Data fusion method, device and equipment based on dynamic atlas remodeling and medium

Abstract

The invention relates to the field of artificial intelligence, can be applied to business system platforms of finance, medical health and the like, and discloses a data fusion method, device, equipment and medium based on dynamic atlas remodeling, and the method comprises the following steps: carrying out semantic perception and extraction on each data source to obtain data entities and attributes; according to data entities and attributes, constructing a corresponding dynamic logic view map, and performing map fusion remodeling of semantic association; generating a corresponding global logic view based on the remodeled dynamic logic view map; receiving a query request submitted by a user based on the global logic view, and splitting and rewriting the query request to obtain a plurality of sub-query requests; and executing all the sub-query requests, fusing query results of the sub-query requests, and returning a fused query result to the user. By constructing a dynamic logic view map and remodeling the map, a global logic view is flexibly generated after dynamic fusion based on semantic perception, and the flexibility and reliability of data fusion and query are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to data fusion methods, apparatus, devices and media based on dynamic graph reshaping. Background Technology

[0002] With the advent of the big data era, data sources within and between enterprises and organizations are experiencing explosive growth. These data sources are often structurally heterogeneous, semantically inconsistent, and dynamically changing. For example, in the medical field, the data source structures and semantics of hospitals, disease control centers, and pharmaceutical companies differ significantly. The semantics of data from different institutions lack a unified standard, and the data changes dynamically with the patient's treatment process, the development of the epidemic, and the progress of research and development. Similarly, in the financial field, the data sources of banks, securities firms, and insurance companies also exhibit significant heterogeneity and dynamism. Different institutions have different semantic standards and need to obtain dynamically updated data such as market fluctuations, exchange rate changes, and policy regulatory documents in real time.

[0003] Traditional data fusion solutions, such as data warehouses and ETL tools, require extensive schema mapping and semantic alignment before fusion, focusing on syntactic and structural matching. This not only results in rigid integration methods but also lowers the accuracy of the fusion results, reducing the flexibility and reliability of data fusion. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this invention is to provide a data fusion method, apparatus, device and medium based on dynamic graph reshaping that can be applied to the financial field, medical field or other related fields. Its main purpose is to realize semantically self-aware dynamic data fusion and improve the flexibility and reliability of data fusion.

[0005] The technical solution of the present invention is as follows: The first aspect of this invention provides a data fusion method based on dynamic map reshaping, comprising: Connect to several data sources to be processed, perform semantic awareness and extraction on each data source, and obtain the data entities and attributes in each data source; Based on the data entities and attributes in each data source, a corresponding dynamic logical view graph is constructed and semantically associated graph fusion and reshaping are performed to obtain the reshaped dynamic logical view graph. Generate a corresponding global logical view based on the reshaped dynamic logical view graph; Receive a query request submitted by a user based on the global logical view, and split and rewrite the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests; Execute all subquery requests, merge the query results of the subquery requests, and return the merged query result to the user.

[0006] A second aspect of the present invention provides a data fusion apparatus based on dynamic map reshaping, comprising: The semantic awareness module is used to connect to several data sources to be processed, perform semantic awareness and extraction on each data source, and obtain the data entities and attributes in each data source. The graph construction and reshaping module is used to construct a corresponding dynamic logical view graph based on the data entities and attributes in each data source and to perform semantic association graph fusion and reshaping to obtain the reshaped dynamic logical view graph. A global logical view generation module is used to generate a corresponding global logical view based on the reshaped dynamic logical view graph. The query rewriting module is used to receive query requests submitted by users based on the global logical view, and to split and rewrite the query requests according to the reshaped dynamic logical view graph to obtain several sub-query requests. The query execution module is used to execute all sub-query requests, merge the query results of the sub-query requests, and return the merged query results to the user.

[0007] A third aspect of the present invention provides a computer device including at least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the above-described data fusion method based on dynamic map reshaping.

[0008] A fourth aspect of the present invention provides a non-volatile computer-readable storage medium storing computer-executable instructions, which, when executed by one or more processors, cause the one or more processors to perform the above-described data fusion method based on dynamic graph reshaping.

[0009] Beneficial Effects: This invention discloses a data fusion method, apparatus, device, and medium based on dynamic graph reshaping. Compared to existing technologies, this invention accesses several data sources to be processed, performs semantic awareness and extraction on each data source to obtain data entities and attributes in each data source; constructs a corresponding dynamic logical view graph based on the data entities and attributes in each data source and performs semantic association graph fusion and reshaping to obtain a reshaped dynamic logical view graph; generates a corresponding global logical view based on the reshaped dynamic logical view graph; receives a query request submitted by a user based on the global logical view, splits and rewrites the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests; executes all sub-query requests, and returns the fused query result to the user after fusing the query results of the sub-query requests. By constructing corresponding dynamic logical view graphs and performing graph reshaping after semantic awareness of different data sources, dynamic fusion based on semantic awareness is achieved. This allows for flexible generation of a global logical view based on the semantic association of different data sources, effectively improving the flexibility and reliability of data fusion and querying. Attached Figure Description

[0010] To more clearly illustrate the solutions in this invention, the accompanying drawings used in the description of the embodiments of this invention will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of an application environment for the data fusion method based on dynamic map reshaping provided in an embodiment of the present invention; Figure 2 A flowchart of a data fusion method based on dynamic map reshaping provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the functional modules of the data fusion device based on dynamic map reshaping provided in an embodiment of the present invention; Figure 4 A schematic diagram of the hardware structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention is further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. The embodiments of the invention are described below in conjunction with the accompanying drawings.

[0013] The data fusion method based on dynamic map reshaping provided in this invention can be applied to, for example... Figure 1In the application environment, it includes a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0014] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (for example only).

[0015] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0016] Server 105 can be a server providing various services, such as a backend server supporting the content browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend server can analyze and process received user requests and other data, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices. Server 105 can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the shortcomings of traditional physical hosts and VPS services ("Virtual Private Server", or simply "VPS"), such as high management difficulty and weak business scalability. Server 105 can also be a server for a distributed system or a server combined with blockchain.

[0017] It should be noted that the data fusion method based on dynamic map reshaping provided in this application embodiment can generally be executed by the first terminal device 101, the second terminal device 102, or the third terminal device 103. Correspondingly, the data fusion device based on dynamic map reshaping provided in this embodiment can also be located in the first terminal device 101, the second terminal device 102, or the third terminal device 103. Alternatively, the data fusion method based on dynamic map reshaping provided in this embodiment can generally be executed by the server 105. Correspondingly, the data fusion device based on dynamic map reshaping provided in this embodiment can generally be located in the server 105.

[0018] It should be understood that the number of terminal devices, networks, and servers listed above is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be used.

[0019] like Figure 2 As shown, the data fusion method based on dynamic map reshaping provided in this embodiment of the invention specifically includes the following steps: S201. Connect to several data sources to be processed, perform semantic awareness and extraction on each data source, and obtain the data entities and attributes in each data source.

[0020] In this embodiment, by configuring corresponding access parameters (such as database connection address, API key, access path, etc.), a stable connection with each data source is established, and several data sources to be processed are accessed for data fusion processing. The data sources to be processed may include structured data, semi-structured data, and unstructured data. Structured data includes policy tables and account tables in relational databases; semi-structured data includes health record data in JSON format and transaction records in XML format; unstructured data includes documents and reports containing customer information.

[0021] Next, semantic awareness and extraction processing are performed on each connected data source. Specifically, the data structure of the data sources is parsed, and information is extracted from each data source using natural language processing techniques such as existing NLP models to obtain data entities and associated attributes. Specifically, for structured data sources, metadata information such as table structure, field types, and primary key constraints is automatically parsed to identify corresponding data entities (such as "insurance customer," "bank account," and "platform user") and their associated attributes (such as "policy number," "account balance," and "ID number"). For semi-structured data sources, entity and attribute information is extracted by parsing tags and key-value pair structures; for example, attributes such as "user ID," "blood pressure value," and "consultation date" are extracted from JSON-formatted health records. For unstructured data sources, key entities and attributes are extracted from documents using natural language processing technologies such as text segmentation, entity recognition, and relation extraction; for example, information such as "customer name," "risk preference," and "demand description" is extracted from customer survey reports. By automatically parsing the structure and semantics of heterogeneous data sources, core entities and attributes are quickly extracted without manual preprocessing and semantic alignment, reducing the time cost of data source integration.

[0022] For example, in the cross-business data integration scenario of the group, the "Insurance Customer Policy Table" of Ping An Life Insurance, the "Bank Account and Transaction Record Table" of Ping An Bank, and the "User Health Record Table" of Ping An Good Doctor are deeply integrated to build a "360-degree panoramic view of customers" for precise marketing, personalized health insurance product recommendations, and comprehensive customer risk profiles.

[0023] After accessing data from Ping An Life Insurance, Ping An Bank, and Ping An Good Doctor, semantic perception and extraction are used to extract the entity "insurance customer" and its attributes "policy number, policyholder name, insured's ID number, insurance type, annual premium, and health status disclosure" from the insurance customer policy table; the entity "bank customer" and its attributes "customer number, customer name, ID number, account balance, monthly transaction history, and consumption habit tags" from the bank account and transaction record table; and the entity "platform user" and its attributes "user ID, name, ID number, historical consultation records, health indicators, and medication purchase records" from the user health record table.

[0024] S202. Based on the data entities and attributes in each data source, construct the corresponding dynamic logical view graph and perform semantic association graph fusion and reshaping to obtain the reshaped dynamic logical view graph.

[0025] In this embodiment, the dynamic logical view graph is a graph structure model used to represent the semantic association of data. The self-aware fusion of data semantics is achieved through the construction and reshaping process of the dynamic logical view graph. First, based on the data entities and attributes extracted from each data source, a corresponding dynamic logical view graph is constructed. In the graph, nodes represent data entities from different data sources (such as "user", "product", "order"), and nodes have corresponding attributes (such as "name", "production batch"). Edges represent the semantic association relationships between entities / attributes (such as "equivalent to", "belong to", "partially overlap"). Each association edge is also attached with a quantifiable semantic connection weight. This weight is dynamically calculated based on factors such as semantic similarity calculation, frequency of co-occurrence in historical queries, and user feedback, and is presented in a visual form (such as line thickness and color depth), thus initially forming an initial graph containing semantic associations from multiple data sources.

[0026] After constructing the dynamic logical view graph, the deeper semantic relationships between different nodes and their attributes are further identified, thereby reshaping the nodes and edges in the graph. This includes merging nodes with the same semantics from different data sources, mining semantic relationships between attributes across data sources to establish new edges, and so on. This completes the graph reshaping, ensuring that the graph can optimally represent the current global data semantics. Semantic relationship mining eliminates data semantic heterogeneity, forming a globally unified semantic representation. Simultaneously, the dynamic reshaping mechanism ensures that the graph can adapt to data changes, solving the rigidity problem of traditional integration solutions.

[0027] For example, based on entities and attributes extracted from three data sources—Ping An Life Insurance, Ping An Bank, and Ping An Good Doctor—the graph is constructed with "Insurance Customer," "Bank Customer," and "Platform User" as independent nodes. Semantically equivalent association edges are established between attributes such as "Insured's ID Number - ID Document Number" and "ID Document Number - ID Number." The semantic connection weights of these edges, calculated based on semantic similarity, historical query frequency of collaboration, and user feedback, are, for example, all 0.99. The graph is then fused and reshaped through semantic association. Since the semantic connection weights between the Insured's ID Number (Life Insurance), ID Document Number (Bank), and ID Number (Medical) are extremely high, representing semantic equivalence, the three nodes—Insurance Customer, Bank Customer, and Platform User—are merged into a single logical entity, "Ping An Customer," representing the same natural person in the real world. The merged "Ping An Customer" entity possesses rich attributes from the three data sources. By further mining the semantic relationships between the merged attributes, corresponding edges are established and assigned semantic connection weights, ultimately forming a reshaped dynamic logical view graph containing the core node "Ping An Customer" and multi-dimensional semantic relationships.

[0028] S203. Generate a corresponding global logical view based on the reshaped dynamic logical view graph.

[0029] In this embodiment, the global logical view serves as a unified data access point for users. Users do not need to concern themselves with the physical location, storage format, and original structure of the underlying data source; they can submit query requests simply through this view. Based on the reshaped dynamic logical view graph, the system automatically parses the entity fusion results, attribute relationships, and semantic hierarchy within the graph, extracts global data patterns, and generates corresponding global logical views. The final global logical view is presented in an intuitive hierarchical structure, providing users with a unified and intuitive data access point. This shields users from differences in the physical location and storage format of the underlying data sources, eliminating the need for users to concern themselves with the underlying data details and lowering the barrier to entry for cross-data source queries.

[0030] For example, a global logical view called "Ping An Customer Panoramic View" is generated after parsing the reshaped dynamic logical view graph. This global logical view takes "Ping An Customer" as the root node and has four sub-nodes: "Policy Information," "Account Information," "Health Information," and "Consumption Preferences." Each sub-node corresponds to the integrated relevant attributes, forming a complete global logical view that allows marketing personnel and underwriters to directly query the cross-domain data they need through this view.

[0031] S204. Receive a query request submitted by a user based on the global logical view, and split and rewrite the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests.

[0032] In this embodiment, users can submit corresponding query requests based on a global logical view through a terminal device. These requests can take various forms, such as SQL queries, graphical interface filtering requests, or natural language query text. Upon receiving a query request, the parsed query information is split and mapped to the corresponding underlying physical data source based on the semantic mapping relationships in the reshaped dynamic logical view graph. This rewrites the original query request into a series of specific sub-queries targeting the underlying physical data source. This query rewriting process automatically decomposes the global query into sub-queries adapted to each underlying data source, decoupling the query logic from the physical data source, improving the flexibility and adaptability of query execution, and avoiding the tedious operation of manually writing multi-data source query statements.

[0033] For example, if a user wants to query "the bank asset overview of customers with an annual premium of over 50,000 yuan and a history of hypertension", then "annual premium" corresponds to the insurance customer policy table of Ping An Life Insurance, "history of hypertension" corresponds to the user health record table of Ping An Good Doctor, and "bank asset overview" corresponds to the bank account and transaction record table of Ping An Bank. Based on the reshaped dynamic logical view graph, the original query request is rewritten into three sub-query requests targeting different data sources. Each sub-query request includes information such as the access path, query fields, and filtering conditions of the corresponding data source.

[0034] S205. Execute all subquery requests, merge the query results of the subquery requests, and return the merged query results to the user.

[0035] In this embodiment, the query execution engine preferably adopts a distributed architecture, simultaneously sending corresponding subquery requests to each underlying physical data source and coordinating the execution progress of each subquery to ensure query efficiency. The returned subquery results undergo multi-dimensional fusion processing, including data alignment and conflict resolution. Data alignment involves uniformly organizing the query results returned from different data sources according to the field specifications of the global logical view. Conflict resolution handles different values ​​for the same attribute; for example, when the name of the same customer differs in different data sources, priority is determined based on semantic join weight and data credibility to select the optimal value. Through distributed execution and semantic-level result fusion, the fused query results are returned to the user after the fusion processing is complete, achieving an integrated query experience across data sources, providing users with complete and consistent query results, realizing penetrating data fusion queries, and improving data query and usage efficiency.

[0036] In one embodiment, step S202 includes: Initialize the graph structure by using the data entities as nodes and the semantic relationships between the attributes of different data entities as edges, construct an initial dynamic logical view graph and calculate the semantic connection weight of each edge; Initiate the graph fusion and reshaping mechanism to identify semantically equivalent nodes in the initial dynamic logical view graph based on the semantic connection weights; The corresponding semantically equivalent nodes are fused to obtain fused nodes, and all attributes of the fused nodes are aligned and integrated. Based on the semantic associations between all attributes of the integrated nodes, corresponding edges are established and the semantic connection weight of each edge is calculated to obtain the reshaped dynamic logical view graph.

[0037] In this embodiment, when constructing and reshaping the dynamic logical view graph, the graph structure is first initialized, and each data entity extracted from each data source is added to the graph as an independent node. The semantic relationships between the attributes of different data entities are identified through semantic similarity calculation, domain rule matching, and other methods. Based on the relationships, edges (such as "equivalent to", "belong to", "partially overlapping") are established between nodes. For example, the "insured person's ID number" attribute of the "insurance customer" entity and the "ID card number" attribute of the "bank customer" entity have a semantically equivalent relationship. Therefore, an "equivalent to" edge can be established between the two nodes to accurately represent the semantic relationship between entities and attributes through the graph. At the same time, the corresponding semantic relationship weights are calculated to quantify the semantic relationship of each edge. Specifically, the weights can be dynamically calculated based on factors such as semantic similarity calculation, frequency of historical query collaboration, and user feedback, and presented in a visual form (such as line thickness and color depth).

[0038] The graph fusion and reshaping mechanism is then activated. Based on preset semantic equivalence judgment rules and semantic connection weights, semantically equivalent nodes in the graph are identified. For example, a weight threshold (e.g., 0.9) is set. When two nodes have an edge with a weight greater than or equal to the threshold indicating they are "equivalent," these two nodes are determined to be semantically equivalent nodes. Simultaneously, secondary verification can be performed using auxiliary information such as the core identifier fields of entities and attribute set similarity to ensure the accuracy of equivalence node identification. For instance, in the initial graph, among the nodes "Insurance Customer," "Bank Customer," and "Platform User," the edge weights corresponding to "Insured's ID Number - ID Document Number," "ID Document Number - ID Number," and "Insured's ID Number - ID Number" are all greater than 0.9, and the core identifier fields of all three entities point to the user's unique identity identifier. Furthermore, there is significant overlap in the attribute sets; therefore, these three nodes are determined to be semantically equivalent nodes.

[0039] The identified semantically equivalent nodes are fused together. For example, “insurance customer”, “bank customer”, and “platform user” are fused into the “Ping An customer” node. During the fusion, the attributes of all semantically equivalent nodes are merged into the fused node, and all attributes of the fused node are aligned and integrated to eliminate duplicate attributes while covering attributes from different data sources, forming an attribute set containing all semantic information.

[0040] After the attribute integration of the fusion nodes is completed, the deep semantic relationships between all attributes are further explored, including causal relationships, corroborative relationships, and complementary relationships. Corresponding edges are established based on different relationship types. For example, there is a "payment capacity association" edge between "annual premium" and "account balance," a "mutual corroboration" edge between "health status disclosure" and "health indicators," and a "product preference" edge between "consumption habit tags" and "insurance type," etc. Similarly, the semantic connection weight of each edge is calculated based on factors such as semantic similarity, the frequency of attribute co-occurrence in historical queries, and user feedback correction information to reflect the semantic relationships between attributes of different nodes. After all the associated edges and weights are constructed, a reshaped dynamic logical view graph with optimized structure and accurate semantics is formed. By reshaping the initial graph through graph node fusion and semantic relationship mining, data redundancy is eliminated, and the dynamic logical view graph can better adapt to changes in data sources, improving the flexibility of data fusion.

[0041] In one embodiment, the step of fusing the corresponding semantically equivalent nodes to obtain a fused node, and aligning and integrating all attributes of the fused node, includes: Merge semantically equivalent nodes that correspond to the same data entity, and use logical entities with unified semantics as fusion nodes; Align all attributes of the merged semantically equivalent nodes, and identify common and unique attributes among all attributes based on the alignment results; The common attributes are merged, and the unique attributes are integrated and assigned to the fusion node.

[0042] In this embodiment, during node fusion, semantically equivalent nodes corresponding to the same data entity are merged. Specifically, a dual determination can be made based on the uniqueness of the identifier field and semantic consistency. The identifier field can include unique fields such as ID card number, unique user ID, and unified social credit code of enterprise. When the identifier fields of different nodes are semantically equivalent and the weight meets the threshold requirement, they are determined to correspond to the same data entity, which means that these nodes correspond to the same data object in the real world. By fusing nodes with the same data entity, the duplicate existence of semantically equivalent nodes is avoided, and the redundancy of graph data is eliminated.

[0043] Next, logical entities with unified semantics are created as fusion nodes. The fused logical entities will have rich attributes from all participating fusion nodes. These attributes are aligned to identify common and unique attributes. Specifically, when two attributes have the same semantic label, match data types, and have an overlap in value range exceeding a preset threshold (e.g., 80%), they are determined to be common attributes. For example, the semantic label of "policyholder name", "customer name", and "name" are all "username", the data types are all string types, and the value ranges completely overlap, so they are determined to be common attributes. On the other hand, when an attribute exists only in one original node and the semantic similarity of all attributes in other nodes is lower than a preset threshold (e.g., 30%), it is determined to be a unique attribute. For example, "insurance type" exists only in the original "insurance customer" node, and "account balance" exists only in the original "bank customer" node, both of which are determined to be unique attributes. Through attribute alignment, the categories of common and unique attributes are clearly distinguished.

[0044] Based on the attribute recognition results, all common attributes are merged. For example, "Insured's Name", "Customer's Name", and "Name" are uniformly named "Customer's Name". Unique data is integrated, and the unique attributes of all semantically equivalent nodes are fully attributed to the fusion node. For example, unique attributes such as insurance type, account balance, and health indicators are integrated into the fusion node. Finally, the fusion node forms an attribute set containing unified common attributes and complete unique attributes, which not only ensures the integrity of the data but also improves the uniformity of the data, making the fused graph more in line with the global logic of the business scenario.

[0045] In one embodiment, the step of establishing corresponding edges based on the semantic associations between all attributes of the integrated nodes and calculating the semantic connection weight of each edge to obtain the reshaped dynamic logical view graph includes: Semantic feature extraction is performed on all attributes of the integrated node to obtain feature labels; Based on the feature labels of all attributes, perform label matching and statistical association analysis to identify the association patterns between pairs of attributes and establish corresponding edges; The semantic connection weight of each edge is calculated based on the semantic similarity, collaboration frequency, and user correction coefficient between the attributes associated with each edge, thus obtaining the reshaped dynamic logical view graph.

[0046] In this embodiment, when further mining the deep semantic relationships between attributes based on all attributes of the fusion node to achieve graph reshaping, semantic features are first extracted from all attributes attached to the fusion node. For example, core feature labels can be extracted through pre-trained financial domain semantic models (such as BERT-Finance), including data type labels (such as "text description", "numerical", "time", "classification label"), business semantic labels (such as "health status", "cash flow", "product holding", "consumption behavior", "identity"), and functional usage labels (such as "risk assessment", "payment verification", "demand preference", "information verification"), etc., to provide reliable feature basis for accurately identifying potential relationships between attributes.

[0047] Then, based on the feature labels of all attributes, label matching and statistical association analysis are performed to identify the association patterns between pairs of attributes and establish corresponding edges. Specifically, the matching degree of the feature labels of each pair of attributes can be calculated first. When the matching degree exceeds a preset threshold (such as 50%), it is preliminarily determined that there is a semantic association. Then, combined with historical query logs, business scenario rules, etc., the frequency of collaborative occurrence of each pair of attributes in the query and the association relationship in the business process are statistically analyzed to further verify the effectiveness of the association. Based on the association analysis results, corresponding association patterns are identified according to preset association rules. For example, if the business semantic tags of the two are consistent (e.g., both are "health status") and the data type is text description, it is determined to be the "information complement / mutual confirmation" pattern; if the business semantic tags of the two are a combination of fund-related and product-related (e.g., "cash flow" + "product holding") and there is a numerical / time correlation, it is determined to be the "capability and behavior matching" pattern; if the business semantic tags of the two are a combination of behavioral characteristics and product-related (e.g., "consumption behavior" + "product holding") and there is a classification mapping relationship, it is determined to be the "preference product matching" pattern; if there is a time series correlation (e.g., "insurance time" + "physical examination time"), it is determined to be the "time series correlation" pattern. Specifically, the corresponding preset association rules can be flexibly set according to actual needs for identification, and this embodiment does not limit this.

[0048] The semantic connection weights are calculated based on the semantic similarity, collaboration frequency, and user correction coefficient between the attributes associated with each edge. Semantic similarity is calculated using algorithms such as cosine similarity and Jaccard similarity coefficient, with a value ranging from 0 to 1. Collaboration frequency refers to the percentage of times each pair of attributes appears simultaneously in historical queries, also ranging from 0 to 1. The user correction coefficient reflects the user's manual confirmation or correction feedback regarding the association. If the user confirms the association is valid, the coefficient is 1; if the user corrects the association strength, the coefficient is adjusted proportionally (e.g., changing a strong association to a medium association results in a coefficient of 0.6). The coefficient without user correction is 1 by default. (The text abruptly ends here, likely due to an incomplete sentence or missing information.) After calculating the degree, collaboration frequency, and user correction coefficient, a weighted calculation is performed to obtain the final semantic connection weight. For example, the semantic similarity between "annual premium" and "account balance" is 0.6, the collaboration frequency is 0.7, and there is no user correction coefficient (coefficient 0.8). Therefore, the weight value is 0.5×0.6+0.3×0.7+0.2×1=0.71. After the weights of all edges are calculated, the graph structure is updated to form a reshaped dynamic logical view graph. The quantitative weights reflect the correlation strength between nodes in the reshaped graph, enhancing the semantic expressiveness and interpretability of the graph. At the same time, it provides a quantitative basis for conflict resolution and semantic fusion of query results.

[0049] For example, after merging the three nodes of insurance customers, bank customers, and platform users into a single node called Ping An Customer, deep semantic association mining is performed based on the integrated attributes to establish new edges and assign corresponding weights, for example: Establish a "mutually corroborating" edge between health status disclosures (life insurance) and historical medical records (medical), with weights based on text similarity analysis (e.g., "history of hypertension" in insurance disclosures and "diagnosis of hypertension" in medical records are highly matched, with a weight of 0.95). Establish a "payment capacity association" edge between monthly transaction records (bank) and annual premiums (life), with the weight calculated by analyzing the consistency between large expenditures in bank statements and premium deduction dates; Establish a "product preference" edge between consumption habit tags (banks) and insurance types (life insurance). The weight is obtained by using a machine learning model to mine association rules from historical data (for example, the "high-frequency travel" consumption tag is strongly correlated with the purchase record of "travel accident insurance", with a weight of 0.88).

[0050] In one embodiment, step S203 includes: The reshaped dynamic logical view graph is semantically aggregated using a graph neural network to generate a global semantic feature vector. Based on the global semantic feature vector, the hierarchical structure in the global data pattern is parsed and output, and the hierarchical relationship between the root node, child nodes, and attributes is determined. The global logical view is generated after field standardization processing based on the hierarchical structure of the global data model.

[0051] In this embodiment, the Graph Neural Network (GNN) adopts a Graph Convolutional Neural Network (GCN) or Graph Attention Network (GAT) architecture to process semantic aggregation of graph-structured data. By taking the node features and edge features (including semantic connection weights) of the reshaped dynamic logical view graph as input, multi-layer graph convolution operations are used to fuse the local semantic features of the nodes with the structural features of the global graph, realizing the semantic hierarchical aggregation of the dynamic logical view graph and generating a global semantic feature vector. This vector contains the core information of all entities, attributes, and semantic relationships, providing a foundation for subsequent global data pattern parsing.

[0052] Next, the global semantic feature vector is parsed and combined with domain business logic to output the hierarchical structure of the global data model. Specifically, the global semantic feature vector is first clustered, and feature clusters with high semantic similarity are divided into the same level or category. Then, combined with domain business logic, a hierarchical structure from core entities to subdivided attributes is constructed, confirming the hierarchical relationship between root nodes, child nodes, and attributes in the hierarchical structure. For example, the top level of the hierarchical structure is the root node, which is usually the merged logical entity (such as "Ping An Customer"). Several child nodes are set under the root node, and the child nodes are divided according to business domain or attribute category (such as "Policy Information", "Account Information", "Health Information", "Consumption Preference", "Risk Information", etc.). Each child node is attached with a corresponding specific attribute, forming a three-level hierarchical structure of "root node-child node-attribute". Through hierarchical structure parsing, the global data model is made clear and orderly, which is convenient for users to understand and query.

[0053] Based on the hierarchical structure of the global data model, further field standardization is performed, specifically including field naming standardization, data type standardization, value range standardization, and field description standardization. Attributes from different data sources are standardized in multiple dimensions. After standardization, the hierarchical structure and standardized fields are organized into a user-accessible global logical view, which can take various display forms, including relational views, tree views, and form views. Users can choose the appropriate view format for data access according to their needs. The generated global logical view possesses both global semantic integrity and a clear hierarchical structure and standardized fields, improving the view's usability and versatility, enabling different business users to quickly locate the data they need, and reducing the difficulty of query operations.

[0054] In one embodiment, step S204 includes: Receive user query requests for the global logical view, perform semantic parsing on the query requests, and extract the corresponding query fields and filtering conditions; Based on the semantic mapping relationship in the reshaped dynamic logical view graph, the query fields and filtering conditions are split and mapped into sub-query fields and sub-filtering conditions corresponding to the physical data source; The query request is rewritten into several corresponding sub-query requests based on the sub-query fields and sub-filter conditions.

[0055] In this embodiment, users can submit query requests for the global logical view through various query methods, such as SQL queries or natural language queries. The query requests are semantically parsed to extract the corresponding query fields and filtering conditions. For example, if a user submits the SQL query "SELECT customer name, bank asset overview, health indicators FROM Ping An customer panoramic view WHERE annual premium > 50000 AND health status disclosure LIKE '%hypertension%'", the parsed query fields are "customer name, bank asset overview, health indicators", and the filtering conditions are "annual premium > 50000" and "health status disclosure includes 'hypertension'". Similarly, for the natural language query "query customer names and account balances who frequently shop online and have purchased critical illness insurance", the parsed query fields are "customer name, account balance", and the filtering conditions are "consumption habit tag = high-frequency online shopping" and "insurance type = critical illness insurance".

[0056] Based on the semantic mapping relationships in the reshaped dynamic logical view graph (i.e., the correspondence between global logical fields recorded in the graph and the fields of the underlying physical data source), query fields and filtering conditions are split and mapped into sub-query fields and sub-filtering conditions corresponding to the physical data source. Specifically, the physical data source and physical field name corresponding to each query field are first determined; for example, "Bank Asset Overview" corresponds to the "Account Balance + Monthly Transaction Flow" combined field in Ping An Bank's "Bank Account and Transaction Record Table", and "Health Indicators" corresponds to fields such as "Blood Pressure" and "Blood Sugar" in Ping An Good Doctor's "User Health Record Table". Then, the filtering conditions are converted into sub-filtering conditions that conform to the syntax rules of the data source. For example, the filtering condition "Annual Premium > 50,000" is split and mapped to "Annual Premium > 50,000" in Ping An Life's "Insurance Customer Policy Table"; "Health Status Disclosure Includes 'Hypertension'" is split and mapped to "Historical Consultation Records LIKE '%Hypertension%'" in Ping An Good Doctor's "User Health Record Table". Based on the splitting results, the original query request is rewritten into several corresponding sub-query requests, which then initiate corresponding queries on the physical data sources. Each rewritten sub-query request can be executed independently on its corresponding physical data source, obtaining partial query results before subsequent fusion and feedback. Through accurate mapping and splitting of query logic, and adaptation to the query syntax of different types of underlying data sources, joint data queries from multiple data sources are achieved, improving data efficiency.

[0057] In one embodiment, after step S202, the method further includes: Real-time monitoring of new data source access events; when a new data source is accessed, extract the corresponding newly added entities and attributes. Based on the newly added entities and attributes, the current reshaped dynamic logical view graph is further reshaped by semantic association through graph fusion, and the global logical view is updated.

[0058] In this embodiment, new data source access events are monitored in real time. By listening to changes in metadata at the data access port, the type, structure, and access time of the new data source are identified. After the new data source is accessed, the semantic awareness and extraction process is automatically initiated, using the same method as in the previous embodiment to extract the new entities and attributes from the new data source. For example, when Ping An Group subsequently accesses Ping An Property & Casualty Insurance's "Vehicle Violation Data Table," semantic awareness and extraction are used to extract the new entity "Vehicle Owner" and the attributes "Vehicle Owner Name, Vehicle Owner ID Number, License Plate Number, Violation Record, and Violation Time." Simultaneously, metadata such as the source information, data format, and update frequency of the new entities and attributes are recorded to provide a basis for subsequent graph reshaping.

[0059] Simultaneously, after connecting to the new data source, the graph fusion and reshaping mechanism is restarted. Based on the newly added entities and attributes, the currently reshaped dynamic logical view graph is again subjected to semantic association graph fusion and reshaping. The specific graph fusion and reshaping process is consistent with the reshaping logic in the above embodiment, the difference being that when the new data source is connected, the focus is on the integration of newly added entities and attributes. Specifically, the newly added entities can be added as candidate nodes to the current dynamic logical view graph first, and then the semantic association weight between the newly added entities and all nodes in the existing graph is calculated to identify whether there are semantically equivalent nodes; for example, the "vehicle owner's ID number" of the "vehicle owner" entity is semantically equivalent to the "ID number" of the existing "Ping An customer" entity, the weight is calculated to be 0.99, and it is determined to be a semantically equivalent node. Subsequently, semantically equivalent nodes are merged, integrating the attributes of newly added entities (such as license plate numbers and traffic violation records) into existing fusion nodes (Ping An customers). Semantic relationships between new and existing attributes are then mined, establishing new association edges and calculating weights. For example, an association edge is established between traffic violation records (such as speeding) and insurance types (auto insurance). Based on the correlation between traffic violation records and auto insurance, corresponding semantic association weights are calculated, and the customer's auto insurance premium risk score is dynamically adjusted. The updated dynamic logical view graph is used to update the global logical view accordingly, attaching new attributes to corresponding child nodes according to business categories (e.g., attaching "traffic violation records" to the "risk information" child node). This achieves dynamic updates to the global logical view, ensuring that user queries cover information from newly added data sources. Through this dynamic evolution graph data fusion method, new data sources can be "plug-and-play," automatically integrating into the existing data system without manual adjustments to the integration logic. This continuously enriches the global data dimensions, ensuring the system can adapt to dynamically changing data environments, extending the system's lifecycle and reducing long-term maintenance costs.

[0060] It should be noted that there is no necessary order between the above steps. Those skilled in the art will understand from the description of the embodiments of the present invention that the above steps may have different execution orders in different embodiments, that is, they may be executed in parallel or in turn, etc.

[0061] Further reference Figure 3 As a response to the above Figure 2 The present invention provides an embodiment of a data fusion device based on dynamic map reshaping, which implements the method shown. Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0062] like Figure 3 As shown, the data fusion device 30 based on dynamic map reshaping described in this embodiment includes: The semantic awareness module 301 is used to access several data sources to be processed, perform semantic awareness and extraction on each data source, and obtain the data entities and attributes in each data source. The graph construction and reshaping module 302 is used to construct a corresponding dynamic logical view graph based on the data entities and attributes in each data source and to perform semantic association graph fusion and reshaping to obtain the reshaped dynamic logical view graph. The global logical view generation module 303 is used to generate a corresponding global logical view based on the reshaped dynamic logical view map. The query rewriting module 304 is used to receive a query request submitted by a user based on the global logical view, and to split and rewrite the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests. The query execution module 305 is used to execute all sub-query requests, merge the query results of the sub-query requests, and return the merged query results to the user.

[0063] The module referred to in this invention is a series of computer program instruction segments that can perform specific functions. It is more suitable than a program for describing the data fusion execution process based on dynamic graph reshaping. For the specific implementation of each module, please refer to the corresponding method embodiments above, which will not be repeated here.

[0064] In one embodiment, the map construction and reshaping module 302 includes: An initial construction unit is used to initialize the graph structure, using the data entities as nodes and the semantic relationships between the attributes of different data entities as edges, to construct an initial dynamic logical view graph and calculate the semantic connection weight of each edge. The node identification unit is used to initiate the graph fusion and reshaping mechanism and identify semantically equivalent nodes in the initial dynamic logical view graph based on the semantic connection weights. The node fusion unit is used to perform fusion processing on the corresponding semantically equivalent nodes to obtain fused nodes, and to align and integrate all the attributes of the fused nodes; The graph reshaping unit is used to establish corresponding edges based on the semantic associations between all attributes of the integrated nodes and to calculate the semantic connection weight of each edge, thereby obtaining the reshaped dynamic logical view graph.

[0065] In one embodiment, the node fusion unit includes: The fusion unit is used to merge semantically equivalent nodes that correspond to the same data entity, with logical entities of unified semantics as fusion nodes; The attribute alignment unit is used to align all attributes of the merged semantically equivalent nodes and to identify common and unique attributes among all attributes based on the alignment results. The attribute integration unit is used to merge the general attributes and integrate the unique attributes and assign them to the fusion node.

[0066] In one embodiment, the map reshaping unit includes: The feature label extraction unit is used to extract semantic features from all attributes of the integrated fusion node to obtain feature labels; The association pattern recognition unit is used to perform label matching and statistical association analysis based on the feature labels of all attributes, identify the association patterns between pairs of attributes and establish corresponding edges; The connection weight calculation unit is used to calculate the semantic connection weight of each edge based on the semantic similarity, collaboration frequency and user correction coefficient between the attributes associated with each edge, so as to obtain the reshaped dynamic logical view graph.

[0067] In one embodiment, the global logical view generation module 303 includes: The semantic aggregation unit is used to perform semantic hierarchical aggregation on the reshaped dynamic logical view graph through a graph neural network to generate a global semantic feature vector. The global structure output unit is used to parse and output the hierarchical structure in the global data pattern according to the global semantic feature vector, and to determine the hierarchical relationship between the root node, child nodes and attributes. The global logical view generation unit is used to generate the global logical view after performing field standardization processing based on the hierarchical structure of the global data model.

[0068] In one embodiment, the query rewriting module 304 includes: The query receiving unit is used to receive user query requests for the global logical view, perform semantic parsing on the query requests, and extract the corresponding query fields and filtering conditions. The query mapping splitting unit is used to split and map the query fields and filter conditions into sub-query fields and sub-filter conditions corresponding to the physical data source based on the semantic mapping relationship in the reshaped dynamic logical view graph. The query rewriting unit is used to rewrite the query request into several corresponding sub-query requests based on the sub-query fields and sub-filter conditions.

[0069] In one embodiment, the device 30 further includes: The real-time monitoring module is used to monitor new data source access events in real time, and extract the corresponding newly added entities and attributes when a new data source is accessed. The view update module is used to perform semantic association graph fusion reshaping on the currently reshaped dynamic logical view graph based on the newly added entities and attributes, and update the global logical view.

[0070] In the above embodiments, this invention discloses a data fusion device based on dynamic graph reshaping. It connects to several data sources to be processed, performs semantic awareness and extraction on each data source to obtain data entities and attributes in each data source; constructs a corresponding dynamic logical view graph based on the data entities and attributes in each data source and performs semantic association graph fusion and reshaping to obtain a reshaped dynamic logical view graph; generates a corresponding global logical view based on the reshaped dynamic logical view graph; receives a query request submitted by a user based on the global logical view, splits and rewrites the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests; executes all sub-query requests, and returns the fused query result to the user after fusing the query results of the sub-query requests. By constructing corresponding dynamic logical view graphs and performing graph reshaping after semantic awareness of different data sources, dynamic fusion based on semantic awareness is achieved. This allows for flexible generation of a global logical view based on the semantic association of different data sources, effectively improving the flexibility and reliability of data fusion and querying.

[0071] Another embodiment of the present invention provides a computer device, such as... Figure 4 As shown, the computer device 40 includes: One or more processors 401 and memory 402, Figure 4 The following section uses a processor 401 as an example. The processor 401 and the memory 402 can be connected via a bus or other means. Figure 4 Taking the example of a connection between China and Israel via a bus.

[0072] Processor 401 performs various control logic functions for computer device 40. It can be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), microcontroller, ARM (Acorn RISC Machine) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination of these components. Furthermore, processor 401 can also be any conventional processor, microprocessor, or state machine. Processor 401 can also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors combined with a DSP, and / or any other such configuration.

[0073] The memory 402, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions corresponding to the data fusion method based on dynamic map reshaping in the embodiments of the present invention. The processor 401 executes various functional applications and data processing of the computer device 40 by running the non-volatile software programs, instructions, and units stored in the memory 402, thereby realizing the data fusion method based on dynamic map reshaping in the above method embodiments.

[0074] The memory 402 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device 40. Furthermore, the memory 402 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 402 may optionally include memory remotely located relative to the processor 401, and these remote memories may be connected to the computer device 40 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. One or more units stored in the memory 402, when executed by one or more processors 401, perform the steps of the data fusion method based on dynamic graph reshaping in any of the above method embodiments.

[0075] In the above embodiments, the present invention discloses a computer device that, by accessing several data sources to be processed, performs semantic awareness and extraction on each data source to obtain data entities and attributes in each data source; based on the data entities and attributes in each data source, constructs a corresponding dynamic logical view graph and performs semantic association graph fusion and reshaping to obtain a reshaped dynamic logical view graph; generates a corresponding global logical view based on the reshaped dynamic logical view graph; receives a query request submitted by a user based on the global logical view, splits and rewrites the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests; executes all sub-query requests, and returns the fused query result to the user after fusing the query results of the sub-query requests. By performing semantic awareness on different data sources, constructing corresponding dynamic logical view graphs, and performing graph reshaping, semantic awareness-based dynamic fusion is achieved. This allows for flexible generation of a global logical view based on the semantic associations of different data sources, effectively improving the flexibility and reliability of data fusion and querying.

[0076] This invention provides a non-volatile computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are executed by one or more processors, they perform the steps of the data fusion method based on dynamic graph reshaping in any of the above method embodiments.

[0077] In the above embodiments, the present invention discloses a non-volatile computer-readable storage medium. By accessing several data sources to be processed, semantic awareness and extraction are performed on each data source to obtain data entities and attributes in each data source. Based on the data entities and attributes in each data source, a corresponding dynamic logical view graph is constructed and semantically related graph fusion and reshaping are performed to obtain a reshaped dynamic logical view graph. A corresponding global logical view is generated based on the reshaped dynamic logical view graph. A query request submitted by a user based on the global logical view is received, and the query request is split and rewritten according to the reshaped dynamic logical view graph to obtain several sub-query requests. All sub-query requests are executed, and the query results of the sub-query requests are fused and returned to the user. By performing semantic awareness on different data sources, constructing corresponding dynamic logical view graphs, and reshaping the graphs, semantically aware dynamic fusion is achieved. This allows for flexible generation of a global logical view based on the semantic relationships of different data sources, effectively improving the flexibility and reliability of data fusion and querying.

[0078] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0079] This invention can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0080] In summary, the data fusion method, apparatus, device, and medium based on dynamic graph reshaping disclosed in this invention include the following steps: accessing several data sources to be processed; performing semantic awareness and extraction on each data source to obtain data entities and attributes in each data source; constructing a corresponding dynamic logical view graph based on the data entities and attributes in each data source and performing semantic association graph fusion and reshaping to obtain a reshaped dynamic logical view graph; generating a corresponding global logical view based on the reshaped dynamic logical view graph; receiving a query request submitted by a user based on the global logical view; splitting and rewriting the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests; executing all sub-query requests; and fusing the query results of the sub-query requests and returning the fused query result to the user. By performing semantic awareness on different data sources, constructing corresponding dynamic logical view graphs, and performing graph reshaping, dynamic fusion based on semantic awareness is achieved. This allows for flexible generation of a global logical view based on the semantic associations of different data sources, effectively improving the flexibility and reliability of data fusion and querying.

[0081] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The computer program can be stored in a non-volatile, computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The storage medium can be a memory, magnetic disk, floppy disk, flash memory, optical storage, etc.

[0082] It should be noted that any software tools or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. It should be understood that the application of this invention is not limited to the examples described above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A data fusion method based on dynamic map reshaping, characterized in that, include: Connect to several data sources to be processed, perform semantic awareness and extraction on each data source, and obtain the data entities and attributes in each data source; Based on the data entities and attributes in each data source, a corresponding dynamic logical view graph is constructed and semantically associated graph fusion and reshaping are performed to obtain the reshaped dynamic logical view graph. Generate a corresponding global logical view based on the reshaped dynamic logical view graph; Receive a query request submitted by a user based on the global logical view, and split and rewrite the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests; Execute all subquery requests, merge the query results of the subquery requests, and return the merged query result to the user.

2. The data fusion method based on dynamic map reshaping according to claim 1, characterized in that, The step of constructing a corresponding dynamic logical view graph based on the data entities and attributes in each data source and performing semantic association graph fusion and reshaping to obtain the reshaped dynamic logical view graph includes: Initialize the graph structure by using the data entities as nodes and the semantic relationships between the attributes of different data entities as edges, construct an initial dynamic logical view graph and calculate the semantic connection weight of each edge; Initiate the graph fusion and reshaping mechanism to identify semantically equivalent nodes in the initial dynamic logical view graph based on the semantic connection weights; The corresponding semantically equivalent nodes are fused to obtain fused nodes, and all attributes of the fused nodes are aligned and integrated. Based on the semantic associations between all attributes of the integrated nodes, corresponding edges are established and the semantic connection weight of each edge is calculated to obtain the reshaped dynamic logical view graph.

3. The data fusion method based on dynamic map reshaping according to claim 2, characterized in that, The process of fusing the corresponding semantically equivalent nodes to obtain fused nodes, and aligning and integrating all attributes of the fused nodes, includes: Merge semantically equivalent nodes that correspond to the same data entity, and use logical entities with unified semantics as fusion nodes; Align all attributes of the merged semantically equivalent nodes, and identify common and unique attributes among all attributes based on the alignment results; The common attributes are merged, and the unique attributes are integrated and assigned to the fusion node.

4. The data fusion method based on dynamic map reshaping according to claim 2, characterized in that, The process of establishing corresponding edges based on the semantic relationships between all attributes of the integrated nodes and calculating the semantic connection weight of each edge to obtain the reshaped dynamic logical view graph includes: Semantic feature extraction is performed on all attributes of the integrated node to obtain feature labels; Based on the feature labels of all attributes, perform label matching and statistical association analysis to identify the association patterns between pairs of attributes and establish corresponding edges; The semantic connection weight of each edge is calculated based on the semantic similarity, collaboration frequency, and user correction coefficient between the attributes associated with each edge, thus obtaining the reshaped dynamic logical view graph.

5. The data fusion method based on dynamic map reshaping according to claim 1, characterized in that, The generation of a corresponding global logical view based on the reshaped dynamic logical view graph includes: The reshaped dynamic logical view graph is semantically aggregated using a graph neural network to generate a global semantic feature vector. Based on the global semantic feature vector, the hierarchical structure in the global data pattern is parsed and output, and the hierarchical relationship between the root node, child nodes, and attributes is determined. The global logical view is generated after field standardization processing based on the hierarchical structure of the global data model.

6. The data fusion method based on dynamic map reshaping according to claim 1, characterized in that, The process involves receiving a query request submitted by a user based on the global logical view, and then splitting and rewriting the query request according to the reshaped dynamic logical view graph to obtain several sub-query requests, including: Receive user query requests for the global logical view, perform semantic parsing on the query requests, and extract the corresponding query fields and filtering conditions; Based on the semantic mapping relationship in the reshaped dynamic logical view graph, the query fields and filtering conditions are split and mapped into sub-query fields and sub-filtering conditions corresponding to the physical data source; The query request is rewritten into several corresponding sub-query requests based on the sub-query fields and sub-filter conditions.

7. The data fusion method based on dynamic map reshaping according to claim 1, characterized in that, After constructing a corresponding dynamic logical view graph based on the data entities and attributes in each data source and performing semantic association graph fusion and reshaping to obtain the reshaped dynamic logical view graph, the method further includes: Real-time monitoring of new data source access events; when a new data source is accessed, extract the corresponding newly added entities and attributes. Based on the newly added entities and attributes, the current reshaped dynamic logical view graph is further reshaped by semantic association through graph fusion, and the global logical view is updated.

8. A data fusion device based on dynamic map reshaping, characterized in that, include: The semantic awareness module is used to connect to several data sources to be processed, perform semantic awareness and extraction on each data source, and obtain the data entities and attributes in each data source. The graph construction and reshaping module is used to construct a corresponding dynamic logical view graph based on the data entities and attributes in each data source and to perform semantic association graph fusion and reshaping to obtain the reshaped dynamic logical view graph. A global logical view generation module is used to generate a corresponding global logical view based on the reshaped dynamic logical view graph. The query rewriting module is used to receive query requests submitted by users based on the global logical view, and to split and rewrite the query requests according to the reshaped dynamic logical view graph to obtain several sub-query requests. The query execution module is used to execute all sub-query requests, merge the query results of the sub-query requests, and return the merged query results to the user.

9. A computer device, characterized in that, Includes at least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the data fusion method based on dynamic map reshaping as described in any one of claims 1-7.

10. A non-volatile computer-readable storage medium, characterized in that, The non-volatile computer-readable storage medium stores computer-executable instructions that, when executed by one or more processors, cause the one or more processors to perform the data fusion method based on dynamic map reshaping as described in any one of claims 1-7.

Images