Data integration system and semantic self-evolution method and data query method thereof

Abstract

The invention discloses a data integration system and a semantic self-evolution method and a data query method thereof, and belongs to the technical field of artificial intelligence. The system comprises a distributed semantic autonomous network composed of a plurality of data source nodes which are in communication connection, and when any data source node receives a service request, the distributed semantic autonomous network receives the service request; part or all of the data source nodes execute operations corresponding to the tasks in a decentration mode in a distributed mode; the service request comprises a system internal maintenance service and a system external service service; each data source node is deployed with a semantic proxy network, and the semantic proxy network comprises a user interaction module, a knowledge graph processing module, a semantic negotiation module and an intelligent contract module and is used for being responsible for local semantic modeling, cross-source semantic mapping and access strategy execution. According to the method, a distributed data integration framework which is free of a central node, semantic self-evolution and capable of guaranteeing privacy is adopted, and the problems of expansibility, collaborative efficiency and privacy protection in a dynamic heterogeneous scene in the prior art can be solved.

Description

Technical Field

[0001] This invention belongs to the field of data integration technology, specifically relating to a data integration system and its semantic self-evolution method and data query method. Background Technology

[0002] Current mainstream data integration solutions (such as Denodo's virtual data port technology) rely on a centralized logic layer and a predefined metadata model, which have the following core drawbacks: (1) Rigid semantic extension: Cross-source data mapping rules need to be manually defined, which makes it difficult to cope with dynamically added data sources such as real-time streams and unstructured data. For example, sensor data generated in real time by IoT devices is diverse in protocol and dynamically changes. Traditional solutions require frequent manual adjustment of metadata models, resulting in low integration efficiency.

[0003] (2) Distributed collaboration is inefficient: query execution depends on the central engine scheduling, and there is a significant delay when integrating data across regions. For example, the data interaction delay between branches of multinational companies can reach hundreds of milliseconds, and the failure of the central node can easily lead to the overall paralysis of the system.

[0004] (3) Privacy compliance risks: Data needs to be processed at a central node, which poses a risk of leakage of raw data and makes it difficult to meet the compliance requirements of industries such as healthcare and finance for localized data storage. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by providing a data integration system and its semantic self-evolution method and data query method. It adopts a distributed data integration architecture with no central node, semantic self-evolution, and guaranteed privacy, which can solve the problems of scalability, collaborative efficiency, and privacy protection in dynamic heterogeneous scenarios.

[0006] This invention provides the following technical solution: Firstly, a data integration system is provided, comprising: a distributed semantic autonomous network composed of several data source nodes connected by communication, wherein when any data source node receives a business request, some or all data source nodes perform the operation corresponding to the task in a decentralized manner; the business request includes internal system maintenance business and external system service business; Each data source node deploys a semantic proxy network, which includes a user interaction module, a knowledge graph processing module, a semantic negotiation module, and a smart contract module. The user interaction module receives external service requests from users and returns the request results to the users. The knowledge graph processing module stores a local data semantic model and supports dynamically adding entity types and relationship attributes. The semantic negotiation module negotiates semantic mappings with neighboring nodes in real time based on a Raft variant consensus algorithm. The smart contract module automatically generates access control smart contracts according to the local data sharing strategy of the data source nodes and verifies the requester's identity and access permissions in real time when receiving external service requests.

[0007] Optionally, the data source node has a built-in consortium blockchain module, and all consortium blockchain modules in the distributed semantic autonomous network synchronously maintain and update the same blockchain; the consortium blockchain module is used to record all execution operations and data flow of its corresponding data source node.

[0008] Optionally, it also includes: an elastic resource management layer, which sets up several cache spaces, each of which corresponds to one or more data source nodes, for caching the semantic mapping results of the data source nodes and the intermediate results of external service business requests.

[0009] Optionally, the data source node is also deployed with a local LSTM model to predict future traffic based on historical traffic of external service business; the smart contract module is also used to record the load status and resource usage of its corresponding data source node; the elastic resource management layer is also used to perform elastic scaling based on the load status, resource usage and predicted future traffic of each data source node.

[0010] Optionally, the elastic resource management layer performs elastic scaling based on the load status, resource usage, and predicted future traffic of each data source node. Specifically, when any data source node predicts an increase in its local future traffic, it selects to create a new semantic proxy instance on a data source node with a load below a set threshold. When any data source node predicts a decrease in its local future traffic, it marks the instance of the current node as reclaimable and performs an instance shutdown operation.

[0011] Secondly, a data query method based on the data integration system described in any one of the first aspects is provided, comprising the following steps: S1: Upon receiving a user's query request, generate a semantic query graph; S2: Decompose the query request into several sub-tasks based on the semantic query graph, and distribute each sub-task to its corresponding target data source node; S3: Perform hierarchical hash verification on the subtask query results returned by the target data source node. If the verification passes, merge the query results of each subtask into the final query result and return it to the user. If the verification fails, trigger a retry operation and re-execute steps S1-S3 until the verification passes or the preset number of retries is reached. Then, record the evidence storage exception in the blockchain and provide feedback to the user.

[0012] Optionally, in step S3, when the target data source node executes a subtask, it first filters invalid data according to the subtask, and then uses homomorphic encryption technology to encrypt the local data; when multiple target data source nodes perform joint data processing, a secure multi-party computation method is used to share intermediate results.

[0013] Optionally, in step S2, when distributing subtasks to the corresponding target data source nodes, the method for selecting target data source nodes related to the subtasks is: global knowledge graph index, distributed hash table, or based on contract event listening predefined in the smart contract.

[0014] Optionally, in step S2, when distributing subtasks to the corresponding target data source nodes, a routing path is dynamically selected using a reinforcement learning algorithm based on the semantic similarity between the subtasks and the target data source nodes and the node load.

[0015] Thirdly, a semantic self-evolution method for a data integration system based on any one of the first aspects is provided, comprising the following steps: A semantic proxy network is deployed for the newly added data source node C, and data patterns are extracted and a knowledge graph is generated through its own knowledge graph processing module. When a newly added data source node C sends a semantic discovery request to all data source nodes, each data source node that receives the semantic discovery request analyzes the data patterns of data source node C through federated learning to identify potential semantic associations. Based on the identified potential semantic associations, semantic mapping rules are generated through a consensus algorithm, and the consensus results are broadcast to all data source nodes, triggering all data nodes to update the knowledge graph cache.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention employs a distributed semantic integration, privacy-preserving data interaction, and decentralized system architecture design involving heterogeneous data sources. It enables distributed semantic autonomy of the system, addressing the scalability, collaborative efficiency, and privacy protection issues of existing technologies in dynamic heterogeneous scenarios. Specifically, it achieves dynamic evolution of the semantic model through a semantic proxy network and federated learning, eliminating the need for manually defining cross-source mapping rules. The time for new data source nodes to connect is reduced from hours to minutes, significantly lowering cross-regional query response time and supporting processing 100,000 query requests per second. It eliminates the need for centralized server deployment, reducing hardware costs and significantly decreasing the workload of manually maintaining semantic mappings. Furthermore, data from each data source node remains within its local node, minimizing the risk of sensitive data leakage. It complies with GDPR, HIPAA, and other regional compliance requirements, avoiding legal risks associated with cross-border business. Simultaneously, it supports the dynamic addition of millions of data source nodes, with the semantic grid automatically adapting to changes in data patterns, making it suitable for dynamic heterogeneous scenarios such as the Internet of Things and fintech. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the specific process of the data integration system of the present invention performing business operations; Figure 2 This is a structural framework diagram of the data integration system of the present invention; Figure 3 This is a sequence diagram of the semantic negotiation process during the execution of internal system maintenance services according to the present invention. Figure 4 This is a structural framework diagram of a specific example given in Embodiment 4 of the present invention. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the scope of protection of the present invention. It should be noted that the term "comprising" and any variations thereof in the specification, claims and the above-mentioned drawings of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or devices.

[0019] Example 1 like Figure 1 and Figure 2As shown, a data integration system includes: a distributed semantic autonomous network composed of several communicatively connected data source nodes. When any data source node receives a business request, some or all of the data source nodes perform the operation corresponding to the task in a decentralized distributed manner. The business requests include internal system maintenance services and external system service services. In this embodiment, the internal system maintenance service is a semantic self-evolution service, and the external system service service is a user data query service. The distributed semantic autonomous network is a peer-to-peer network composed of semantic proxies distributed across various data source nodes. It achieves dynamic evolution of the global knowledge graph through federated learning and distributed consensus algorithms, without the need for a predefined unified metadata model.

[0020] Each data source node deploys a semantic agent network, which includes a user interaction module, a knowledge graph processing module, a semantic negotiation module, and a smart contract module.

[0021] The user interaction module is used to receive external service requests from users and return the request results to the users.

[0022] The knowledge graph processing module stores a local data semantic model and supports dynamically adding entity types and relationship attributes.

[0023] The local data semantic model (such as patient-medical record-diagnosis) supports dynamically adding entity types (such as adding a new gene testing data entity) and relational attributes (such as adding an allergy history attribute to a medical record entity).

[0024] The semantic negotiation module, based on a Raft variant consensus algorithm, negotiates semantic mappings with neighboring nodes in real time.

[0025] Specifically, when the Customer ID of node A needs to be associated with the CustomerNo of node B, the mapping rule CustomerID=CustomerNo is automatically generated through a distributed voting mechanism and synchronized to all nodes in the network.

[0026] The smart contract module automatically generates access control smart contracts based on the local data sharing strategy of the data source node, and verifies the requester's identity and access permissions in real time when it receives external service business requests.

[0027] Local data sharing policies (such as only allowing partner medical institutions to access anonymous_patient name) can use digital certificates stored on the blockchain to verify the identity and access rights of requesters in real time.

[0028] In this embodiment, the data source node has a built-in consortium blockchain module. All consortium blockchain modules in the distributed semantic autonomous network synchronously maintain and update the same blockchain. The consortium blockchain module is used to record all execution operations and data flow of its corresponding data source node.

[0029] When the data integration system performs semantic self-evolutionary business, a new data source node initiates a semantic discovery request. The data source node that receives the semantic discovery request proposes semantic mapping rules, and then the global knowledge graph is dynamically self-evolved through a consensus algorithm. When performing user data query business, the data source node that receives the query request breaks down the sub-tasks and distributes them to the target data source node for execution. After receiving the query results of the sub-tasks, the system verifies and merges the query results of the sub-tasks and then feeds them back to the user.

[0030] Specifically, when performing internal system maintenance tasks, such as Figure 3 As shown, when a new data source node C is added, its semantic agent automatically extracts data patterns (such as field names in CSV files and key-value pairs in JSON data) and generates an initial knowledge graph fragment. Data source node C broadcasts a semantic discovery request to the entire network, and neighboring nodes collaboratively analyze its data patterns through federated learning to identify potential semantic relationships (such as the order date of node C and the OrderDate of node D may be the same concept). The semantic mapping relationship is confirmed through a consensus algorithm, the global knowledge graph is updated, and the cache of relevant nodes is triggered to update.

[0031] When executing external service requests, i.e., when there is no central user query, a user initiates a query request at one of the data source nodes. The semantic proxy of that initiating node breaks down the query into subtasks (such as filtering Beijing patients or filtering diabetes diagnosis records) and distributes them to the target data source node via a smart contract. The target data source node's local query optimizer adjusts the execution plan based on local resource status. For example, edge nodes prioritize data filtering (such as excluding invalid medical records) to reduce the amount of data uploaded to the cloud; then, a privacy computing module processes sensitive data. After each node returns encrypted results, the initiating node verifies the data using hierarchical hashing (such as constructing a Merkle Patricia tree) to ensure the data has not been tampered with. If an inconsistency is found in the results of a node, the task retry for that node is automatically triggered, and an anomaly log is recorded on the blockchain.

[0032] In this embodiment, the data integration system further includes an elastic resource management layer, which sets up several cache spaces, each of which corresponds to one or more data source nodes, for caching the semantic mapping results of the data source nodes and the intermediate results of external service business requests.

[0033] In some other embodiments, the data source nodes also deploy a local LSTM model to predict future traffic based on historical traffic from external service traffic; the smart contract module is used to record the load status and resource usage of its corresponding data source node; and the elastic resource management layer is used to perform elastic scaling based on the load status, resource usage, and predicted future traffic of each data source node. Specifically: when any data source node predicts an increase in its local future traffic, it selects to create a new semantic proxy instance on a data source node with a load below a set threshold; when any data source node predicts a decrease in its local future traffic, it marks the current node's instance as reclaimable and performs an instance shutdown operation.

[0034] Example 2 A data query method based on the data integration system of Embodiment 1 is provided, comprising the following steps: S1: Upon receiving a user's query request, a semantic query graph is generated. The query request is an external service business request of the system.

[0035] S2: Based on the semantic query graph, the query request is broken down into several sub-tasks, and each sub-task is distributed to its corresponding target data source node.

[0036] The methods for generating semantic query graphs and splitting subtasks can refer to existing technologies. When distributing subtasks to corresponding target data source nodes, the methods for selecting target data source nodes related to the subtasks can be: global knowledge graph index, distributed hash table, or based on predefined contract event listening in smart contracts.

[0037] The global knowledge graph index works as follows: When each node joins the network, it registers its data patterns or capabilities with a global registry center (which can be decentralized). The initiating node queries this global index to find the appropriate node to handle each subtask.

[0038] Specifically, a distributed hash table (DHT) allows nodes to find the node responsible for a given data based on its key (e.g., entity type, attribute name).

[0039] The process of listening to pre-defined contract events within a smart contract is as follows: The initiating node deploys the smart contract to the blockchain (if the contract is new) or invokes an existing contract template. Then, the target node listens for smart contract events relevant to itself, or the initiating node directly sends the contract to the target node via a peer-to-peer network. Upon receiving the smart contract, the target node's smart contract engine executes the subtasks specified in the contract. This execution may include accessing local data and performing local computations. After execution, the target node returns the result according to the contract's specifications.

[0040] In this embodiment, when distributing subtasks to the corresponding target data source nodes, a routing path is dynamically selected using a reinforcement learning algorithm based on the semantic similarity between the subtasks and the target data source nodes and the node load.

[0041] Specifically, after the query request is converted into a semantic query graph, a routing path is dynamically selected based on real-time semantic similarity (such as entity association and attribute matching) and node load (such as CPU utilization and network bandwidth) using a reinforcement learning algorithm. For example, when querying medical records of diabetic patients in Beijing in 2023, the request is preferentially routed to the node that stores the entities "patient" and "medical records" and is located in Beijing. If the node is too overloaded, it is automatically switched to other semantically matching nodes.

[0042] S3: Perform hierarchical hash verification on the subtask query results returned by the target data source node. If the verification passes, merge the query results of each subtask into the final query result and return it to the user. If the verification fails, trigger a retry operation and re-execute steps S1-S3 until the verification passes or the preset number of retries is reached. Then, record the evidence storage exception in the blockchain and provide feedback to the user.

[0043] When a target data source node executes a subtask, it first filters invalid data based on the subtask, and then encrypts the local data using homomorphic encryption technology. When multiple target data source nodes perform joint data processing, a secure multi-party computation method is used to share intermediate results.

[0044] Hierarchical hash verification: Based on the Merkle tree structure, multi-level hash verification is performed on the results of distributed computing, and blockchain notarization is combined to realize the integrity and traceability of the data aggregation process.

[0045] In some other embodiments, when any data source node predicts an increase in its local query traffic, it selects to create a new semantic proxy instance on a data source node with a load below a set threshold. When any data source node predicts a decrease in its local query traffic, it marks the current node's instance as reclaimable and performs an instance shutdown operation.

[0046] Example 3 A semantic self-evolution method based on the data integration system described in Embodiment 1 includes the following steps: A semantic proxy network is deployed for the newly added data source node C, and data patterns are extracted and a knowledge graph is generated through its own knowledge graph processing module. When a newly added data source node C sends a semantic discovery request to all data source nodes, each data source node that receives the semantic discovery request analyzes the data patterns of data source node C through federated learning to identify potential semantic associations. Based on the identified potential semantic associations, semantic mapping rules are generated through a consensus algorithm, and the consensus results are broadcast to all data source nodes, triggering all data nodes to update the knowledge graph cache.

[0047] Specifically: Node initialization: Data source nodes deploy semantic agents to automatically extract local data patterns (such as fields and table structures) and generate initial knowledge graph fragments (such as patient-medical record relationships). Semantic automatic alignment: New nodes broadcast semantic discovery requests, and neighboring nodes collaboratively analyze semantic associations (such as matching order dates with OrderDate) through federated learning. Mapping rules are generated within minutes using a consensus algorithm and synchronized across the entire network.

[0048] Example 4 A specific application example of a data integration system is provided. The scenario of this application example is that a multinational medical group needs to integrate patient data from its hospitals in various countries for diabetes epidemiological research, while also meeting the requirements of the EU GDPR and China's "Guidelines for the Security of Health and Medical Data".

[0049] like Figure 4 As shown, the data integration system specifically includes: I. Distributed Semantic Autonomous Network Layer: Semantic Proxy Network: Each data source node deploys a semantic proxy, including: Lightweight knowledge graph processing module: It uses OWL ontology to store local data semantic models and supports dynamically adding entity types and relationship attributes.

[0050] Semantic negotiation module: Based on the Raft variant consensus algorithm, it negotiates semantic mapping with neighboring nodes in real time.

[0051] Smart Contract Module: Automatically generates access control smart contracts based on local data sharing policies. When a query request is received, it verifies the requester's identity and permissions in real time.

[0052] II. Distributed Execution Layer Centralized query execution: The semantic proxy of the initiating node breaks down the query into subtasks (such as "screening Beijing patients" and "screening diabetes diagnosis records"), and distributes them to the target nodes through smart contracts. The local query optimizer of the target node adjusts the execution plan based on the local resource status. For example, edge nodes prioritize data filtering (such as excluding invalid medical records) to reduce the amount of data uploaded to the cloud. The privacy computing module processes sensitive data: homomorphic encryption is used for "patient names," allowing the number of records that meet the criteria to be counted in an encrypted state without exposing the specific names; secure multi-party computation is used for "diagnosis results," and when multiple hospital nodes jointly calculate the prevalence rate, they only share the encrypted intermediate results and do not disclose the original diagnosis data. Result aggregation and verification: After each node returns the encrypted result, the initiating node verifies the data using hierarchical hashing (such as constructing a Merkle Patricia tree) to ensure that the data has not been tampered with. If an inconsistency is found in the result of a node, the task of that node is automatically retried, and the exception is logged through the blockchain.

[0053] III. Flexible Resource Management Distributed caching network: A decentralized cache is built based on the IPFS protocol. Frequently accessed semantic mapping rules (such as "Customer ID-CustomerNo" mapping) and query result shards (such as popular report data) are stored in the cache of neighboring nodes, reducing redundant calculations and network transmission overhead.

[0054] Adaptive scaling: The traffic prediction agent analyzes historical query patterns (such as 10:00 AM as the daily query peak) using an LSTM model, and creates semantic proxy instances in advance on nodes with low load to dynamically expand computing resources; during query troughs, it automatically reduces instances to lower energy consumption costs.

[0055] IV. Trusted Governance Layer Blockchain Evidence Preservation: Consortium blockchain records the entire data integration process: Query Request Evidence Preservation: records the requester's identity, query time, and semantic query graph content; Semantic Negotiation Evidence Preservation: records the creator of the mapping rule, creation time, and consensus voting results; Data Flow Evidence Preservation: records the data transmission path from the source node to the target node and processing operations (such as de-identification and aggregation).

[0056] Federal Access Control: Attribute-Based Encryption (ABE) is used to achieve fine-grained authorization. For example, the authorization rule "{Department: Medical R&D}∧{Role: Analyst}∧{Data Tag: Anonymous}" means that only analysts in the Medical R&D department can access anonymized medical data, and each access requires dynamic verification of the attribute combination through a smart contract.

[0057] Business process: Data source access: Semantic proxies are deployed at each hospital node to automatically extract the semantic model of the local electronic medical record system (such as "patient ID", "diagnosis time" and "disease type"), and negotiate semantic mapping with other nodes through federated learning (such as unifying the enumeration value of the "disease type" field).

[0058] Query Initiation: Researchers submit the query "Analyze the age distribution of diabetes patients in Central Europe in 2022-2023". The semantic agent converts it into a semantic query graph, which includes entities such as "patients" and "diagnosis records" as well as filtering conditions such as "time range" and "region".

[0059] Distributed execution: The European hospital node uses a privacy computing module to homomorphically encrypt "patient age" and calculates a histogram of age distribution that meets the criteria. The Chinese hospital node desensitizes "patient age" (e.g., by replacing it with an age range) and aggregates the results jointly with the European node through secure multi-party computation.

[0060] Result Verification and Delivery: The initiating node verifies the consistency of the results from the China-Europe nodes through hierarchical hashing, merges them to generate the final report, and the blockchain records all data operations for auditing by regulatory agencies.

[0061] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Those skilled in the art will clearly understand that the technologies in the embodiments of this invention can be implemented using software and necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments of this invention.

[0062] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A data integration system, characterized in that, include: A distributed semantic autonomous network consisting of several data source nodes connected by communication allows some or all data source nodes to perform the corresponding operations in a decentralized manner when any data source node receives a business request; the business requests include internal system maintenance services and external system service services. Each data source node deploys a semantic proxy network, which includes a user interaction module, a knowledge graph processing module, a semantic negotiation module, and a smart contract module. The user interaction module receives external service requests from users and returns the request results to the users. The knowledge graph processing module stores a local data semantic model and supports dynamically adding entity types and relationship attributes. The semantic negotiation module negotiates semantic mappings with neighboring nodes in real time based on a Raft variant consensus algorithm. The smart contract module automatically generates access control smart contracts according to the local data sharing strategy of the data source nodes and verifies the requester's identity and access permissions in real time when receiving external service requests.

2. The data integration system according to claim 1, characterized in that, The data source node has a built-in consortium blockchain module, and all consortium blockchain modules in the distributed semantic autonomous network synchronously maintain and update the same blockchain; the consortium blockchain module is used to record all execution operations and data flow of its corresponding data source node.

3. The data integration system according to claim 1, characterized in that, Also includes: The elastic resource management layer has several cache spaces, each of which corresponds to one or more data source nodes. These cache spaces are used to cache the semantic mapping results of the data source nodes and the intermediate results of external service business requests.

4. The data integration system according to claim 3, characterized in that, The data source node is also equipped with a local LSTM model, which is used to predict future traffic based on the historical traffic of external service business. The smart contract module is also used to record the load status and resource usage of its corresponding data source node. The elastic resource management layer is also used to perform elastic scaling based on the load status, resource usage and predicted future traffic of each data source node.

5. The data integration system according to claim 4, characterized in that, The elastic resource management layer performs elastic scaling based on the load status, resource usage, and predicted future traffic of each data source node. Specifically, when any data source node predicts an increase in its local future traffic, it selects to create a new semantic proxy instance on a data source node with a load below a set threshold. When any data source node predicts a decrease in its local future traffic, it marks the instance of the current node as reclaimable and performs an instance shutdown operation.

6. A data query method based on the data integration system according to any one of claims 1-5, characterized in that, Includes the following steps: S1: Upon receiving a user's query request, generate a semantic query graph; S2: Decompose the query request into several sub-tasks based on the semantic query graph, and distribute each sub-task to its corresponding target data source node; S3: Perform hierarchical hash verification on the subtask query results returned by the target data source node. If the verification passes, merge the query results of each subtask into the final query result and return it to the user. If the verification fails, trigger a retry operation and re-execute steps S1-S3 until the verification passes or the preset number of retries is reached. Then, record the evidence storage exception in the blockchain and provide feedback to the user.

7. The data query method of the data integration system according to claim 6, characterized in that, In step S3, when the target data source node executes a subtask, it first filters invalid data according to the subtask, and then uses homomorphic encryption technology to encrypt the local data; when multiple target data source nodes perform joint data processing, a secure multi-party computation method is used to share intermediate results.

8. The data query method of the data integration system according to claim 6, characterized in that, In step S2, when distributing subtasks to the corresponding target data source nodes, the method for selecting target data source nodes related to the subtasks is: global knowledge graph index, distributed hash table, or contract event listening based on predefined rules in the smart contract.

9. The data query method of the data integration system according to claim 6, characterized in that, In step S2, when distributing subtasks to the corresponding target data source nodes, a routing path is dynamically selected using a reinforcement learning algorithm based on the semantic similarity between the subtasks and the target data source nodes and the node load.

10. A semantic self-evolution method based on the data integration system according to any one of claims 1-5, characterized in that, Includes the following steps: A semantic proxy network is deployed for the newly added data source node C, and data patterns are extracted and a knowledge graph is generated through its own knowledge graph processing module. When a newly added data source node C sends a semantic discovery request to all data source nodes, each data source node that receives the semantic discovery request analyzes the data patterns of data source node C through federated learning to identify potential semantic associations. Based on the identified potential semantic associations, semantic mapping rules are generated through a consensus algorithm, and the consensus results are broadcast to all data source nodes, triggering all data nodes to update the knowledge graph cache.

Images