, boundary coordinates of nature reserves, ecological protection red line demarcation areas, descriptions of natural landscape characteristics, vegetation dynamic monitoring records, and detailed information such as typical ecosystems; obtaining satellite analysis of tropical cyclone scale data from the official website of the China Meteorological Administration; obtaining attribute information related to island atlas entities from text materials, such as the "Guangdong Island Tourism Resources" published in 2015; searching for island-related policies from government websites at all levels, and marking the provinces, cities, regions or sea areas involved in the policies. Conduct in-depth mining on the theme of "islands" from well-known academic resource libraries (such as CNKI), sort by citation volume, screen out journals and papers, and obtain top-ranked search information, including paper titles, keywords, authors, author units, abstracts, etc.
[0028] Since there are many types of island data, a wide range, and a large amount of data, it is necessary to sort and classify the data according to the application scenarios of the knowledge graph, and screen the data of specific categories to construct the knowledge graph. In a preferred embodiment, the island information is divided into six categories: island attributes, island environment, island protection and development, island resources, island basic geographic information, and island research and consultation. Each level is further divided into secondary or tertiary categories. Among them, the island attributes are the inherent information of the island itself, the island environment refers to the disaster situation of the island and the quality of the surrounding waters, the island protection and development section records the development and utilization of the island and related planning and protection projects, the island resources record the natural resources and cultural resources of the island and its neighborhood, the island basic geographic information is the artificially demarcated island information (including the administrative region to which it belongs, area classification, etc.), and the island research and consultation module includes the papers and relevant scholars related to the island. For specific classification, see Figure 2 shown.
[0029] The collected data is classified and preprocessed: 1. The unstructured data is extracted by artificial induction to extract multiple concepts and attribute information reflecting island resources, ecological environment, development and utilization, and the extracted attribute information is linked to the map through spatial association; 2. The semi-structured data is denoised by python and converted into structured data, and converted into map spots by coordinate instantiation method. The typhoon map spots are buffered according to the average radius of the four directions of its level 8 wind circle; 3. All layer data in the structured data are unified into CGCS2000 coordinate system, and the 19-bit layer code is designed with reference to the landmark "Marine Data Classification and Coding Standard" (DB44 / T 2463-2024) encoding method. It consists of 13-bit seven-level standard classification code + 2-bit type code + 4-bit layer serial number. A data coding field is added to each spatial element in the layer and a 27-bit unified data code is designed. The data code is based on the layer code and then adds an 8-bit element serial number.
[0030] Construction of knowledge graph model layer 1.1. Entity classification. The model layer abstracts entity classes from the data and defines the relationships between different entities. The present invention divides entity categories into concepts, geographic entities, and general entities. Concepts are artificial definitions of the division of things. This application mainly includes the island classification concepts such as island environments mentioned above (excluding island attributes) and island type concepts such as uninhabited islands. Geographic entities refer to natural or artificial features with a certain spatial position and size in the real world, such as nature reserves, geothermal hot springs, lighthouses, etc. In addition to geographic entities, other entities are defined as general entities, such as vegetation coverage.
[0031] 1.2. Classification of entity attributes. Attributes are the descriptions of certain features of an entity, and attribute values are specific values assigned to attributes. The present invention divides entity attributes into general attributes, spatial attributes, temporal attributes, and unique attributes; defines the concepts to which entities are subordinate as general attributes, and each entity is subordinate to at most 3 levels of concepts; defines layer codes and data codes as spatial attributes for geographic entities; some entities have temporal features, and the present invention describes temporal attributes in terms of year, month, and day; designs unique attributes for island entities, including but not limited to information such as island area and development and utilization status.
[0032] 1.3. Relationship classification. According to the different categories of entities, the relationships between entities are divided into spatial relationships, temporal relationships, and semantic relationships. The relationship between geographic entities is generally a spatial relationship, which is described by statements such as direction and distance or "located at" to indicate relative spatial positions. For example, a nature reserve is located 1,000 meters north of a certain island. Temporal relationships focus on the time ordering of event sequences, such as the "after" time relationship between "Typhoon Tali on July 15, 2023" and "Typhoon Tali" the next day, which clarifies the sequence of events. A semantic relationship refers to the relationship between two entities in a linguistic sense, such as the triple <a certain island, island area type, extra-large island>. The present invention uses the concepts of the hierarchical classification of island information and the island attributes as attributes of nodes in the graph, and the concept of island type is used as a node in the graph to construct a semantic relationship with the island node.
[0033] 1.4. Relationship attribute design. Some relationships may change over time, and the present invention introduces time attributes into the relationship. For example, the first and last nodes of vegetation coverage, water quality evaluation, pollution emission evaluation, etc. may be the same. Relationships exist in events in different time periods. Quadruples are constructed, and time is added to the relationship as an attribute, such as <Water Quality Monitoring Station A, Water Quality Evaluation (Time: 2024), Category 1>, <Water Quality Monitoring Station A, Water Quality Evaluation (Time: 2023), Category 1>.
[0034] According to the above classification, the knowledge graph model layer of island management data is obtained as follows: Figure 3 shown.
[0035] Construction of knowledge graph data layer 2.1. Extraction of regular entities. For data that meets certain rules, regular expressions and other methods are used to extract data. For example, in the island natural landscape data, "Qingao Bay; Huanghuashan Forest Park; Baisha Bay" is stored in the data table as a scenic spot name. The regular expression "([^;]+);([^;]+);([^;]+)" is used for entity extraction to obtain "Qingao Bay", "Huanghuashan Forest Park" and "Baisha Bay".
[0036] 2.2. Extraction of text entities. For the extraction of paper entities, the relevant islands are determined from the keywords, titles and abstracts of the paper. For keywords, long-sorted string retrieval is used. The island names are sorted in descending order by string length. The island names are searched by index to see if they are included in the keywords. If they are in the string, the keywords are removed and the subsequent island name searches are performed. For example, for the keywords "Lilian Island; island coastline; coral reef", after searching for "Lilian Island", the string is removed to avoid incorrect searches for "Lilian Island".
[0037] For the extraction of entities in titles and abstracts, the Bert+BiLSTM+CRF model is used to train the corpus data and perform entity extraction. The specific steps are: Ⅰ. Annotation of the dataset. Some island-related research papers are limited. Through similar prompts such as "You are an island research expert. Please write a paper title and abstract on island A. Please note that the topic of the paper should be similar to the topic of island A in the previous historical dialogue. The keyword island A should not appear in the same position as the previous title and abstract.", the LLM large model is used to simulate and generate paper abstracts and titles. For the simulated and real paper titles and abstracts, an annotation dataset is established according to the standard BIOES (B indicates the beginning of an entity, E indicates the end of an entity, I indicates inside an entity, O indicates a non-entity, and S indicates a single-character entity). For example, "A study on the coastline of island A" is annotated as "O, O, B, I, E, O, O, O, O, O".
[0038] Ⅱ. Construct a Bert+BiLSTM+CRF model. First, use the pre-trained BERT model to encode the preprocessed text. This model maps each word in the text to a high-dimensional space vector that can fully capture the contextual relationship of the word. Then, the word vector sequence output by the BERT model is input into a bidirectional long short-term memory network (BiLSTM). This network layer further refines and strengthens the model's understanding of the relationships and attributes between entities in the text sequence by forward and reverse processing of the word vector sequence. Finally, the feature sequence output by the BiLSTM layer is passed to a conditional random field (CRF) layer. The CRF layer uses these features and learns the dependencies between labels, optimizes the annotation results of the entire sequence through a dynamic programming algorithm, and ensures that the output sequence is globally optimal.
[0039] III. Divide the labeled dataset into training set, test set and validation set, and then train the model.
[0040] IV. Use the trained model to annotate the paper title and abstract. According to the annotation sequence output by the CRF layer, extract and construct entities from the text, and combine consecutive "B", "I" and "E" characters into one entity.
[0041] Regarding the extraction of policy entities, when studying policies associated with islands, from the perspective of geographical regions, the names of islands and the provincial, municipal and national documents to which the islands belong are associated with the islands.
[0042] 2.3. Entity alignment and disambiguation. The entity alignment related to islands is mainly for different expressions of unified names of geographical entities. For example, "Guangzhou" and "Guangzhou City" both refer to Guangzhou City, Guangdong Province. For geographical entities with geographic spatial locations, the topological analysis of spatial locations is used to unify the names of entities that may cause ambiguity. For example, the attribute table of a natural landscape shows that the landscape is located in "Sanzhou Island". Through the spatial superposition analysis of natural landscapes and islands, it can be determined whether "Sanzhou Island" and the standard name "Dianbai Sanzhou Island" are the same entity; for text and other data, an ambiguous association dictionary is constructed to establish standard names and their potentially ambiguous names. For example, "Sui", "Yangcheng", "Huacheng", "Wuyangcheng", etc. all represent the same entity as the standard name "Guangzhou City". According to the ambiguous association dictionary, the entities in different triples are aligned.
[0043] The locations of the plots drawn over the years may be different, and the geographic locations of the same plots need to be aligned. For example, due to the different mapmakers and floating water flow, the locations of the same water quality monitoring station in different years may be slightly different. The geographic location information can be aligned based on the average location or the latest location.
[0044] Entity relationship construction 3.1. Extracting relations from data tables. For information that already exists in the original data table, the corresponding data can be constructed into triples according to the list name. For example, each row of the island basic archive data table stores the name of the island, its area, coastline length, etc., and each row can be read in turn to construct a triple <island name, island area, specific value of island area>, etc.
[0045] 3.2. Construction of diffraction relationships. Some directly acquired relationships may not fully meet the analysis and application requirements. In view of this, the present invention adopts an active construction strategy, that is, by customizing a set of standards and rules, further refining and deriving new and more insightful relationships from existing direct relationships.
[0046] Regarding the classification standard of island area, the division scheme shown in Table 1 is adopted to map the specific area value to the category label of “island area type” to realize the transformation of area attribute into semantic relationship.
[0047] Table 1 Island area classification indicators
[0048] The offshore characteristics of islands are defined according to Table 2, specifically by quantifying the spatial distance between the islands and the nearest continental coastline and encapsulating it into a standardized semantic expression of “island offshore type”.
[0049] Table 2 Judgment criteria for offshore island types
[0050] According to the proportion of the length of the artificial shoreline of the island in the total shoreline length, calculate the semantic relationship of the artificial shoreline proportion of the island, which reflects the development and utilization situation of the island. The formula (1) for the artificial shoreline proportion is as follows: (1) Among them, r is the artificial shoreline proportion, p is the length of the artificial shoreline, and n is the length of the natural shoreline.
[0051] Using topological analysis technology, overlay and analyze the area affected by the typhoon circle of level 8 in the historical typhoon records with the geographical location data of the islands. By identifying and recording all island instances that intersect with the typhoon-affected area, form a standardized triple <typhoon, affects, island>, and transform the spatial intersection relationship into the semantic relationship "affects" in this form, which deepens the understanding and description of the interaction between islands and natural disasters in the knowledge graph.
[0052] Through the geographical feature of spatial regional association, the present invention intelligently links policy documents to the corresponding island entities, realizing the transformation from geographical spatial relationship to high-order semantic relationship. Specifically, for example, Hailing Island is confirmed to be under the management of Guangdong Province, and the policy document "Notice of the People's Government of Guangdong Province on Accelerating the Development of Islands" aims to guide the development and utilization of islands in the whole province, then construct the triple <Hailing Island, relevant policy, Notice of the People's Government of Guangdong Province on Accelerating the Development of Islands>.
[0053] 3.3 Construction of spatial relationships between geographical entities.
[0054] Based on the spatial relationships between geographical patches can be divided into three types. First, geographical entities that are located on the island or exist in the adjacent waters attached to the island or have a serious impact on the island, the spatial relationship is expressed by "located in", such as territorial sea baseline marker monuments, mangroves, ecological protection red line areas, etc. Second, entities that can exist independently of the island are expressed by azimuth and distance, such as the triple <Island A, (azimuth) 1000m north, Water Quality Monitoring Station B>. Finally, the relationships between geographical entities that focus on spatial distance or topological degree are expressed by "far", "near", "intersect", etc.
[0055] For the "located in" relationship and spatial topological relationship, the corresponding relationship can be obtained by using GIS spatial intersection analysis and GIS proximity analysis methods. There are two cases for the representation of direction and distance. First, for geographical entities with certain island affiliation, the principle of entity proximity to islands is adopted. Taking water quality monitoring stations as an example, a water quality monitoring station may be set up to monitor the water quality of the waters near its nearest island. Therefore, the proximity analysis method is used to match the nearest island for each water quality monitoring station and calculate its distance and direction to the island. Secondly, for geographical entities that have an impact on multiple islands, the principle of island proximity to entities is adopted. Taking nature reserves as an example, the proximity analysis method is used to match the nearest nature reserve for each island and calculate the spatial position of the nature reserve relative to the island.
[0056] There may be more than one way to express the spatial relationship between two entities. For the spatial location of a nature reserve and an island, if the two are separated, they are expressed by distance and direction; if the two are tangent, they are expressed by "tangent"; if the two intersect, they are expressed by "located at". However, due to differences in mapping, there may be ambiguity when constructing spatial relationships. For example Figure 4 , which may actually be a tangent relationship, but is mistakenly represented as an intersection. This paper calculates the overlap ratio of all intersecting nature reserves and islands based on the overlap ratio formula (2). . All overlap ratios are arranged in ascending order, and the mutation value is selected from the overlap ratios below 10% as the threshold for distinguishing intersection and tangency. For example, if the data in the previous column are kept below 0.5% and the value suddenly changes from 2.3% to 3.5%, 3.5% is used as the threshold. If the overlap ratio is less than 3.5%, it is considered that the island and the nature reserve are tangent; otherwise, they are considered to intersect.
[0057] (2) in Indicates the area of two different geographic entities. is the overlapping area of the two patterns.
[0058] According to the above steps and combined with the mode layer, the data layer diagram can be obtained as follows: Figure 5 .
[0059] Graph Application of Data Platform In the data preprocessing stage, two fields, "layer code" and "data code", have been added to all vector layer data. The 19-bit layer code is designed with reference to the landmark "Marine Data Classification and Coding Standard" (DB44 / T 2463-2024) encoding method, which consists of 13-bit seven-level standard classification code + 2-bit type code + 4-bit layer serial number. A data coding field is added to each spatial element in the layer and a 27-bit unified data code is designed. The data code is based on the layer code and adds an 8-bit element serial number. Both codes are unique and correspond to the layer code and data code in the spatial attributes of the island map geographic entity.
[0060] The schematic diagram of the map association between the knowledge graph and the data platform is as follows: Figure 6 First, according to the layer code in the atlas spatial attribute, all layer data in the data platform are traversed through the code and the layer with the same layer code is located, and it is loaded and displayed in the two-dimensional base map of the data platform. Then, according to the data code in the atlas spatial attribute, all elements in the loaded layer are traversed, and the element with the same data code is located. The geographic information of the element, such as longitude and latitude, is read, and the code is written to locate the current view of the map system to the longitude and latitude coordinates, so as to obtain the specific geographic location of the atlas geographic entity.
[0061] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.
[0062] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.
Claims
1. A knowledge graph construction method for island data management, characterized in that: include: Step S1: obtaining the original data of the island and preprocessing the original data; Step S2: Establish entity categories and entity relationship categories that reflect the temporal and spatial relevance of island data, and construct the island knowledge graph model layer; Step S3: extracting entities and building relationships from the processed raw data according to the island knowledge graph model layer to obtain the island knowledge graph data layer; Step S4: Generate an island knowledge graph based on the island knowledge graph model layer and the island knowledge graph data layer.
2. The knowledge graph construction method for island data management according to claim 1 is characterized in that: In the step S1, the original multi-dimensional data of the island is obtained, including: extracting multi-dimensional literature data related to the island from the island database, obtaining journals and papers related to the island from the academic resource library, obtaining policy information of the island, and obtaining the geographical space spots of the island.
3. The knowledge graph construction method for island data management according to claim 1 is characterized in that: The entity categories in step S2 include concepts, geographic entities and general entities; The concept is an artificial definition of the division of things. The original data of the island is classified, and the data category obtained is the concept; the geographic entity is a natural or artificial feature with a specific spatial location and size; the general entity includes other entities except the geographic entity.
4. The knowledge graph construction method for island data management according to claim 3 is characterized in that: The entity relationship categories include spatial relationship, temporal relationship, and semantic relationship; The spatial relationship is the position and distance relationship between geographical entities, the temporal relationship is the time ordering relationship of time series, and the semantic relationship is the relationship between two entities in the linguistic sense.
5. The knowledge graph construction method for island data management according to claim 1 is characterized in that: The extraction of entities in step S3 includes: for data that meets specific rules, using regularized expressions to extract entities; When extracting island literature data, we determine whether it is a relevant island based on the keywords, title and abstract of the literature: For the keywords of the literature, long sorted string search is used, and the island names are sorted in descending order according to the string length. The island name is searched by index to see if it is included in the keyword. If it is in the string, the keyword is removed and the subsequent island name search is performed; For the extraction of entities in titles and abstracts, the Bert+BiLSTM+CRF model is used to train the corpus data and perform entity extraction.
6. The knowledge graph construction method for island data management according to claim 5 is characterized in that: The extraction of entities in titles and abstracts includes: Step A1: By setting prompt words, the LLM model is used to simulate the generation of paper abstracts and titles. For the simulated and real paper titles and abstracts, annotated datasets are established according to the BIOES standard; Step A2: Build a Bert+BiLSTM+CRF model, divide the labeled data set into training set, test set, and validation set, and train the Bert+BiLSTM+CRF model; Step A3: Use the trained model to annotate the paper title and abstract. According to the annotation sequence output by the CRF layer, extract and construct entities from the text, and combine consecutive "B", "I" and "E" characters into one entity.
7. The knowledge graph construction method for island data management according to claim 1 is characterized in that: The relationship building in step S3 includes: For the information already existing in the original data table, the corresponding data is constructed into triples according to the list name; In the existing relationship, by customizing the division of attributes, a derivative relationship between corresponding entities and attributes is established; Establish spatial relationships between geographic entities based on geographic patches.
8. The knowledge graph construction method for island data management according to claim 7 is characterized in that: The spatial relationship between the geographic entities is calculated using GIS spatial intersection analysis and GIS proximity analysis methods; if the two are separated, they are represented by distance and direction; if the two are tangent, they are represented by "tangent"; if the two intersect, they are represented by "located at".
9. A data management system for islands, characterized in that: It includes an island knowledge graph and an island geographic map. The island knowledge graph is obtained by using the knowledge graph construction method for island data management described in any one of claims 1 to 8. Layer attributes and number attributes are set for entities in the knowledge graph, and the island knowledge graph and the island geographic map are associated according to the layer attributes and number attributes.
Citation Information
Patent Citations
Island data management method and system
CN103049549A
Island data management method and system
CN117349463A
Geospatial information knowledge service method and device based on semantic network
CN116226302A
Crop remote sensing image intelligent sample library construction method fused with time sequence knowledge graph
CN117972116A
