A method for constructing spatiotemporal knowledge graph based on cultivated land protection laws and regulations

Abstract

In the embodiment of the present invention, a method for constructing a spatiotemporal knowledge graph based on cultivated land protection regulations is provided, which belongs to the field of data processing technology, and specifically includes: using a model layer to define the entity and relationship expression of the knowledge graph for cultivated land protection regulations, using the entity as the basic node in the knowledge graph of cultivated land protection regulations, and using the relevance or dependency between different entities at the semantic level as a relationship expression; using a data layer to sort out the basic data, discretionary benchmark data and reasoning decision data related to cultivated land protection in the land-type administrative discretion benchmark file into the form of entity and relationship expression to fill in the knowledge graph for cultivated land protection regulations; according to the form of entity and relationship expression, the reasoning layer is divided into a rule reasoning mechanism, a spatiotemporal reasoning mechanism and a learning reasoning mechanism; the application layer generates a decision-making plan according to the reasoning analysis results of the rule reasoning mechanism, the spatiotemporal reasoning mechanism and the learning reasoning mechanism. Through the scheme of the present invention, the pertinence and adaptability are improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing technology, and in particular to a method for constructing a spatiotemporal knowledge graph based on farmland protection laws and regulations. Background Art

[0002] At present, there are methods for constructing legal knowledge graphs in the fields of finance, transportation, and medical care, but the application field of farmland protection has its own unique features. In the structuring process of non-agricultural discretionary regulations, the types of knowledge involved are more diverse, including multiple dimensions such as text, numerical values, time relationships, and spatial relationships. Therefore, it is particularly important to construct a targeted spatiotemporal knowledge expression model in order to fully capture and present these complex relationships and support specific discretionary and application processes.

[0003] It can be seen that there is an urgent need for a targeted and adaptable spatiotemporal knowledge graph construction method based on farmland protection regulations. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method for constructing a spatiotemporal knowledge graph based on farmland protection regulations, which at least partially solves the problems of poor pertinence and adaptability in the prior art.

[0005] The embodiment of the present invention provides a method for constructing a spatiotemporal knowledge graph based on farmland protection laws and regulations, including:

[0006] Step 1: Use the model layer to define the entity and relationship expressions of the knowledge graph for cultivated land protection regulations, use the entities as the basic nodes in the knowledge graph of cultivated land protection regulations, and use the relevance or dependency between different entities at the semantic level as the relationship expression;

[0007] Step 2: Use the data layer to sort out the basic data, discretionary benchmark data, and reasoning decision data related to cultivated land protection in the land-related administrative discretionary benchmark files into entity and relationship expressions and fill them into the knowledge graph for cultivated land protection regulations;

[0008] Step 3: Divide the reasoning layer into rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism according to the forms of entity and relationship expression;

[0009] Step 4: When receiving a user's instruction to query knowledge of farmland protection regulations, the application layer generates a decision plan corresponding to the farmland protection regulations queried by the user based on the reasoning analysis results of the rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism.

[0010] According to a specific implementation of the embodiment of the present invention, step 1 specifically includes:

[0011] Step 1.1: Take the name and attributes of the farmland protection administrative discretion benchmark data as nodes and define the entity as

[0012]

[0013] in, Represents an entity, Represents the name of the knowledge graph entity, The unique identifier of the knowledge graph entity. Represents the entity type of the knowledge graph;

[0014] Step 1.2: Define the relationship between entities in the knowledge graph as edges and express the relationship

[0015]

[0016] in, Represents the relationship name of the knowledge graph, A unique identifier that represents a knowledge graph relationship. Represents the relationship type of the knowledge graph, which includes attribute relationship, spatial relationship and temporal relationship.

[0017] According to a specific implementation of the embodiment of the present invention, the expression of the attribute relationship is:

[0018]

[0019] in, Represents the starting node entity, Indicates the attribute termination node, is the qualitative attribute relationship between the entity node and other nodes, It is the quantitative attribute relationship between the entity node and other nodes;

[0020] The expression of the spatial relationship is

[0021]

[0022] in, and Indicates the start and end node entities of the discretionary datum element. is the topological relationship between discretionary elements and land types, It is a measure of the overlap area between discretionary elements and land types, calculated from spatial coordinates based on projection;

[0023] The expression of the time relationship is

[0024]

[0025] in, and Indicates the start and end node entities of the discretionary datum element. The temporal relationship of events in the discretionary basis.

[0026] According to a specific implementation of the embodiment of the present invention, step 2 specifically includes:

[0027] Step 2.1, define basic data and express each basic data in the same way as defined in step 1.1, where the expression for defining basic data is

[0028]

[0029] in, is the name of the underlying data, is the code of the type to which the basic data element belongs, is the type of underlying data;

[0030] Step 2.2: Define the discretionary benchmark data as

[0031] ;

[0032] in, For discretionary matters, The legal basis for discretion, For matters of discretion, for discretionary measures;

[0033] Step 2.3, express the discretionary matters, the legal basis for discretion, the circumstances of the discretionary matters and the discretionary measures in the manner of the definitions in step 1.1, express the relations of the discretionary legal basis, the circumstances of the discretionary matters and the discretionary measures in the manner of step 1.2, and express the relations of the basic data corresponding to the circumstances of the discretionary matters and the discretionary benchmark data in the manner of step 1.2;

[0034] Step 2.4, define the inference decision data as

[0035]

[0036] in, Indicates the analytical means required to obtain decision-making, Indicates the object of analysis;

[0037] Step 2.5, perform entity expression on the analytical means required for decision-making, perform entity expression on the object of analysis and its corresponding basic data, and perform relational expression on the discretionary benchmark data corresponding to the object of analysis.

[0038] According to a specific implementation of the embodiment of the present invention, step 3 specifically includes:

[0039] Step 3.1, based on the form of entity and relationship expression, the behavior of occupying cultivated land is judged according to the preset rules, and a rule reasoning mechanism is generated;

[0040] Step 3.2, according to the temporal relationship and the form of relational expression, analyze the data in a specific time point or time period to generate a temporal reasoning mechanism, and according to the spatial relationship and the form of relational expression, use the geographic spatial relationship to analyze the spatial data related to land use to obtain a spatial reasoning mechanism, and combine the temporal reasoning mechanism and the spatial reasoning mechanism to obtain a spatiotemporal reasoning mechanism;

[0041] Step 3.3, through supervised learning of historical behaviors, build a discriminant model and add rule constraints based on the knowledge graph to its loss function. Then use the discriminant model to vectorize the form of entity and relationship expression to obtain a learning reasoning mechanism.

[0042] According to a specific implementation of the embodiment of the present invention, step 4 specifically includes:

[0043] Step 4.1, perform semantic search on the query words in the user query according to the form of entity and relationship expression, and return the search results, wherein the search results include documents containing the query words and semantically similar content, and the expression of the search results is

[0044]

[0045] in, Represents the search function, For user queries, is an entity in the knowledge graph, Represents the relationship between the query and the entity, Represents the constructed knowledge graph;

[0046] Step 4.2, using rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism to obtain the answer content from the search results, wherein the expression of the answer content is

[0047]

[0048] in, Indicates the question-and-answer function. represents the answer obtained from the search results through the reasoning mechanism in step 3;

[0049] Step 4.3, integrate the multi-source data into the knowledge graph according to the answer content to obtain a decision solution, where the expression of the decision solution is

[0050]

[0051] in, Represents a decision plan, For different farmland protection policies, Represents an assessment of the impact of a policy on the information provided by the question-answering system.

[0052] The spatiotemporal knowledge graph construction scheme based on cultivated land protection laws and regulations in the embodiment of the present invention includes: step 1, using the model layer to define the entity and relationship expression of the knowledge graph for cultivated land protection laws and regulations, taking the entity as the basic node in the knowledge graph of cultivated land protection laws and regulations, and taking the relevance or dependency between different entities at the semantic level as the relationship expression; step 2, using the data layer to sort out the basic data, discretion benchmark data and reasoning decision data related to cultivated land protection in the land-type administrative discretion benchmark file into the form of entity and relationship expression and fill it into the knowledge graph for cultivated land protection laws and regulations; step 3, according to the form of entity and relationship expression, the reasoning layer is divided into rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism; step 4, when receiving the user's instruction to query the knowledge of cultivated land protection laws and regulations, the application layer generates a decision plan corresponding to the cultivated land protection laws and regulations queried by the user according to the reasoning analysis results of the rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism.

[0053] The beneficial effects of the embodiments of the present invention are as follows: through the scheme of the present invention, based on the four levels of model layer, data layer, reasoning layer and application layer, the association relationship between entities of laws, regulations, data and scenarios is constructed from the three dimensions of attributes, space and time, and the reasoning mechanism is sorted out from the three aspects of logic, time and space, and learning, and decision-making plans are generated accordingly, thereby improving the pertinence and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 A schematic diagram of a process for constructing a spatiotemporal knowledge graph based on farmland protection laws and regulations provided in an embodiment of the present invention;

[0056] Figure 2 A schematic diagram of the application framework of a method for constructing a spatiotemporal knowledge graph based on farmland protection regulations provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0058] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0059] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present invention, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0060] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention. The drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0061] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0062] The implementation process of farmland protection is not isolated. It relies on multi-dimensional information input and systematic technical support. It is usually based on laws and regulations, and takes advance prediction and real-time monitoring of surveying and mapping data as the benchmark, combined with spatial analysis and management knowledge, to jointly build a farmland protection administrative discretion service framework. When constructing knowledge on farmland protection laws and regulations, there are problems such as diverse data sources, inconsistent knowledge systems, and lack of technical systems. These problems limit the construction of the knowledge model of farmland protection laws and regulations. As an important knowledge carrier, the accuracy and scientificity of farmland protection laws and regulations are crucial. It is urgent to formalize and express this knowledge to supplement the overall system of legal knowledge.

[0063] Constructing a knowledge expression model that is suitable for farmland protection is an important prerequisite for knowledge organization and association. Knowledge service is a further extension of information service. It is not only the transmission and acquisition of information, but also involves the organization, integration, generation and application of knowledge. In general, knowledge service is a higher-level form of information utilization, emphasizing intelligence and user orientation. Knowledge graphs represent and store various types of knowledge and their relationships in a structured way, which can provide a basic framework for the representation, management, reasoning and application of knowledge.

[0064] The process of building a knowledge graph usually includes four main steps: data collection, data processing, knowledge representation, and reasoning. First, domain-related information is obtained through multi-source data, and then the data is cleaned and standardized to ensure its quality and consistency. In the knowledge representation stage, a graph structure is used to model entities and their attributes and relationships, and finally a semantically rich knowledge network is constructed. The reasoning process combines rule reasoning and machine learning algorithms to achieve intelligent retrieval and inference of knowledge, thereby supporting decision making and problem solving.

[0065] At present, there are methods for constructing legal knowledge graphs in the fields of finance, transportation, and medical care, but the application field of farmland protection has its own unique features. In the structuring process of non-agricultural discretionary regulations, the types of knowledge involved are more diverse, including multiple dimensions such as text, numerical values, time relationships, and spatial relationships. Therefore, it is particularly important to construct a targeted spatiotemporal knowledge expression model in order to fully capture and present these complex relationships and support specific discretionary and application processes. It is mainly reflected in:

[0066] Characteristic 1. The degree of refinement of rule design. The rule design of the knowledge graph of cultivated land protection needs to be defined in detail for specific elements such as specific occupation behavior, use, and regional type, such as whether it is used for non-agricultural construction, whether it violates the red line of cultivated land, etc., so the granularity of the rules is relatively fine; while the rules of the general regulations knowledge graph are usually relatively general and applicable to a wider range of legal clauses and situations.

[0067] Feature 2: Spatial requirements for data sources. The knowledge graph of cultivated land occupation needs to be highly integrated with spatial data (such as GIS geographic information) to capture dynamic information such as the location, area changes, and use of cultivated land. This combination of spatial data has a direct impact on the penalty results, while general regulatory knowledge graphs often only need to process text data.

[0068] Feature 3: Geographic and temporal reasoning in the reasoning layer. Since farmland protection regulations are closely related to land type, regional policy and timeliness, graph reasoning needs to consider geographic spatial reasoning and temporal reasoning. For example, reasoning about illegal land occupation activities within a specific period of time, analyzing seasonal farmland changes, etc. This is rarely involved in general regulatory knowledge graphs.

[0069] Feature 4. Dynamic update frequency of regulations. Farmland protection regulations are often affected by multiple factors such as national policies, natural conditions, and regional development. The content of regulations and penalty standards may be updated frequently, so the update mechanism of the graph requires higher frequency and automation support. The update requirements of general regulatory knowledge graphs are relatively low, and usually only need to follow up on changes in legal revisions.

[0070] Feature 5. Specificity and operability of the application layer. The application of the farmland occupation penalty map is more focused on diagnosis and prediction, such as real-time warning of illegal land occupation, classification of illegal behaviors, and regional protection policy recommendations, which are highly operational; while the application of general legal knowledge maps is more extensive, usually used for legal consultation, case retrieval, etc.

[0071] An embodiment of the present invention provides a method for constructing a spatiotemporal knowledge graph based on farmland protection laws and regulations, which can be applied to the knowledge graph communication process of farmland protection scenarios.

[0072] See also Figure 1 , which is a flow chart of a method for constructing a spatiotemporal knowledge graph based on farmland protection laws and regulations provided by an embodiment of the present invention. Figure 1 As shown, the method mainly comprises the following steps:

[0073] Step 1: Use the model layer to define the entity and relationship expressions of the knowledge graph for cultivated land protection regulations, use the entities as the basic nodes in the knowledge graph of cultivated land protection regulations, and use the relevance or dependency between different entities at the semantic level as the relationship expression;

[0074] Furthermore, the step 1 specifically includes:

[0075] Step 1.1: Take the name and attributes of the farmland protection administrative discretion benchmark data as nodes and define the entity as

[0076]

[0077] in, Represents an entity, Represents the name of the knowledge graph entity, The unique identifier of the knowledge graph entity. Represents the entity type of the knowledge graph;

[0078] Step 1.2: Define the relationship between entities in the knowledge graph as edges and express the relationship

[0079]

[0080] in, Represents the relationship name of the knowledge graph, A unique identifier that represents a knowledge graph relationship. Represents the relationship type of the knowledge graph, which includes attribute relationship, spatial relationship and temporal relationship.

[0081] Furthermore, the expression of the attribute relationship is

[0082]

[0083] in, Represents the starting node entity, Indicates the attribute termination node, is the qualitative attribute relationship between the entity node and other nodes, It is the quantitative attribute relationship between the entity node and other nodes;

[0084] The expression of the spatial relationship is

[0085]

[0086] in, and Indicates the start and end node entities of the discretionary datum element. is the topological relationship between discretionary elements and land types, It is a measure of the overlap area between discretionary elements and land types, calculated from spatial coordinates based on projection;

[0087] The expression of the time relationship is

[0088]

[0089] in, and Indicates the start and end node entities of the discretionary datum element. The temporal relationship of events in the discretionary basis.

[0090] During the specific implementation, the model layer is constructed to define the entity expression and relationship expression of the knowledge graph, which is used to describe the laws, subjects, behaviors and their relationships related to farmland protection.

[0091] 1.1. Entity expression: The names and attributes of knowledge related to farmland protection are defined as nodes. The expression form is defined as follows:

[0092]

[0093] in, Represents an entity, A unique identifier representing a knowledge graph entity; Represents the entity type of the knowledge graph, that is, the concept to which it belongs.

[0094] 1.2. Relationship expression: The relationships between entities in the knowledge graph constructed by the present invention are defined as edges, including attribute relationships, spatial relationships, and temporal relationships.

[0095] 1.2.1 Relationship expression, the semantic description information of entities in the knowledge graph of administrative discretion for farmland protection covers textual semantic information in terms of entity name, time, space, attributes, etc., and emphasizes the meaning connection and inherent semantic characteristics between entity concepts. Relationship expression is defined as:

[0096]

[0097] In the formula, Represents a relationship expression, Represents the relationship name of the knowledge graph; A unique identifier that represents a knowledge graph relationship; Represents the relationship types of the knowledge graph, including attribute relationships, spatial relationships, and temporal relationships.

[0098] 1.2.2. Attribute relations are a specific clarification of the semantic relations in 1.2.1. for . It refers to the relationship between each element entity and the attribute item. According to the data characteristics in the field of cultivated land protection and the needs of subsequent application services, by analyzing the attribute expression system and information expression characteristics of the element entities in the administrative discretion of cultivated land protection, the attribute relationship is classified into two subcategories: qualitative relationship and quantitative relationship. Qualitative relationship refers to the specific entity or attribute associated with the element entity, such as "something-basis-Article 4 of a certain regulation", where "basis" belongs to the qualitative relationship between "something" and "Article 4 of a certain regulation". Quantitative relationship refers to the numerical attribute relationship of the entity, such as the specific area of ​​land occupied. The attribute relationship model RA is defined as:

[0099]

[0100] In the formula Represents the starting node entity, Indicates the attribute termination node; is the qualitative attribute relationship between the entity node and other nodes, It is the quantitative attribute relationship between the entity node and other nodes.

[0101] 1.2.3 Spatial relationship is a specific clarification of the semantic relationship in 1.2.1, that is, for , refers to the spatial relationship between the target element and the land type element in the discretion benchmark due to the geographical location, mainly referring to the topological relationship and the metric relationship (such as overlapping area) that may be involved. According to the data characteristics and application service needs in the field of cultivated land protection, the expression of spatial relationships mainly considers qualitative topological relationships and quantitative spatial metric relationships. Spatial relationship Defined as:

[0102]

[0103] in, and Indicates the start and end node entities of the discretionary datum element. is the topological relationship between discretionary elements and land types, It is a measure of the overlapping area between discretionary elements and land types, calculated based on spatial coordinates based on projection.

[0104] 1.2.4. The temporal relationship is a specific clarification of the semantic relationship in 1.2.1, that is, for , refers to the relationship between the target element and the benchmark event or legal clause in the time dimension in the discretionary benchmark, mainly involving the sequence of events, duration and scope of influence. According to the data characteristics and application requirements in the field of farmland protection, the expression of time relationship mainly considers the quantitative time interval relationship to determine the sequence of events, such as the relationship between the promulgation time and the implementation time of regulations. Time relationships include: the duration of effectiveness of regulations or policies, the implementation cycle of a certain farmland protection measure, etc. Time relationship definition for:

[0105]

[0106] in, and Indicates the start and end node entities of the discretionary datum element. The temporal relationship of events in the discretionary basis.

[0107] Step 2: Use the data layer to sort out the basic data, discretionary benchmark data, and reasoning decision data related to cultivated land protection in the land-related administrative discretionary benchmark files into entity and relationship expressions and fill them into the knowledge graph for cultivated land protection regulations;

[0108] Based on the above embodiment, step 2 specifically includes:

[0109] Step 2.1, define basic data and express each basic data in the same way as defined in step 1.1, where the expression for defining basic data is

[0110]

[0111] in, is the name of the underlying data, is the code of the type to which the basic data element belongs, is the type of underlying data;

[0112] Step 2.2: Define the discretionary benchmark data as

[0113] ;

[0114] in, For discretionary matters, The legal basis for discretion, For matters of discretion, for discretionary measures;

[0115] Step 2.3, express the discretionary matters, the legal basis for discretion, the circumstances of the discretionary matters and the discretionary measures in the manner of the definitions in step 1.1, express the relations of the discretionary legal basis, the circumstances of the discretionary matters and the discretionary measures in the manner of step 1.2, and express the relations of the basic data corresponding to the circumstances of the discretionary matters and the discretionary benchmark data in the manner of step 1.2;

[0116] Step 2.4, define the inference decision data as

[0117]

[0118] in, Indicates the analytical means required to obtain decision-making, Indicates the object of analysis;

[0119] Step 2.5, perform entity expression on the analytical means required for decision-making, perform entity expression on the object of analysis and its corresponding basic data, and perform relational expression on the discretionary benchmark data corresponding to the object of analysis.

[0120] In the specific implementation, the data layer is constructed, and the basic data, discretionary benchmark data and reasoning decision data of the spatiotemporal knowledge graph of cultivated land protection are sorted out, and sorted into the form of entity expression in step 1.1 and relationship expression in step 1.2. It involves converting the relevant data and regulatory topics of land use into structured information that can be used for discretionary understanding and analysis. The goal of this stage is to supplement and improve the semantic content of the discretionary knowledge graph by defining clear rules, concepts and decision-making processes. Among them, the acquisition of discretionary benchmark data is based on the basic information of laws and administrative documents; the acquisition of basic data is based on the basic text information of laws and administrative documents, as well as the basic data space, attributes and other contents produced by GIS; the acquisition of reasoning decision data is based on the topological relationship of relevant elements of GIS spatial analysis and other analysis results.

[0121] 2.1. Basic data focuses on expressing the basic data types and attributes related to farmland protection scenarios, such as general farmland information, permanent farmland protection information, spatial utilization planning information, three-zone and three-line demarcation information, etc. Basic data is defined as:

[0122]

[0123] DataName describes the name of the basic data; DataTypeID is the code of the type of the basic data element; DataType is the type of the basic data. Each basic data is expressed in the same way as defined in step 1.1.

[0124] 2.2. Discretion benchmark data focuses on sorting out the basic information in the land-related administrative penalty discretion benchmark documents, such as matters concerning the occupation of cultivated land in the context of cultivated land protection, the legal provisions and specific articles for determination, the circumstances that can be used as discretion benchmarks, and specific measures. Discretion benchmark data is defined as:

[0125] ;

[0126] Among them, PenaltyMatter describes the discretionary matters, including information such as the time type involved in farmland protection; PenaltyLaw is the legal basis for discretion, including specific information such as cited articles; PenaltyPlot is the circumstances of the discretionary matters, including specific information such as the type of land occupied and the area of ​​land occupied; PenaltyPunish is the discretionary measure, including specific information such as the disposal method and the amount involved. The PenaltyMatter, PenaltyLaw, PenaltyPlot, and PenaltyPunish data are all expressed in entities in the way defined in step 1.1, and the PenaltyLaw, PenaltyPlot, and PenaltyPunish under each PenaltyMatter are expressed semantically in the way defined in step 1.2. In addition, the basic data involved in PenaltyPlot and the discretionary benchmark data in step 2.2 are expressed in a relationship in the way defined in step 1.2.

[0127] 2.3. Reasoning decision data focuses on expressing the decision results that can be obtained through spatial reasoning analysis, such as whether the ground features suspected of occupying cultivated land have a specific spatial relationship with a certain basic land type through spatial geometric analysis. Reasoning decision data is defined as:

[0128]

[0129] Among them, AnalyType describes the analytical means required for decision-making, such as spatial overlay analysis; AnalyGeoFrom represents the object of analysis. AnalyType is expressed in the way defined in step 1.1. AnalyGeoFrom includes the discretionary target object expressed in the way defined in step 1.1, and the basic data in step 2.1 is expressed in the way defined in step 1.1. In addition, the relationship between the entity expressed by AnalyGeoFrom and the discretionary benchmark data entity involved in step 2.2PenaltyPlot is expressed in the way defined in step 1.2.

[0130] Step 3: Divide the reasoning layer into rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism according to the forms of entity and relationship expression;

[0131] Based on the above embodiment, step 3 specifically includes:

[0132] Step 3.1, based on the form of entity and relationship expression, the behavior of occupying cultivated land is judged according to the preset rules, and a rule reasoning mechanism is generated;

[0133] Step 3.2, according to the temporal relationship and the form of relational expression, analyze the data in a specific time point or time period to generate a temporal reasoning mechanism, and according to the spatial relationship and the form of relational expression, use the geographic spatial relationship to analyze the spatial data related to land use to obtain a spatial reasoning mechanism, and combine the temporal reasoning mechanism and the spatial reasoning mechanism to obtain a spatiotemporal reasoning mechanism;

[0134] Step 3.3, through supervised learning of historical behaviors, build a discriminant model and add rule constraints based on the knowledge graph to its loss function. Then use the discriminant model to vectorize the form of entity and relationship expression to obtain a learning reasoning mechanism.

[0135] In specific implementation, the construction of the reasoning layer, which is subdivided into three parts: rule reasoning, spatiotemporal reasoning, and learning reasoning, is a reasonable design for the reasoning needs of complex problems, which can significantly improve the reasoning ability and application efficiency of the knowledge graph. By comprehensively applying these three reasoning methods, a comprehensive analysis of administrative discretionary behavior in farmland protection can be achieved, thereby enhancing the application ability of the knowledge graph in different situations.

[0136] 3.1. Rule reasoning is an important part of the reasoning layer. It uses predefined rules to perform logical inferences on input data to obtain reasoning results that meet actual needs. In the administrative discretion scenario, rule reasoning is mainly used to judge the behavior of occupying cultivated land based on the entities and relationships constructed in step 2. The characteristics of rule reasoning are its determinism and explanatory power, which can effectively support land management decisions.

[0137] The present invention uses entity set E to represent all entities constructed in step 2 in the knowledge graph, for example , each is a specific object. The relationship set R represents the relationship between entities constructed in step 2, for example , each r is a specific relationship between two entities. The attribute set A represents the attributes of the entities constructed in step 2, for example , each a is a feature or state. The rule set P represents all the rules in the knowledge graph reasoning, such as , each p is a fact rule. Each inference rule can be expressed as:

[0138]

[0139] Represents the conditional rule part of the reasoning process, which can be formalized as:

[0140]

[0141] Representing Entities With entity The relationship between Representing Entities Properties Satisfy a condition. It represents the conclusion part of the reasoning process and represents the derived result.

[0142] 3.2. Spatiotemporal reasoning combines time and space information in the knowledge graph for complex reasoning, which is particularly suitable for fields that need to consider geographical location changes and temporal dynamics. The occupation of cultivated land often has spatiotemporal characteristics. Spatiotemporal reasoning combines temporal logic and spatial reasoning to analyze the time and geographical location changes of related behaviors, and comprehensively uses geographic spatial data and time dimension data for inference.

[0143] 3.2.1. Temporal reasoning is based on the temporal relationship defined in step 1.2.4 and the specific temporal relationship constructed in step 2, analyzing the data at a specific time point or time period, and is formally expressed as temporal logic. Spatial reasoning is based on the spatial relationship defined in step 1.2.3 and the specific spatial relationship constructed in step 2, analyzing the spatial data related to land use, and using geographic spatial relationships (such as inclusion relationships, intersection relationships, etc.) for inference. Joint reasoning of time and space combines the constraints of time and space to infer the legality of land use through spatiotemporal information.

[0144] In the spatiotemporal knowledge graph of cultivated land protection regulations designed by the present invention, temporal reasoning can be subdivided into time point reasoning and time period reasoning, where the relationship between time periods and time points can be described by interval logic and set operations. Assume that the time point is represented by Indicates that the time period is Indicates that the following main relations can be defined:

[0145] 3.2.1.1. Time inclusion relationship (including time points): for any time period and time point ,if lie in If , then there is a containment relationship, which can be expressed as:

[0146]

[0147] 3.2.1.2. Time inclusion relationship (including time periods): For any two time periods and ,if Completely included in Inside, then:

[0148]

[0149] 3.2.1.3. Time Adjacency: If any two time periods and are adjacent, then one of the following two conditions is satisfied:

[0150]

[0151] 3.2.1.4. Time overlap: If any two time periods and If there is a partial overlap, then:

[0152]

[0153] 3.2.1.5. Time relationship: If a time period Occurs in another time period Before or After, then:

[0154]

[0155]

[0156] 3.2.2. When the spatial reasoning of the spatiotemporal knowledge graph of cultivated land protection regulations designed by the present invention is performed, the discretionary benchmark data in step 2.2 The plot data PenaltyPlot mainly involves the Rtopology topological relationships of overlap, inclusion, separation and coincidence, which are described as follows:

[0157] 3.2.2.1. Spatial overlap relationship: Two spatial objects partially overlap, that is, they share some spatial areas, but each still has independent parts. For example, two adjacent plots of cultivated land overlap in some areas, indicating that part of the area of ​​the plots overlaps. This relationship is usually used to indicate that there is a partial intersection between two spatial objects. If entities A and B have a spatial overlap relationship, then:

[0158]

[0159] 3.2.2.2. Spatial inclusion relationship: A spatial object completely contains another spatial object, and the boundary and vertices of the contained object are located within the containing object. For example, the boundary of a protected area completely surrounds a piece of farmland. In this case, the protected area contains the farmland, showing a complete inclusion relationship. This relationship is suitable for describing geographic relationships of complete enclosure and dependence. If entities A and B have a spatial inclusion relationship, then:

[0160]

[0161] 3.2.2.3. Spatial separation relationship: There is no intersection or overlap between two spatial objects, indicating that they are independent of each other and have no contact relationship. For example, two independent plots of cultivated land are separated, do not intersect or overlap each other. This relationship is used to indicate the state of no contact and no shared space between spatial objects. If entities A and B have a spatial overlap relationship, then:

[0162]

[0163] 3.2.2.4. Spatial coincidence relationship: Two or more geographic objects completely cover or overlap in space, forming the same geometric shape and area. Specifically, if the geometric boundaries and internal areas of the objects are completely consistent, they can be considered to be spatially coincident. If entities A and B have a spatial coincidence relationship, then:

[0164]

[0165] 3.3. Learning reasoning combines machine learning technology to give full play to the explanatory power of the knowledge graph to enhance the model. Through supervised learning of historical behaviors, a discriminant model is constructed, and the legal rules and contextual information in the knowledge graph constructed in step 2 are combined to interpret and optimize the prediction results. By vectorizing the entities and relationships in the knowledge graph constructed in step 2, complex semantic information is converted into a computable form. This process uses representation learning methods to capture the potential features and relevance of the knowledge graph. Through vectorized representation, the efficiency of the model in knowledge reasoning and information retrieval is improved, and the understanding of the relationship between different entities is enhanced.

[0166] Knowledge graphs can provide additional legal constraints or factual supplements for the model. By adding constraints based on knowledge graphs to the model's loss function, the interpretability of model reasoning can be enhanced and the model output can be optimized. For example, introducing rule constraints

[0167]

[0168] in, is the loss of the machine learning model, is the graph-based rule constraint loss term, is the weight coefficient. By introducing this knowledge constraint, the model can infer relevant behaviors more accurately.

[0169] Step 4: When receiving a user's instruction to query knowledge of farmland protection regulations, the application layer generates a decision plan corresponding to the farmland protection regulations queried by the user based on the reasoning analysis results of the rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism.

[0170] Based on the above embodiment, step 4 specifically includes:

[0171] Step 4.1, perform semantic search on the query words in the user query according to the form of entity and relationship expression, and return the search results, wherein the search results include documents containing the query words and semantically similar content, and the expression of the search results is

[0172]

[0173] in, Represents the search function, For user queries, is an entity in the knowledge graph, Represents the relationship between the query and the entity, Represents the constructed knowledge graph;

[0174] Step 4.2, using rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism to obtain the answer content from the search results, wherein the expression of the answer content is

[0175]

[0176] in, Indicates the question-and-answer function. represents the answer obtained from the search results through the reasoning mechanism in step 3;

[0177] Step 4.3, integrate the multi-source data into the knowledge graph according to the answer content to obtain a decision solution, where the expression of the decision solution is

[0178]

[0179] in, Represents a decision plan, For different farmland protection policies, Represents an assessment of the impact of a policy on the information provided by the question-answering system.

[0180] In specific implementation, an intelligent comprehensive system is constructed by combining step 3.1 logical reasoning, step 3.2 spatiotemporal reasoning, and step 3.3 learning reasoning with the three functions of semantic search, question-answering system, and decision support. In this system, step 3.1 logical reasoning enhances the accuracy of semantic search, step 3.2 spatiotemporal reasoning provides dynamic spatiotemporal information for the question-answering system, and step 3.3 learning reasoning provides data-driven suggestions for the decision support system by analyzing historical data. This integration not only enhances the intelligence level of information retrieval, enabling users to obtain relevant knowledge more efficiently, but also provides instant and accurate answers through the question-answering system to deal with complex problems. In addition, the decision support function provides users with analysis and reasoning based on knowledge graphs to help them make more scientific decisions in a changing environment. This multifunctional comprehensive application model will greatly enhance the value and utility of knowledge graphs in actual scenarios.

[0181] 4.1. Semantic search uses the entity recognition and relationship reasoning of the knowledge graph to not only return documents containing the query terms, but also mine semantically similar content.

[0182]

[0183] in, Represents the search function, For user queries, The entities in the knowledge graph are constructed for step 2. Represents the relationship between a query and an entity.

[0184] 4.2. The question-answering system uses knowledge graphs to enhance the accuracy and contextual understanding of answers and parse users’ natural language questions to obtain the required information.

[0185]

[0186] in, Indicates the question-and-answer function. Represents the answer obtained from the search results through the reasoning mechanism in step three.

[0187] 4.3. Decision Support By integrating multi-source information into the graph, decision makers can analyze the relationship between various factors from a unified perspective.

[0188]

[0189] in, represents the decision support function, For different farmland protection policies, Represents an assessment of the impact of a policy on the information provided by the question-answering system.

[0190] At the same time, if Figure 2 As shown in the figure, the application layer can also be divided into four types of applications: descriptive, diagnostic, predictive, and solution-based, which reflects the rationality and necessity of systematically handling complex problems. Descriptive applications provide a comprehensive understanding of data and information through detailed descriptions and summaries, which helps to build a basic understanding of knowledge graphs. Diagnostic applications conduct in-depth analysis of specific issues based on existing data and rules, identify and explain potential problems and anomalies, and provide a basis for decision-making. Predictive applications predict possible future trends and developments through the analysis and model construction of historical data, which helps to formulate forward-looking strategies. Solution-based applications propose practical solutions or interventions based on the diagnosis and prediction results to achieve the goal of optimizing management and decision-making. Such a hierarchical design not only improves the application level and sophistication of knowledge graphs, but also enhances their pertinence and practicality in actual decision-making and problem solving.

[0191] Specifically, 4.1 Description Application Service

[0192] 4.1.1 Application Type Concept

[0193] The core of the descriptive service is to classify, describe and query the illegal occupation of cultivated land through relevant laws and regulations, and help users quickly find relevant legal provisions, penalty standards, case analysis and other information. This service focuses on providing users with detailed descriptions within the legal framework to facilitate policy inquiries, legal interpretation and case retrieval.

[0194] 4.1.2 Service Steps Process

[0195] 4.1.2.1 Graph Construction

[0196] Based on steps one and two, the laws and regulations related to farmland protection and related data are converted into a knowledge graph, including various legal provisions, penalty standards, etc., to build a hierarchical relationship between laws and regulations and data.

[0197] 4.1.2.2 Reasoning Method

[0198] Through keywords or related questions (such as "penalty standards for illegal occupation of cultivated land", "cases of illegal occupation of cultivated land in a certain province"), combined with the logical reasoning method of step 3.1, search for relevant legal provisions, penalty measures or historical cases.

[0199] 4.1.2.3 Matching and Presentation

[0200] Based on the input, the logical reasoning of step 3.1 is performed to return relevant regulations or penalty cases. The returned information may include the name of the regulation, the content of the clause, the penalty level, etc., and provide links or further case analysis.

[0201] 4.1.3 Formulaic Expression

[0202] Assume the query keyword is , then the most relevant regulations and other information are returned based on the query word reasoning, which can be expressed as:

[0203]

[0204] in, It returns a result set based on the query words; It is a complete set of all regulatory information, such as the requirements in step 2.2 of data layer construction; Indicates query terms and demand information The logical reasoning result of step three 3.1.

[0205] 4.2 Diagnostic Application Services

[0206] 4.2.1 Application Type Concept

[0207] The diagnostic service aims to diagnose possible risk areas and violations through spatial analysis of illegal occupation of cultivated land, and impose penalties or handle them based on laws and regulations. It focuses on spatial diagnosis of illegal occupation of cultivated land.

[0208] 4.2.2 Service Steps Process

[0209] 4.2.2.1 Graph Construction

[0210] Based on steps one and two, the laws and regulations related to farmland protection and related data are converted into a knowledge graph, including various legal provisions, penalty standards, etc., to build a hierarchical relationship between laws and regulations and data.

[0211] 4.2.2.2 Spatial data acquisition and preprocessing

[0212] Obtain the spatial data of the cultivated land to be diagnosed, including the boundaries of the cultivated land, land use status, location of illegal activities, etc. Clean, standardize and preprocess the spatial data.

[0213] 4.2.2.3 Spatiotemporal analysis and pattern recognition

[0214] Use GIS technology to conduct spatial analysis, and combine the logical reasoning in step 3.1 and spatiotemporal reasoning in step 3.2 to identify hot spots or specific patterns of illegal occupation of cultivated land. For example, whether it is permanent basic farmland or general cultivated land, and how much area is occupied.

[0215] 4.2.2.4 Penalty Rule Matching

[0216] According to relevant laws and regulations, the specific penalty results are generated by combining factors such as the severity of the illegal behavior, the occupied area, and whether it involves the ecological environment. For example, if the illegal land occupation area in a certain area exceeds the standards stipulated by laws and regulations, the penalty result can be automatically determined.

[0217] 4.2.3 Formulaic Expression

[0218] Assumptions The area illegally occupied is The severity of the violation and the penalty It can be expressed by the following formula:

[0219]

[0220] in, The area illegally occupied is The severity of the illegal behavior, as required in step 2, data layer construction; The penalty standards in relevant laws and regulations, such as the requirements in step 2, data layer construction 2.2; It is a mapping function for penalty rules, and gives the final penalty result according to area, severity and regulatory standards. .

[0221] 4.3 Predictive Application Services

[0222] 4.3.1 Application Type Concept

[0223] The predictive service uses historical data, geographic information, environmental changes and other factors to predict possible future illegal land occupation incidents and their risk distribution. Based on statistical and machine learning methods, the service analyzes potential trends of illegal activities and high-risk areas and provides early warning.

[0224] 4.3.2 Service Steps Process

[0225] 4.3.2.1 Graph Construction

[0226] Based on steps one and two, the laws and regulations related to farmland protection and related data are converted into a knowledge graph, including various legal provisions, penalty standards, etc., to build a hierarchical relationship between laws and regulations and data.

[0227] 4.3.2.2 Data collection and feature extraction

[0228] Collect historical data on illegal occupation of cultivated land, including the time, location, type of land use, etc. Extract various factors that affect illegal occupation of cultivated land, such as changes in land policies, population density, geographical location, and level of economic development.

[0229] 4.3.2.3 Model training and learning

[0230] Use historical data to train the prediction model. Use statistical regression models and machine learning models, combined with step 3.2 spatiotemporal reasoning and step 3.3 learning reasoning methods, to jointly learn the patterns and characteristics of illegal behaviors.

[0231] 4.3.2.4 Risk assessment and prediction

[0232] Based on the trained model, the risk of illegal occupation of land in each area in the future time period is evaluated. The prediction result is usually the probability of illegal occupation of land in a certain time period or the risk of occupation of each land type in a certain area.

[0233] 4.3.2.5 Result display and early warning

[0234] Visualize the prediction results, such as generating risk heat maps, time series graphs, etc., and provide risk warning prompts.

[0235] 4.3.3 Formulaic Expression

[0236] Assume that the prediction model for the risk of illegal land occupation at a certain time in the future can be expressed as:

[0237]

[0238] in, For time ( ) Time and space position Risk of illegal land occupation; Various factors that may influence illegal behavior; are the parameters or weights of the model.

[0239] 4.4 Solution-based application services

[0240] 4.4.1 Application Type Concept

[0241] The solution-based service automatically generates penalty results by matching the specific circumstances of the occupation of cultivated land with relevant regulatory templates. This service can formulate personalized penalty plans based on different types of violations and specific circumstances.

[0242] 4.4.2 Service Steps Process

[0243] 4.4.2.1 Graph Construction

[0244] Based on steps one and two, the laws and regulations related to farmland protection and related data are converted into a knowledge graph, including various legal provisions, penalty standards, etc., to build a hierarchical relationship between laws and regulations and data.

[0245] 4.4.2.2 Event Data Acquisition and Template Matching

[0246] Obtain data such as the region, area, and type of the occupation behavior, and use step 3.1 logical reasoning and step 3.2 spatiotemporal reasoning to match the reasoning results with the preset penalty templates in the regulations library. Each penalty template contains information such as the corresponding violation type, corresponding penalty standard, and penalty range.

[0247] 4.4.2.3 Template Standard Application

[0248] According to the specific circumstances of events such as occupation of cultivated land, appropriate penalty content is automatically filled in from the penalty template.

[0249] 4.4.2.4 Automatic report generation

[0250] The system automatically generates a detailed report, including basic information about the incident, applicable legal provisions, basis for sentencing, and sentencing results.

[0251] 4.4.3 Formula Expression

[0252] Assuming the template includes multiple factors (such as occupied area, disposal plan, legal provisions, etc.), the penalty result It can be expressed by the following formula:

[0253]

[0254] in, is the occupied area, For the disposal plan, is the relevant legal provisions, such as the requirements in step 2, data layer construction; m is the number of standards in the template, Refers to the i-th penalty criterion (such as fine amount, penalty level, etc.).

[0255] Four service types - description, diagnosis, prediction and solution services, combine spatiotemporal knowledge graphs, reasoning technology and data mining methods, and can provide comprehensive services for farmland protection from data display to decision support.

[0256] The method for constructing a spatiotemporal knowledge graph based on farmland protection regulations provided in this embodiment constructs regulations from the three dimensions of attributes, space, and time based on four levels: model layer, data layer, reasoning layer, and application layer. It constructs the relationship between entities among regulations, data, and scenarios, organizes the reasoning mechanism from the three aspects of logic, space, and learning, and generates decision-making plans based on this, thereby improving the pertinence and adaptability.

[0257] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof.

[0258] The above is only a 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 shall be subject to the protection scope of the claims.

Claims

1. A method for constructing a spatiotemporal knowledge graph based on farmland protection laws and regulations, characterized in that: include: Step 1: Use the model layer to define the entity and relationship expressions of the knowledge graph for cultivated land protection regulations, use the entities as the basic nodes in the knowledge graph of cultivated land protection regulations, and use the relevance or dependency between different entities at the semantic level as the relationship expression; Step 2: Use the data layer to sort out the basic data, discretionary benchmark data, and reasoning decision data related to cultivated land protection in the land-related administrative discretionary benchmark files into entity and relationship expressions and fill them into the knowledge graph for cultivated land protection regulations; Step 3: Divide the reasoning layer into rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism according to the forms of entity and relationship expression; Step 4: When receiving a user's instruction to query knowledge of farmland protection regulations, the application layer generates a decision plan corresponding to the farmland protection regulations queried by the user based on the reasoning analysis results of the rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism.

2. The method according to claim 1, characterized in that The step 1 specifically includes: Step 1.1: Take the name and attributes of the farmland protection administrative discretion benchmark data as nodes and define the entity as in, Represents an entity, Represents the name of the knowledge graph entity, The unique identifier of the knowledge graph entity. Represents the entity type of the knowledge graph; Step 1.2: Define the relationship between entities in the knowledge graph as edges and express the relationship in, Represents the relationship name of the knowledge graph, A unique identifier that represents a knowledge graph relationship. Represents the relationship type of the knowledge graph, which includes attribute relationship, spatial relationship and temporal relationship.

3. The method according to claim 2, characterized in that The expression of the attribute relationship is in, Represents the starting node entity, Indicates the attribute termination node, is the qualitative attribute relationship between the entity node and other nodes, It is the quantitative attribute relationship between the entity node and other nodes; The expression of the spatial relationship is in, and Indicates the start and end node entities of the discretionary datum element. is the topological relationship between discretionary elements and land types, It is a measure of the overlap area between discretionary elements and land types, calculated from spatial coordinates based on projection; The expression of the time relationship is in, and Indicates the start and end node entities of the discretionary datum element. The temporal relationship of events in the discretionary basis.

4. The method according to claim 3, characterized in that Step 2 specifically includes: Step 2.1, define basic data and express each basic data in the same way as defined in step 1.1, where the expression for defining basic data is in, is the name of the underlying data, is the code of the type to which the basic data element belongs, is the type of underlying data; Step 2.2: Define the discretionary benchmark data as ； in, For discretionary matters, The legal basis for discretion, For matters of discretion, for discretionary measures; Step 2.3, express the discretionary matters, the legal basis for discretion, the circumstances of the discretionary matters and the discretionary measures in the manner of the definitions in step 1.1, express the relations of the discretionary legal basis, the circumstances of the discretionary matters and the discretionary measures in the manner of step 1.2, and express the relations of the basic data corresponding to the circumstances of the discretionary matters and the discretionary benchmark data in the manner of step 1.2; Step 2.4, define the inference decision data as in, Indicates the analytical means required to obtain decision-making, Indicates the object of analysis; Step 2.5, perform entity expression on the analytical means required for decision-making, perform entity expression on the object of analysis and its corresponding basic data, and perform relational expression on the discretionary benchmark data corresponding to the object of analysis.

5. The method according to claim 4, characterized in that The step 3 specifically includes: Step 3.1, based on the form of entity and relationship expression, the behavior of occupying cultivated land is judged according to the preset rules, and a rule reasoning mechanism is generated; Step 3.2, according to the temporal relationship and the form of relational expression, analyze the data in a specific time point or time period to generate a temporal reasoning mechanism, and according to the spatial relationship and the form of relational expression, use the geographic spatial relationship to analyze the spatial data related to land use to obtain a spatial reasoning mechanism, and combine the temporal reasoning mechanism and the spatial reasoning mechanism to obtain a spatiotemporal reasoning mechanism; Step 3.3, through supervised learning of historical behaviors, build a discriminant model and add rule constraints based on the knowledge graph to its loss function. Then use the discriminant model to vectorize the form of entity and relationship expression to obtain a learning reasoning mechanism.

6. The method according to claim 5, characterized in that The step 4 specifically includes: Step 4.1, perform semantic search on the query words in the user query according to the form of entity and relationship expression, and return the search results, wherein the search results include documents containing the query words and semantically similar content, and the expression of the search results is in, Represents the search function, For user queries, is an entity in the knowledge graph, Represents the relationship between the query and the entity, Represents the constructed knowledge graph; Step 4.2, using rule reasoning mechanism, spatiotemporal reasoning mechanism and learning reasoning mechanism to obtain the answer content from the search results, wherein the expression of the answer content is in, Indicates the question-and-answer function. represents the answer obtained from the search results through the reasoning mechanism in step 3; Step 4.3, integrate the multi-source data into the knowledge graph according to the answer content to obtain a decision solution, where the expression of the decision solution is in, Represents a decision plan, For different farmland protection policies, Represents an assessment of the impact of a policy on the information provided by the question-answering system.

Images