A method for constructing a knowledge graph of river and lake health

Abstract

The present invention discloses a method for constructing a river and lake health knowledge graph, comprising the following main steps: based on the analysis of relevant water conservancy industry standards and river and lake health-related data resource types, defining river and lake health metadata types and knowledge service models based on catalog classification, determining the ontology set of the river and lake health ontology model and clarifying the attributes, modeling the river and lake health ontology library model according to attribute mining and establishing the relationship between ontologies; extracting more entities and association relationships from massive heterogeneous data resources through a variety of means such as topic mining, remote supervision, and causal relationship extraction, and further supplementing and improving the ontology library model: using a concept similarity calculation algorithm based on common attributes and a similarity calculation algorithm based on an in-and-out chain set for comprehensive calculation, reducing entity redundancy and realizing knowledge fusion; and realizing semi-automatic updating of the river and lake health knowledge graph by establishing an adaptive update mechanism.

Description

Technical Field

[0001] The present invention belongs to the technical field of knowledge graph construction, and specifically relates to a method for constructing a river and lake health knowledge graph. Background Art

[0002] Rivers and lakes are important carriers of water resources and an important part of the ecosystem. my country's rivers and lakes are scattered all over the country, and the management and protection tasks are heavy and difficult. In the long-term river and lake protection work, various river and lake management units have accumulated a large amount of river and lake health data, mainly including hydrology, water environment, water ecology, water resources, water safety monitoring data, business data, remote sensing data, basic geographic information data and video data, etc., and can provide data resource sharing services for relevant departments of river and lake management at all levels. However, these data resources have large data volume, complex sources, different structures and lack of knowledge. Without the condensation of river and lake health knowledge, it is difficult for managers to make accurate judgments on the current status of river and lake health and the status of emergencies. Therefore, in order to carry out river and lake management work more efficiently, it is necessary to realize the intelligent transformation of "data-information-knowledge-wisdom" and build a knowledge service system for river and lake health. Summary of the invention

[0003] The technical problem to be solved by the present invention is to address the defects involved in the background technology. The present invention provides a method for constructing a river and lake health knowledge map, aiming to establish a communication bridge between data and knowledge in the field of river and lake health, so as to solve the problems of rich, scattered, vague and unguided data in this field, and provide knowledge support for the decision-making of river and lake managers. Only on this basis can we accurately target the prominent problems faced by rivers and lakes in different periods, adopt comprehensive measures, and continuously explore good policies suitable for my country's national conditions.

[0004] In order to achieve the above object, the present invention specifically adopts the following technical solutions:

[0005] A method for constructing a river and lake health knowledge graph, characterized by comprising the following steps:

[0006] Step 1: Develop river and lake health data resource categories and related metadata paradigm documents;

[0007] Step 2: Formulate the queryable metadata items (index configuration files) and returnable metadata items (summary configuration files) required by the "Water Resources Data Directory Service Specification", as well as detailed presentation (details configuration files), and other rule configuration files required for metadata management;

[0008] Step 3: Define the ontology concepts related to river and lake health, determine the ontology set of the river and lake health ontology model and clarify the attributes, and model the river and lake health ontology library model based on attribute mining and the establishment of relationships between ontologies, aiming to build the backbone network of the knowledge graph;

[0009] Step 4: Based on step 3, more entities and relationships are extracted from massive heterogeneous data resources through various means such as topic mining, remote supervision, and causal relationship extraction, and added to the backbone relationship diagram as common sense content;

[0010] Step 5: Calculate the similarity between conceptual entities on the extracted entity triple data, and use the concept similarity calculation algorithm based on common attributes and the similarity calculation algorithm based on the in-and-out link set to perform comprehensive calculations to eliminate ambiguous and duplicate entities;

[0011] Step 6: Figure 5 An adaptive update mechanism for establishing a river and lake health knowledge graph is shown.

[0012] Furthermore, the step 3 specifically includes the following steps:

[0013] Step 3-1: Determine the professional field and scope of the ontology. The ontology model constructed by this method mainly covers the field of river and lake health. Refer to the river and lake health objects in the "Basic Database Table Structure and Identifiers of Water Conservancy Objects" to clarify the objects or facilities included in the river and lake health ontology;

[0014] Step 3-2: Examine the possibility of reusing existing river and lake health ontologies;

[0015] Step 3-3: List the important terms in the river and lake health ontology;

[0016] Step 3-4: Define the river and lake health ontology class and class hierarchy. In the river and lake health hierarchy, there are six layers in total. The parent node is the river and lake, and its child nodes include river and lake objects and wading objects.

[0017] River and lake objects are divided into two categories: natural and non-natural:

[0018] Within the scope of nature, according to its geographical scope, nature is subdivided into two categories: underground and surface. Within these two scopes (underground and surface), according to their properties and respective characteristics, they can be divided into two categories: water transmission channels and water collection and storage units;

[0019] In the non-natural level, there are two scopes: facilities and non-facilities; in the scope of facilities, there are two scopes: independent facilities and combined projects; in the scope of non-facilities, there are two scopes: behavior subjects and behavior objects; in the scope of behavior subjects, there are two ontological concepts: river and lake management related units and natural persons;

[0020] Step 3-5: Define the attributes of the river and lake health ontology. According to the river and lake health ontology class defined above, determine the attributes of the river and lake health ontology class; the attributes of a river and lake health ontology usually include two types, namely intrinsic attributes and extrinsic attributes;

[0021] Step 3-6: Define the facets of the river and lake health ontology attribute. This step determines the range value of the river and lake health ontology attribute and its data type;

[0022] Step 3-7: Construct a river and lake health ontology instance. Based on the preparatory work in the above steps, create a specific river and lake health ontology instance, in which the backbone nodes and association relationships are composed of the river and lake object classification directory and the unit department directory.

[0023] Further, step 4 mainly includes the following categories:

[0024] (1) Topic extraction and association: Through LDA and other topic mining methods, management topics contained in unstructured archival data resources such as academic / official documents, networks and sentiment are extracted; on this basis, through text mining, co-occurrence analysis and other means, the river and lake health management topic map is associated with the ontology structure in the river and lake object directory;

[0025] (2) Extraction and association of facts and principles: Through causal analysis and other methods, key events in the handling of different types of river and lake health events in academic / official documents and the impact relationship between events are extracted; on this basis, through text mining, co-occurrence analysis and other methods, the facts and principles map is associated with the ontology structure in the river and lake health object directory;

[0026] (3) entity / relationship dictionary, which extracts common professional terms in the field of river and lake health knowledge management through various means such as manual filling and text mining;

[0027] (4) Import OWL files and adjust and modify the concept layer of the river and lake knowledge graph through the OWL medium;

[0028] (5) Multi-topic scenario fusion: Through representation learning and other methods, based on relationship hierarchy, temporal relationship, etc., scenario graphs under specific management topics are extracted from the river and lake knowledge graph.

[0029] Further, step 5 specifically includes the following steps:

[0030] Step 5-1: Concept similarity calculation algorithm based on shared attributes, the calculation formula is:

[0031]

[0032] Formula 5-1 calculates the similarity of the river and lake health object concepts by the number of common attributes of the river and lake health objects, where |W1W2| represents the number of common attributes of the river and lake health object concepts W1 and W2, |W1| represents the number of attributes of the river and lake health object concept W1, and |W2| represents the number of attributes of the river and lake health object concept W2;

[0033] Step 5-2: Based on the similarity calculation algorithm of the in-and-out chain set, the calculation formula is:

[0034]

[0035] Calculate the outbound link set similarity and the inbound link set similarity separately, and then combine the two parts to calculate the concept similarity;

[0036] Step 5-3: Calculate the comprehensive similarity of concepts. Finally, use the linear weighted method to combine the common attribute similarity of the river and lake health object concepts and the similarity of the input and output chain sets to calculate the similarity of the two river and lake health object concepts:

[0037] sim(W1,W2)=α×sim1(W1,W2)+β×sim 12 (W1, W2)

[0038] (5-3);

[0039] In formula 5-3, d and β represent the weights of the common attribute similarity and the similarity of the in-and-out link sets respectively. The values ​​of α and β are obtained through multiple experiments, where α+β=1.

[0040] Further, step 6 specifically includes the following steps:

[0041] Step 6-1: Incremental data collection: for Oracle database, four incremental data collection methods, namely trigger-based, CDC-based, log analysis-based, and snapshot differential method-based, are used to obtain incremental data from structured data sources;

[0042] Step 6-2: Data cleaning and compression: compress and clean the captured incremental data. Use a data compression algorithm based on log analysis to parse the incremental data of the required instance according to specific scenarios and user needs.

[0043] Step 6-3: Data synchronization: parse the Oracle RowID mapping table through instructions to achieve data synchronization between heterogeneous and homogeneous databases, ensuring the consistency of data before and after the update;

[0044] Step 6-4: Knowledge graph update, based on the automatic annotation method of river and lake health management corpus under remote supervision, the template generalization technology is used to extract the ontology concept relationship to expand the river and lake health management corpus. The knowledge is completed through the change knowledge captured by the structured database and the expanded corpus, realizing the semi-automatic update of the knowledge graph.

[0045] Beneficial effects of the present invention:

[0046] (1) The present invention is used for the storage and intelligent identification of knowledge in the field of river and lake health, which can solve the problems of scattered, vague and unguided knowledge in the field of river and lake health, and has the ability to merge and summarize knowledge and provide self-learning services;

[0047] (2) Based on the river and lake health knowledge graph constructed by the present invention, a directory service model constructed with a multi-mapping mechanism is adopted, and based on this knowledge graph, efficient query and indexing of river and lake health data can be achieved;

[0048] (3) Through the construction method of the river and lake health knowledge graph of the present invention, the extraction of river and lake health structured data and unstructured text data, themes and facts can be realized more conveniently and efficiently, and river and lake health objects can be linked;

[0049] (4) The present invention proposes an algorithm for calculating the similarity of river and lake health object concepts. The algorithm uses the common attribute similarity of the river and lake health object concepts and the similarity of the input and output chain sets to comprehensively calculate the similarity of two river and lake health object concepts, merges the river and lake health object concepts with high similarity, abstracts the river and lake health object concepts upward by one layer, reduces the number of water conservancy object concepts, thereby reducing the storage cost of the knowledge graph and improving the query speed of the knowledge graph. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Schematic diagram of the process of constructing a knowledge graph for river and lake health.

[0051] Figure 2 This is a diagram of a lake object instance.

[0052] Figure 3 A graph of objects, attributes, and their relationships.

[0053] Figure 4 It is a diagram of things.

[0054] Figure 5 Provides an update mechanism for river and lake health maps.

[0055] Specific implementation methods

[0056] Embodiment 1

[0057] A method for constructing a river and lake health knowledge graph in this embodiment is characterized by comprising the following steps:

[0058] Step 1: Develop river and lake health data resource categories and related metadata paradigm documents;

[0059] Step 2: Formulate the queryable metadata items (index configuration files) and returnable metadata items (summary configuration files) required by the "Water Resources Data Directory Service Specification", as well as detailed presentation (details configuration files), and other rule configuration files required for metadata management;

[0060] Step 3: Define the ontology concepts related to river and lake health, determine the ontology set of the river and lake health ontology model and clarify the attributes, and model the river and lake health ontology library model based on attribute mining and the establishment of relationships between ontologies, aiming to build the backbone network of the knowledge graph;

[0061] Step 4: Based on step 3, more entities and relationships are extracted from massive heterogeneous data resources through various means such as topic mining, remote supervision, and causal relationship extraction, and added to the backbone relationship diagram as common sense content;

[0062] Step 5: Calculate the similarity between conceptual entities on the extracted entity triple data, and use the concept similarity calculation algorithm based on common attributes and the similarity calculation algorithm based on the in-and-out link set to perform comprehensive calculations to eliminate ambiguous and duplicate entities;

[0063] Step 6: Figure 5 An adaptive update mechanism for establishing a river and lake health knowledge graph is shown.

[0064] Embodiment 2

[0065] A method for constructing a river and lake health knowledge graph in this embodiment is characterized in that step 3 specifically includes the following steps:

[0066] Step 3-1: Determine the professional field and scope of the ontology. The ontology model constructed by this method mainly covers the field of river and lake health. Refer to the river and lake health objects in the "Basic Database Table Structure and Identifiers of Water Conservancy Objects" to clarify the objects or facilities included in the river and lake health ontology;

[0067] Step 3-2: Examine the possibility of reusing existing river and lake health ontologies;

[0068] Step 3-3: List the important terms in the river and lake health ontology;

[0069] Step 3-4: Define the river and lake health ontology class and class hierarchy. In the river and lake health hierarchy, there are six layers in total. The parent node is the river and lake, and its child nodes include river and lake objects and wading objects.

[0070] River and lake objects are divided into two categories: natural and non-natural:

[0071] Within the scope of nature, according to its geographical scope, nature is subdivided into two categories: underground and surface. Within these two scopes (underground and surface), according to their properties and respective characteristics, they can be divided into two categories: water transmission channels and water collection and storage units;

[0072] In the non-natural level, there are two scopes: facilities and non-facilities; in the scope of facilities, there are two scopes: independent facilities and combined projects; in the scope of non-facilities, there are two scopes: behavior subjects and behavior objects; in the scope of behavior subjects, there are two ontological concepts: river and lake management related units and natural persons;

[0073] Step 3-5: Define the attributes of the river and lake health ontology. According to the river and lake health ontology class defined above, determine the attributes of the river and lake health ontology class; the attributes of a river and lake health ontology usually include two types, namely intrinsic attributes and extrinsic attributes;

[0074] Step 3-6: Define the facets of the river and lake health ontology attribute. This step determines the range value of the river and lake health ontology attribute and its data type;

[0075] Step 3-7: Construct a river and lake health ontology instance. Based on the preparatory work in the above steps, create a specific river and lake health ontology instance, in which the backbone nodes and association relationships are composed of the river and lake object classification directory and the unit department directory.

[0076] Embodiment 3

[0077] A method for constructing a river and lake health knowledge graph in this embodiment is characterized in that step 4 mainly includes the following categories:

[0078] (1) Topic extraction and association: Through LDA and other topic mining methods, management topics contained in unstructured archival data resources such as academic / official documents, networks and sentiment are extracted; on this basis, through text mining, co-occurrence analysis and other means, the river and lake health management topic map is associated with the ontology structure in the river and lake object directory;

[0079] (2) Extraction and association of facts and principles: Through causal analysis and other methods, key events in the handling of different types of river and lake health events in academic / official documents and the impact relationship between events are extracted; on this basis, through text mining, co-occurrence analysis and other methods, the facts and principles map is associated with the ontology structure in the river and lake health object directory;

[0080] (3) entity / relationship dictionary, which extracts common professional terms in the field of river and lake health knowledge management through various means such as manual filling and text mining;

[0081] (4) Import OWL files and adjust and modify the concept layer of the river and lake knowledge graph through the OWL medium;

[0082] (5) Multi-topic scenario fusion: Through representation learning and other methods, based on relationship hierarchy, temporal relationship, etc., scenario graphs under specific management topics are extracted from the river and lake knowledge graph.

[0083] Embodiment 4

[0084] A method for constructing a river and lake health knowledge graph in this embodiment is characterized in that step 5 specifically includes the following steps:

[0085] Step 5-1: Concept similarity calculation algorithm based on shared attributes, the calculation formula is:

[0086]

[0087] Formula 5-1 calculates the similarity of the river and lake health object concepts by the number of common attributes of the river and lake health objects, where |W1W2| represents the number of common attributes of the river and lake health object concepts W1 and W2, |W1| represents the number of attributes of the river and lake health object concept W1, and |W2| represents the number of attributes of the river and lake health object concept W2;

[0088] Step 5-2: Based on the similarity calculation algorithm of the in-and-out chain set, the calculation formula is:

[0089]

[0090] Calculate the outbound link set similarity and the inbound link set similarity separately, and then combine the two parts to calculate the concept similarity;

[0091] Step 5-3: Calculate the comprehensive similarity of concepts. Finally, use the linear weighted method to combine the common attribute similarity of the river and lake health object concepts and the similarity of the input and output chain sets to calculate the similarity of the two river and lake health object concepts:

[0092] sim(W1,W2)=α×sim1(W1,W2)+β×sim 12 (W1, W2)

[0093] (5-3);

[0094] In formula 5-3, d and β represent the weights of the common attribute similarity and the similarity of the in-and-out link sets respectively. The values ​​of α and β are obtained through multiple experiments, where α+β=1.

[0095] Embodiment 5

[0096] A method for constructing a river and lake health knowledge graph in this embodiment is characterized in that step 6 specifically includes the following steps:

[0097] Step 6-1: Incremental data collection: for Oracle database, four incremental data collection methods, namely trigger-based, CDC-based, log analysis-based, and snapshot differential method-based, are used to obtain incremental data from structured data sources;

[0098] Step 6-2: Data cleaning and compression: compress and clean the captured incremental data. Use a data compression algorithm based on log analysis to parse the incremental data of the required instance according to specific scenarios and user needs.

[0099] Step 6-3: Data synchronization: parse the Oracle RowID mapping table through instructions to achieve data synchronization between heterogeneous and homogeneous databases, ensuring the consistency of data before and after the update;

[0100] Step 6-4: Knowledge graph update, based on the automatic annotation method of river and lake health management corpus under remote supervision, the template generalization technology is used to extract the ontology concept relationship to expand the river and lake health management corpus. The knowledge is completed through the change knowledge captured by the structured database and the expanded corpus, realizing the semi-automatic update of the knowledge graph.

[0101] Based on the above embodiments, the present invention has the following advantages: (1) The present invention is used for the storage and intelligent identification of knowledge in the field of river and lake health, which can solve the problems of scattered, vague and unguided knowledge in the field of river and lake health, and has the service capability of merging and summarizing knowledge and providing self-learning; (2) The river and lake health knowledge graph constructed based on the present invention adopts a directory service model constructed by a multiple mapping mechanism, and based on this knowledge graph, efficient query and indexing of river and lake health data can be achieved; (3) Through the construction method of the river and lake health knowledge graph of the present invention, the extraction of river and lake health structured data and unstructured text data, themes and facts can be realized more conveniently and efficiently, and river and lake health objects can be linked; (4) The present invention proposes an algorithm for calculating the similarity of river and lake health object concepts, which uses the common attribute similarity of river and lake health object concepts and the similarity of the in-and-out chain sets to comprehensively calculate the similarity of two river and lake health object concepts, merges the river and lake health object concepts with high similarity, abstracts the river and lake health object concepts upward by one layer, reduces the number of water conservancy object concepts, thereby reducing the storage cost of the knowledge graph and improving the query speed of the knowledge graph.

[0102] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for constructing a river and lake health knowledge graph, characterized in that: Here are the steps: Step 1: Develop river and lake health data resource categories and related metadata paradigm documents; Step 2: Develop queryable metadata items, returnable metadata items and detailed displays required by the Water Resources Data Directory Service Specification, as well as other rule configuration files required for metadata management; Step 3: Define the ontology concepts related to river and lake health, determine the ontology set of the river and lake health ontology model and clarify the attributes, and model the river and lake health ontology library model based on attribute mining and the establishment of relationships between ontologies, aiming to build the backbone network of the knowledge graph; Step 4: Based on step 3, more entities and relationships are extracted from massive heterogeneous data resources through the means of topic mining, remote supervision, and causal relationship extraction, and added to the backbone relationship diagram as common sense content; Step 5: Calculate the similarity between conceptual entities on the extracted entity triple data, and use the concept similarity calculation algorithm based on common attributes and the similarity calculation algorithm based on the in-and-out link set to perform comprehensive calculations to eliminate ambiguous and duplicate entities; Step 6: Establish an adaptive update mechanism for the river and lake health knowledge graph; Wherein, the step 3 specifically includes: Step 3-1: Determine the professional field and scope of the ontology. The constructed ontology model mainly covers the field of river and lake health. Refer to the river and lake health objects in the "Basic Database Table Structure and Identifiers of Water Conservancy Objects" to clarify the objects or facilities included in the river and lake health ontology; Step 3-2: Examine the possibility of reusing existing river and lake health ontologies; Step 3-3: List the important terms in the river and lake health ontology; Step 3-4: Define river and lake health ontology classes and class hierarchies; In the river and lake health hierarchy system, there are six levels in total. The parent node is the river and lake, and its child nodes include river and lake objects and wading objects; River and lake objects are divided into two categories: natural and non-natural: Within the scope of nature, it is subdivided into two categories, underground and surface, according to its geographical scope, and within the above two scopes, it is further divided into two categories, water transmission channels and water collection and storage units, according to their properties and respective characteristics; In the non-natural level, there are two scopes: facilities and non-facilities. Within the scope of facilities, there are two scopes: independent facilities and combined projects. Within the scope of non-facilities, there are two scopes: behavior subjects and behavior objects. Within the scope of behavior subjects, there are two ontological concepts: river and lake management related units and natural persons. Step 3-5: Define the attributes of the river and lake health ontology. According to the river and lake health ontology class defined above, determine the attributes of the river and lake health ontology class; the attributes of a river and lake health ontology usually include two types, namely intrinsic attributes and extrinsic attributes; Step 3-6: Define the facets of the river and lake health ontology attributes, determine the range value of the river and lake health ontology attributes and its data type; Step 3-7: Construct a river and lake health ontology instance; Based on the preparations in the above steps, create a specific river and lake health ontology instance, in which the backbone nodes and association relationships are composed of the river and lake object classification directory and the unit department directory; Wherein, the step 4 specifically includes: (1) Topic extraction and association: Through LDA topic mining, management topics contained in academic / official documents, network and unstructured archival data resources are extracted; on this basis, through text mining and co-occurrence analysis, the river and lake health management topic map is associated with the ontology structure in the river and lake object directory; (2) Extraction and association of facts and principles: Through the causal analysis of facts and principles, key events in the handling of different types of river and lake health events in academic / official documents and the impact relationship between events are extracted; on this basis, through text mining and co-occurrence analysis, the facts and principles map is associated with the ontology structure in the river and lake health object directory; (3) Entity / relationship dictionary, which extracts common professional terms in the field of river and lake health knowledge management through manual filling and text mining; (4) Import OWL files and adjust and modify the concept layer of the river and lake knowledge graph through the OWL medium; (5) Multi-topic scenario fusion: by using representation learning, based on relationship hierarchy and temporal relationship, scenario graphs under specific management topics are extracted from the river and lake knowledge graph; Wherein, the step 5 specifically includes: Step 5-1: Concept similarity calculation algorithm based on shared attributes, the calculation formula is: （5-1）； Formula 5-1 calculates the similarity of the river and lake health object concepts by the number of common attributes of the river and lake health objects, where |W1W2| represents the number of common attributes of the river and lake health object concepts W1 and W2, |W1| represents the number of attributes of the river and lake health object concept W1, and |W2| represents the number of attributes of the river and lake health object concept W2; Step 5-2: Based on the similarity calculation algorithm of the input and output chain sets, the calculation formula is: （5-2）； Calculate the outbound link set similarity and the inbound link set similarity separately, and then combine the two parts to calculate the concept similarity; Step 5-3: Calculate the comprehensive similarity of concepts; finally, use the linear weighted method to calculate the similarity of the common attributes of the river and lake health object concepts and the similarity of the input and output chain sets to calculate the similarity of the two river and lake health object concepts: （5-3）； In formula 5-3, α and β represent the weights of the shared attribute similarity and the similarity of the input and output link sets respectively. The values ​​of α and β are obtained through multiple experiments, where α+β=1.

2. The method for constructing a river and lake health knowledge graph according to claim 1, characterized in that: The step 6 specifically includes: Step 6-1: Incremental data collection: for Oracle database, four incremental data collection methods, namely trigger-based, CDC-based, log analysis-based, and snapshot differential method-based, are used to obtain incremental data from structured data sources; Step 6-2: Data cleaning and compression: compress and clean the captured incremental data. Use a data compression algorithm based on log analysis to parse the incremental data of the required instance according to specific scenarios and user needs. Step 6-3: Data synchronization: parse the Oracle RowID mapping table through instructions to achieve data synchronization between heterogeneous and homogeneous databases, ensuring the consistency of data before and after the update; Step 6-4: Knowledge graph update, based on the automatic annotation method of river and lake health management corpus under remote supervision, using template generalization technology to realize ontology concept relationship extraction, to expand the river and lake health management corpus, through the change knowledge captured by structured database and the expansion of corpus to complete the knowledge completion, and realize the semi-automatic update of knowledge graph.

Images