Data processing method for enterprise digital intelligence quality management

Abstract

The invention relates to the technical field of data processing, and discloses a data processing method for enterprise digital intelligence quality management. The method comprises the following steps: processing a document to obtain a first structured graph; matching the first structured atlas to obtain neighborhood topological information; aggregating the neighborhood topological information to obtain a first feature vector; acquiring second feature vectors corresponding to the vertexes based on identifiers corresponding to the vertexes in the first structured atlas; based on the first feature vector and the second feature vector, obtaining a weight of an edge between vertexes in the first structured atlas; and obtaining associated information of the elements corresponding to the vertexes of the first structured graph based on the updated weights. The method is used for solving the problems of low efficiency and inaccuracy of processing related data of enterprise quality management by a traditional method. The method is mainly used for enterprise digital intelligent quality management. According to the embodiment of the invention, document local information is combined with industry global knowledge, so that the depth and accuracy of data analysis are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a data processing method for enterprise digital quality management. Background Technology

[0002] For enterprise digital quality management, advanced information technology is used to digitize, automate, and intelligently manage quality management concepts, processes, and activities. The managed objects are not only physical products but also massive amounts of data. With the development of artificial intelligence technology, natural language models can be used to process document content and extract relevant features. Traditional methods are less efficient at processing enterprise quality management data and may also lead to inaccuracies. Summary of the Invention

[0003] At least one aspect and advantage of the invention will be set forth in part in the description which follows, or may be apparent from the description, or may be obtained by practicing the subject matter of this disclosure.

[0004] According to a first aspect of the present invention, a data processing method for enterprise digital quality management includes:

[0005] The document is processed using a multimodal first processor to obtain a first structured map;

[0006] Based on matching the first structured graph within the graph database, the neighborhood topology information corresponding to the vertices in the first structured graph is obtained;

[0007] The neighborhood topology information is aggregated to obtain the first feature vector corresponding to each vertex in the first structured graph;

[0008] The second feature vector corresponding to each vertex is obtained based on the identifier of the vertex in the first structured graph;

[0009] Based on the first and second feature vectors, the weights of the edges between vertices in the first structured graph are obtained;

[0010] Based on the updated weights, obtain the association information of the elements corresponding to the vertices of the first structured graph.

[0011] According to one embodiment of the present invention, the first feature vector is obtained in the following manner:

[0012] The weights of vertices in the neighborhood topology information are determined based on the distance between vertices in the neighborhood topology information and vertices in the first structured graph.

[0013] The first feature vector is obtained based on the vertices and their corresponding weights in the neighborhood topology information.

[0014] According to one embodiment of the present invention, the vertices and their corresponding weights in the neighborhood topology information increase based on the increasing distance between the vertices.

[0015] According to one embodiment of the present invention, the neighborhood topology information corresponding to a vertex in the first structured graph is obtained in the following manner:

[0016] Based on nearest neighbor search selection and a first set of vertices connected by edges in the first structured graph, the number of nodes between vertices in the first set and vertices in the first structured graph is no higher than a first threshold.

[0017] According to an embodiment of the present invention, the process of obtaining the second feature vector includes:

[0018] The first embedding vector is obtained based on the text corresponding to the vertices in the first structured graph;

[0019] The second, third, and fourth vectors are obtained based on the first embedding vector;

[0020] The second feature vector is obtained by concatenating the second, third, and fourth vectors.

[0021] The second, third, and fourth vectors are used to represent the degree of correlation between the first embedded vector and the corresponding first, second, and third preset texts, respectively.

[0022] The first preset text, the second preset text, and the third preset text all contain several preset words.

[0023] According to one embodiment of the present invention, the second vector, the third vector, and the fourth vector are obtained in the following manner:

[0024] Obtain the feature vectors corresponding to several preset words contained in the first preset text, the second preset text, or the third preset text;

[0025] Calculate the correlation between the first embedding vector and the feature vector corresponding to the preset word, respectively;

[0026] The corresponding vector is obtained based on the correlation between the first embedding vector and the feature vector corresponding to the preset word.

[0027] According to one embodiment of the present invention, the weights of edges between vertices in the first structured graph are obtained as follows:

[0028] The first approximation between vertices is obtained based on the identifiers corresponding to the vertices in the first structured graph;

[0029] The second approximation between vertices is obtained based on the first feature vector corresponding to the vertices in the first structured graph;

[0030] The third approximation between vertices is obtained based on the second eigenvector corresponding to the vertices in the first structured graph;

[0031] The weights of edges between vertices are obtained based on the first, second, and third approximations.

[0032] According to an embodiment of the present invention, the process of obtaining the association information of the elements corresponding to the vertices of the first structured graph includes:

[0033] Construct a second directed graph;

[0034] In response to the weight between vertices of the first structured graph being greater than the second threshold, it is determined that there is a correlation between vertices of the first structured graph, and the corresponding vertices are added to the second directed graph;

[0035] The difference edges and associated vertices are determined based on the differences between the first structured graph and the second directed graph.

[0036] According to one embodiment of the present invention, the second directed graph is obtained based on the fully connected graph of the first structured graph.

[0037] According to one embodiment of the present invention, the second directed graph is obtained based on the relationships of the first structured graph.

[0038] The beneficial effects of this invention include: This invention integrates quality knowledge scattered across countless documents into a unified visual network, enabling data-driven, refined management and improving efficiency. Simultaneously, by combining local document information with global industry knowledge, the information extracted from a single document is no longer isolated but imbued with a deep industry context, thereby enhancing the depth and accuracy of data analysis. Attached Figure Description

[0039] Figure 1 This is a flowchart of a data processing method for enterprise digital quality management in the embodiments of this application. Detailed Implementation

[0040] The present disclosure will now be discussed with reference to several exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and thus implement the present disclosure, and are not intended to imply any limitation on the scope of the disclosure.

[0041] This embodiment introduces a data processing method for enterprise digital quality management, including steps 1100-1600.

[0042] Step 1100: Process the document using the first multimodal processor to obtain the first structured map.

[0043] Multimodal processing refers to a primary processor capable of handling different types of documents, such as text, tables, and images. It combines technologies such as optical character recognition and natural language processing.

[0044] The first processor uses natural language processing (NLP) techniques to process documents, including named entity recognition and relation extraction. For enterprises, initial documents can be of various types, such as enterprise standards, processes, reports, and records. These documents contain a wide range of content related to enterprise quality management. Quality management is a systematic framework, methodology, and activities designed to ensure that the products, services, and processes offered by an organization consistently meet or exceed customer requirements and expectations, and to pursue efficiency and excellence.

[0045] For enterprise-wide digital quality management, advanced information technologies (such as big data, artificial intelligence, the Internet of Things, and knowledge graphs) are used to digitize, automate, and intelligently transform the aforementioned quality management concepts, processes, and activities. The managed objects are not only physical products but also massive amounts of data (production parameters, test reports, customer feedback, process documents, etc.). The goal is to achieve more accurate prediction, early warning, optimization, and decision-making, upgrading quality management from primarily "remedial" and "in-process control" to primarily "prevention" and "intelligent-driven" approaches.

[0046] For example, in the manufacturing industry, the initial input document can be a product quality report document, which includes product information, quality inspection and testing data, details of non-conformities and defects, product production process and resource association information, and handling and corrective measures.

[0047] Product information may include data such as product name, product model, production batch number, and production date.

[0048] Quality inspection and testing data may include a list of test items (dimensional tolerances, surface finish, hardness, color consistency, etc.), inspection standards (standard values ​​corresponding to each test item), measured data (results of actual measurements by inspectors or automated equipment), and judgment results (qualified, unqualified, pending judgment).

[0049] The details of nonconformities and defects include defect description, defect category, defect location, number of defects, and severity. The defect description is a textual description of the nonconformity, such as "scratches on the casing," "poor solder joints," or "misaligned label." The defect category and code indicate the specific defect type, such as appearance defect or functional defect. The defect location indicates the specific part of the product where the defect occurs, such as at a particular interface.

[0050] Product manufacturing process and resource-related information includes the production technology, procedures, equipment, tooling, and materials involved in product manufacturing.

[0051] The handling and corrective actions describe the actions taken in response to the identified problem, including initial handling recommendations, root cause analysis, corrective and preventive measures, etc.

[0052] The first structured graph is a preliminary knowledge graph, where vertices represent entities extracted from documents, and edges represent the relationships between these entities directly extracted from the documents. For example, for the quality report document of the aforementioned product, the vertices in the graph include product entities, inspection standard entities, and defect type entities. For instance, the content "Product A" in the report is a product entity vertex in the graph, and the content "shell scratches" in the report is a defect type entity vertex. The content "Product A has shell scratch defects" in the report is the edge between the two vertices.

[0053] Step 1200: Based on the graph database, match the first structured graph to obtain the neighborhood topology information corresponding to the vertices in the first structured graph.

[0054] A graph database is a database specifically designed for storing and querying graph-structured data, and it already contains knowledge graphs related to relevant industries.

[0055] The entities in the first structured graph obtained earlier are searched and matched in the graph database. Upon successful matching, all neighboring nodes and relationships surrounding that entity node can be found in the graph database.

[0056] For each entity in the first structured graph, the neighborhood topology information includes information about other entities adjacent to that entity in the graph database. Since the original document may not explicitly mention all the other related nodes for each node, potential neighbors can be identified by matching against the graph database.

[0057] Step 1300: Aggregate the neighborhood topology information to obtain the first feature vector corresponding to each vertex in the first structured graph.

[0058] A single entity node may have dozens or even hundreds of neighboring nodes, resulting in complex information. By aggregating the information of its neighbors, a feature vector is generated for each vertex. This feature vector encodes the vertex's structural information and local environment within the graph, reducing the complexity of neighboring nodes. For example, a device node that frequently connects to faulty nodes may have a feature vector containing high-risk semantics.

[0059] Step 1400: Obtain the second feature vector corresponding to each vertex based on the identifier of the vertex in the first structured graph.

[0060] Secondary feature extraction is performed on the vertex to analyze its contextual meaning within the document. The natural language model then identifies the text surrounding the identifier corresponding to that vertex in the original document, such as sentences or paragraphs containing that word. This text constitutes its context.

[0061] Step 1500: Based on the first feature vector and the second feature vector, obtain the weights of the edges between vertices in the first structured graph.

[0062] For each edge in the graph, there are two dimensions of information: a first feature vector representing the structural features from industry knowledge, and a second feature vector representing the semantic features from the current document. The system combines these two feature vectors to recalculate the strength or confidence of the relationship. For example, if the document initially extracts a connection between device A and device B, and the knowledge base reveals that they are standard configurations, and the context indicates that they are discussing fault correlations, then the system will increase the weight of this edge, considering it a close relationship. If the contextual semantics contradict the industry knowledge base, the weight may be decreased.

[0063] Step 1600: Obtain the association information of the elements corresponding to the vertices of the first structured graph based on the updated weights.

[0064] Based on the new weights, a more accurate and comprehensive correlation graph is generated. Enterprise managers can quickly understand the company's vast management system through the graph without having to read massive amounts of documents. Hidden risks are discovered through dynamic weights. For example, the system calculates that "supplier A's raw materials" and "minor parameter deviations in production equipment B" are not significant problems when occurring individually, but their simultaneous occurrence creates a very high correlation weight with a specific defect. This complex multi-factor correlation can be discovered through dynamic weights.

[0065] By analyzing the structure and weights of the graph, we can provide quantitative basis for process optimization and strategic decision-making. We can analyze which vertices are the hubs in the graph. If a vertex has too many high-weight connections, it may be a bottleneck. If a node is theoretically important but has very low actual connection weight, it may mean that there is a disconnect in the process execution.

[0066] This invention integrates quality knowledge scattered across numerous documents into a unified, visualized network, enabling data-driven, refined management. Furthermore, by combining local document information with industry-wide knowledge, information extracted from a single document is no longer isolated but imbued with a deep industry context, thereby improving the depth and accuracy of data analysis.

[0067] In this embodiment, the process of obtaining the association information of the elements corresponding to the vertices of the first structured graph includes: establishing a second directed graph; in response to the weight between the vertices of the first structured graph being greater than a second threshold, determining that the vertices of the first structured graph are associated, and adding the corresponding vertices to the second directed graph; and determining the difference edges and associated vertices based on the differences between the first structured graph and the second directed graph.

[0068] The second directed graph is a newly created empty directed graph. Traversing all vertices of the first structured graph, if the weight between two vertices is greater than a second threshold, it indicates a strong correlation between them, and these two vertices are added to the second directed graph. The second threshold can be set by experts based on experience, or it can be automatically determined based on the model's performance on the validation set.

[0069] Some relationships are indeed mentioned in the document, but their importance is insufficient to meet the criteria for a strong association. These relationships are filtered into the difference edges. Analysts can review these difference edges. For example, the system may determine that a certain type of screw and transformer have a low association weight, but the document may be discussing how a loose screw of that type caused a transformer failure. This special case can be discovered through manual review.

[0070] If a large number of important relationships are found to be filtered to the difference edge, it indicates that the second threshold may be set too high, providing direct data feedback for adjusting system parameters.

[0071] The second directed graph is the final product delivered to the business application, and the difference analysis is the quality inspection report and debugging tool that supports the quality of this product.

[0072] In one implementation, the second directed graph is obtained based on the fully connected graph of the first structured graph.

[0073] For example, if the first structured graph contains N vertices and M edges, then its fully connected graph contains N identical vertices. Between these N vertices, there is an edge between any two different vertices, forming a full connection.

[0074] Full connectivity also means that every factor needs to be considered, rather than selecting representative factors for data processing. This approach can help uncover potential correlations.

[0075] In another implementation, the second directed graph is obtained based on the relationships in the first structured graph. This approach considers only the relationships disclosed in the documents, rather than all relationships, thus improving efficiency.

[0076] In this embodiment, the first feature vector is obtained in the following way: the weights of vertices in the neighborhood topology information are determined based on the distance between vertices in the neighborhood topology information and vertices in the first structured graph; the first feature vector is obtained based on the vertices in the neighborhood topology information and their corresponding weights.

[0077] For each vertex in the first structured graph, calculate the distance between that vertex and all vertices in its corresponding neighborhood topology. Here, distance refers to the shortest path hop count; a distance of 1 indicates a direct neighbor, and a distance of 2 indicates a neighbor's neighbor. For example, in the graph, if vertex A is directly connected to vertex B, then the distance between vertex A and vertex B is 1. However, if vertex C is directly connected to vertex B but not directly connected to vertex A, then the distance between vertex A and vertex C is 2.

[0078] Specifically, the vertices and their corresponding weights in the neighborhood topology information increase based on the distance between the vertices.

[0079] In this embodiment, the neighborhood topology information corresponding to the vertices in the first structured graph is obtained in the following way: based on the nearest neighbor search selection and the first set of vertices in the first structured graph that are connected by edges, the number of nodes between the vertices in the first set of vertices and the vertices in the first structured graph is not higher than a first threshold.

[0080] Each vertex in the first vertex set is connected to a vertex in the first structured graph by an edge, which can be a direct or indirect connection. For example, if vertex A is a vertex in the first structured graph, vertex A is directly connected to vertex B, and vertex A is indirectly connected to vertex C through vertex B, then for vertex A, its corresponding first vertex set includes vertices B and C.

[0081] The first threshold is used to filter vertices that are far apart. For example, if there are 2 vertices between vertex A and vertex D, and 4 vertices between vertex A and vertex E, and the first threshold is set to 3, then vertex D belongs to the first vertex set. Although vertex E is also indirectly connected to vertex A, the number of intermediate nodes exceeds the first threshold, so vertex E does not belong to the first vertex set.

[0082] In this embodiment, the process of obtaining the second feature vector includes: obtaining a first embedding vector based on the text corresponding to the vertices in the first structured graph; obtaining a second vector, a third vector, and a fourth vector based on the first embedding vector; connecting the second vector, the third vector, and the fourth vector to obtain a second feature vector; the second vector, the third vector, and the fourth vector are used to represent the correlation between the first embedding vector and the corresponding first preset text, the second preset text, and the third preset text, respectively; the first preset text, the second preset text, and the third preset text all contain several preset words.

[0083] The text corresponding to a vertex contains the text identifier of that vertex and its context in the document. By using a pre-trained language model, an initial vector, namely the first embedding vector, is generated for this text, which already contains the basic semantic information of the entity.

[0084] For each document, multiple dimensions can be set, and for each dimension, several expected related terms can be set. A set of preset texts contains multiple preset terms.

[0085] The correlation between the first embedding vector and the first, second, and third pre-set texts is calculated respectively. The correlation can be represented by the distance between words, which is calculated by a pre-trained language model.

[0086] Specifically, the second, third, and fourth vectors are obtained as follows: feature vectors corresponding to several preset words contained in the first, second, or third preset text are obtained; the correlation between the first embedding vector and the feature vectors corresponding to the preset words are calculated respectively; and the corresponding vectors are obtained based on the correlation between the first embedding vector and the feature vectors corresponding to the preset words.

[0087] A set of pre-defined text contains multiple pre-defined words. A pre-trained language model is used to obtain the feature vector corresponding to each pre-defined word. The correlation between the first embedding vector and the feature vector of each pre-defined word is calculated, resulting in a correlation list. These correlation lists are then aggregated, and the average value is taken to obtain a single vector.

[0088] In this embodiment, the weights of the edges between vertices in the first structured graph are obtained as follows: the first approximation between vertices is obtained based on the identifiers corresponding to the vertices in the first structured graph; the second approximation between vertices is obtained based on the first feature vectors corresponding to the vertices in the first structured graph; the third approximation between vertices is obtained based on the second feature vectors corresponding to the vertices in the first structured graph; and the weights of the edges between vertices are obtained based on the first approximation, the second approximation, and the third approximation.

[0089] When calculating the weights of edges between vertices, the approximation between vertices, the approximation of associated vertices, and their approximation within the context are considered.

[0090] The first approximation represents the similarity between vertices, which can be directly calculated based on the identifiers corresponding to the vertices. A word vector model is used to convert the names of two vertices into vectors, and then the similarity between these two vectors is calculated.

[0091] The second approximation is calculated based on the first eigenvector of the vertex, which encodes the vertex's local topology in the industry knowledge base. If two vertices have similar neighbors in the industry knowledge base, their second approximation is higher, indicating that the two vertices play similar roles or have similar functional attributes.

[0092] The third approximation is calculated based on the second feature vector of the vertex, which encodes the vertex's specific semantics within the current document context. If two vertices consistently appear in similar contexts within a document, such as both appearing in paragraphs about fault diagnosis, then their third approximation will be high even if the two vertices are not directly connected in the global knowledge base.

[0093] Based on the pre-configured weights corresponding to each approximation level, the first, second, and third approximations are weighted and summed to obtain the weights of the edges between vertices, specifically calculated according to the following formula:

[0094] ;

[0095] in, For the first approximation, For the second approximation, The third approximation degree. The weights corresponding to the first approximation are: The weights corresponding to the second approximation are... This represents the weight corresponding to the third approximation. The sum of the weights corresponding to the three approximations is 1.

[0096] This embodiment calculates the first approximation, the second approximation, and the third approximation respectively, and finally calculates the weight of the edge between vertices based on the three approximations, thus verifying the results from multiple perspectives and improving the reliability of the results.

[0097] While specific embodiments of the present invention have been described in detail by way of examples, those skilled in the art should understand that the above examples are for illustrative purposes only and are not intended to limit the scope of the invention. Those skilled in the art should understand that modifications can be made to the above embodiments without departing from the scope and spirit of the invention.

[0098] Those skilled in the art will recognize that the modules and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0099] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and equipment can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0100] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0101] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0102] In addition, the functional modules in the embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0103] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0104] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

[0105] It should be understood that the sequence numbers of the steps in the invention's content and embodiments do not absolutely imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. The foregoing description of embodiments of this disclosure has been provided for illustrative and descriptive purposes. The foregoing description is not exhaustive and is not intended to limit this disclosure to the exact form disclosed. Various modifications and variations may exist based on the foregoing teachings, or various modifications and variations may be derived from the practice of this disclosure. These embodiments were chosen and described to illustrate the principles of this disclosure and its practical application, so that those skilled in the art can utilize this disclosure in various implementations and modifications suitable for the specific purpose of the concept.

Claims

1. A data processing method for enterprise digital quality management, characterized in that, include: The document is processed using a multimodal first processor to obtain a first structured map; Based on matching the first structured graph within the graph database, the neighborhood topology information corresponding to the vertices in the first structured graph is obtained; The neighborhood topology information is aggregated to obtain the first feature vector corresponding to each vertex in the first structured graph; The second feature vector corresponding to each vertex is obtained based on the identifier of the vertex in the first structured graph; Based on the first and second feature vectors, the weights of the edges between vertices in the first structured graph are obtained; Based on the updated weights, obtain the association information of the elements corresponding to the vertices of the first structured graph.

2. The data processing method for enterprise digital quality management as described in claim 1, characterized in that, The first feature vector is obtained in the following way: The weights of vertices in the neighborhood topology information are determined based on the distance between vertices in the neighborhood topology information and vertices in the first structured graph. The first feature vector is obtained based on the vertices and their corresponding weights in the neighborhood topology information.

3. The data processing method for enterprise digital quality management as described in claim 2, characterized in that, The vertices and their corresponding weights in the neighborhood topology information increase as the distance between vertices increases.

4. The data processing method for enterprise digital quality management as described in claim 1, characterized in that, The neighborhood topology information corresponding to the vertices in the first structured graph is obtained in the following way: Based on nearest neighbor search selection and a first set of vertices connected by edges in the first structured graph, the number of nodes between vertices in the first set and vertices in the first structured graph is no higher than a first threshold.

5. The data processing method for enterprise digital quality management as described in claim 1, characterized in that, The process of obtaining the second feature vector includes: The first embedding vector is obtained based on the text corresponding to the vertices in the first structured graph; The second, third, and fourth vectors are obtained based on the first embedding vector; The second feature vector is obtained by concatenating the second, third, and fourth vectors. The second, third, and fourth vectors are used to represent the degree of correlation between the first embedded vector and the corresponding first, second, and third preset texts, respectively. The first preset text, the second preset text, and the third preset text all contain several preset words.

6. The data processing method for enterprise digital quality management as described in claim 5, characterized in that, The second, third, and fourth vectors are obtained as follows: Obtain the feature vectors corresponding to several preset words contained in the first preset text, the second preset text, or the third preset text; Calculate the correlation between the first embedding vector and the feature vector corresponding to the preset word, respectively; The corresponding vector is obtained based on the correlation between the first embedding vector and the feature vector corresponding to the preset word.

7. The data processing method for enterprise digital quality management as described in claim 1, characterized in that, The weights of edges between vertices in the first structured graph are obtained as follows: The first approximation between vertices is obtained based on the identifiers corresponding to the vertices in the first structured graph; The second approximation between vertices is obtained based on the first feature vector corresponding to the vertices in the first structured graph; The third approximation between vertices is obtained based on the second eigenvector corresponding to the vertices in the first structured graph; The weights of edges between vertices are obtained based on the first, second, and third approximations.

8. The data processing method for enterprise digital quality management as described in claim 1, characterized in that, The process of obtaining the association information of the elements corresponding to the vertices of the first structured graph includes: Construct a second directed graph; In response to the weight between vertices of the first structured graph being greater than the second threshold, it is determined that there is a correlation between vertices of the first structured graph, and the corresponding vertices are added to the second directed graph; The difference edges and associated vertices are determined based on the differences between the first structured graph and the second directed graph.

9. The data processing method for enterprise digital quality management as described in claim 8, characterized in that, The second directed graph is obtained from the fully connected graph of the first structured graph.

10. The data processing method for enterprise digital quality management as described in claim 8, characterized in that, The second directed graph is obtained based on the relationships of the first structured graph.

Images