A method, system, device and medium for enterprise architecture artifact conflict detection based on multi-modal knowledge graph

Abstract

The application discloses a kind of based on multimodal knowledge graph Enterprise Architecture artifact conflict detection method, system, equipment and medium, method includes to multimodal data is carried out terminology unification processing, obtains standard data;Entity information is extracted from standard data, and the architecture relationship between each entity information is extracted, while adding multimodal attribute for each entity information, according to entity information, architecture relationship and multimodal attribute constructs multimodal knowledge graph;Definition conflict rule, according to conflict rule carries out conflict detection to multimodal knowledge graph, obtains artifact conflict;According to the priority of artifact conflict is evaluated according to conflict root and preset evaluation index;Combining the sorting result of priority and preset reference knowledge base to each artifact conflict generates corresponding solution.The application solves the problems of low efficiency of conflict detection, insufficient multimodal data processing capability and low accuracy of implicit conflict identification in the prior art.

Description

Technical Field

[0001] This invention relates to the field of enterprise architecture management technology, and in particular to a method, system, device and medium for detecting conflicts in enterprise architecture artifacts based on multimodal knowledge graphs. Background Technology

[0002] With the deepening of digital transformation, enterprise architecture management has become an important means to improve organizational operational efficiency and business collaboration capabilities. Architecture artifacts, as normative documents covering multiple levels such as enterprise strategy, business, applications, and data, encompass various types including management systems, business process diagrams, organizational job descriptions, and technical architecture design documents. These artifacts collectively constitute the knowledge foundation of enterprise operations. However, because enterprise architecture artifacts involve multiple business domains, various document formats, and multiple departments involved in their compilation and maintenance, consistency between different artifacts is difficult to guarantee, and conflicts frequently occur during the actual compilation and maintenance process.

[0003] Currently, conflict detection of enterprise architectural artifacts mainly relies on manual comparison and analysis. This method is not only time-consuming and labor-intensive, but the detection quality also highly depends on the experience level of experts. More challenging is that enterprise architectural artifacts are usually in multiple formats such as Word documents, Excel spreadsheets, Visio flowcharts, and PowerPoint presentations. Most existing detection tools can only process text data in a single format and cannot effectively integrate and analyze the relationships between these multimodal data. In addition, traditional conflict detection methods are mainly based on keyword matching and simple rule verification, lacking the ability to deeply understand semantics. They are unable to accurately identify implicit conflicts such as inconsistent terminology and cross-domain adaptation contradictions. Even if conflicts are detected, they cannot provide actionable solutions, resulting in limited overall efficiency improvements. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method, system, device, and medium for enterprise architecture artifact conflict detection based on multimodal knowledge graphs to solve the problems of low conflict detection efficiency, insufficient multimodal data processing capabilities, and low accuracy of implicit conflict identification in existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for detecting conflicts in enterprise architectural artifacts based on a multimodal knowledge graph, comprising: collecting and extracting multimodal data from enterprise architectural artifacts, and performing terminology standardization processing on the multimodal data to obtain standard data; extracting entity information from the standard data, extracting the architectural relationships between the entity information, and adding multimodal attributes to each entity information; constructing a multimodal knowledge graph based on the entity information, architectural relationships, and multimodal attributes; defining conflict rules, performing conflict detection on the multimodal knowledge graph according to the conflict rules to obtain artifact conflicts, tracing the source of the artifact conflicts, and locating the root cause of the conflicts; prioritizing the artifact conflicts according to the root cause of the conflicts, and generating corresponding solutions for each artifact conflict according to the priority ranking result.

[0007] As a preferred embodiment of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph described in this invention, the step of obtaining standard data includes: selecting the corresponding parsing method according to the format of each enterprise architecture artifact to extract information, obtaining multimodal data, and cleaning the multimodal data; performing terminology unification processing on the cleaned multimodal data according to a preset terminology standard, mapping synonyms with different expressions in the multimodal data to standard terms, and obtaining standard data.

[0008] The beneficial effects of this preferred technical solution are as follows: By selecting the corresponding parsing method based on the format of the enterprise architecture artifacts for information extraction, it can handle artifacts in various formats such as Word documents, Excel spreadsheets, Visio flowcharts, and PPT presentations, avoiding the data loss problem caused by traditional methods that can only handle a single text format; the cleaning process for multimodal data removes invalid data such as duplicate artifacts, blank documents, and files with incorrect formats, ensuring data quality; finally, the terminology is standardized according to the preset terminology standard, mapping synonymous terms with different expressions to standard terms. For example, "marketing control," "marketing management," and "marketing supervision" are uniformly mapped to "marketing control," eliminating the interference of terminological ambiguity on conflict detection. The resulting standard data has the characteristics of standardized terminology and semantic consistency, enabling enterprise architecture artifacts from different sources and in different formats to be compared and analyzed under a unified terminology system.

[0009] As a preferred embodiment of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph described in this invention, the step of constructing a multimodal knowledge graph based on the entity information, architectural relationships, and multimodal attributes includes: extracting entity information from the standard data using a combination of rule matching and entity recognition models; identifying explicit relationships between the entity information and using prompt words to guide a language model to mine implicit relationships between the entities, combining explicit and implicit relationships to obtain architectural relationships; adding multimodal attributes to the entity information, aligning the multimodal attributes using an attention mechanism, and aligning the entity information based on the alignment results of the multimodal attributes; and constructing a multimodal knowledge graph based on the aligned entity information, architectural relationships, and multimodal attributes.

[0010] The beneficial effects of this preferred technical solution are as follows: the combination of rule matching and entity recognition models balances extraction speed and accuracy, avoiding the problems of low efficiency when processing explicit terms or insufficient accuracy when processing complex terms by a single method; by guiding the language model to mine implicit relationships through prompt words, it can identify semantic associations in entity information that are not explicitly recorded but objectively exist, making up for the limitation of traditional methods that can only identify explicit relationships; the use of an attention mechanism to align multimodal attributes establishes an association mapping between text descriptions, graph nodes, and structured fields, solving the problem of isolated existence of different modal data and inability to perform association analysis; the alignment of entity information based on the alignment results of multimodal attributes eliminates the difference in the representation of the same entity in different modalities. For example, it identifies a node in a Visio flowchart and a process clause in a Word document as the same entity, avoiding redundant modeling and conflict misjudgment, so that the constructed multimodal knowledge graph can uniformly represent enterprise architectural artifacts of different formats.

[0011] As a preferred embodiment of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph described in this invention, the step of obtaining artifact conflict includes: matching the multimodal knowledge graph through the conflict rules to detect explicit conflicts; calculating the attribute semantic similarity between the entity information to detect implicit conflicts; performing reasoning analysis on the entity information through the architectural relationships in the multimodal knowledge graph to obtain cross-domain conflicts; and combining the explicit conflicts, implicit conflicts, and cross-domain conflicts to obtain artifact conflicts.

[0012] The beneficial effects of this preferred technical solution are as follows: matching multimodal knowledge graphs using conflict rules can quickly identify explicit conflicts such as term repetition and direct contradictions, while calculating the semantic similarity of attributes between entity information can discover implicit conflicts such as inconsistent terminology. The combination of the two covers multiple conflict types within and between artifacts. Furthermore, reasoning and analyzing entity information through architectural relationships in the multimodal knowledge graph can identify adaptation contradictions and dependency conflicts between different domains such as business architecture, application architecture, and data architecture. Combining explicit conflicts, implicit conflicts, and cross-domain conflicts to obtain artifact conflicts avoids missed detections caused by relying solely on rule matching or misjudgments caused by relying solely on semantic calculation. This allows conflict detection to cover both explicit rule violations and potential semantic contradictions and cross-domain adaptation problems.

[0013] As a preferred embodiment of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph described in this invention, the step of prioritizing artifact conflicts according to the conflict root cause includes: analyzing the multimodal knowledge graph according to the conflict root cause to obtain the number of entity information and the number of enterprise architecture artifacts involved in the artifact conflict, and determining the scope of impact of the artifact conflict based on the number of entity information and the number of enterprise architecture artifacts; determining whether the artifact conflict violates preset compliance requirements and determining the severity of the artifact conflict; determining whether the artifact conflict affects the normal operation of business and determining the urgency of the artifact conflict; calculating the priority scores of the scope of impact, severity, and urgency using the analytic hierarchy process (AHP), and ranking the artifact conflicts according to the priority scores.

[0014] The beneficial effects of this preferred technical solution are as follows: Analyzing the multimodal knowledge graph based on the root causes of conflicts allows for tracing the number of entity information involved in artifact conflicts and the number of enterprise architecture artifacts, thereby quantifying the scope of impact of artifact conflicts and avoiding subjectivity caused by relying solely on experience. A multi-dimensional evaluation system is established by determining the severity of artifact conflicts based on whether they violate preset compliance requirements and the urgency based on whether they affect normal business operations. Finally, the Analytic Hierarchy Process (AHP) is used to calculate priority scores for the scope of impact, severity, and urgency, comprehensively considering the weight relationships of multiple evaluation indicators. Artifact conflicts are ranked according to their priority scores, solving the problem of difficulty in determining the processing order when a large number of conflicts exist simultaneously, enabling enterprise architecture managers to prioritize high-priority conflicts.

[0015] As a preferred embodiment of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph described in this invention, the step of generating corresponding solutions for each artifact conflict according to the priority ranking results includes: retrieving relevant standard specifications from a preset reference knowledge base based on the artifact conflict; determining the adjustment direction of each artifact conflict according to the priority ranking results in conjunction with the standard specifications, and determining a solution based on the adjustment direction.

[0016] The beneficial effects of this preferred technical solution are as follows: It retrieves relevant enterprise architecture standards and industry best practices from a pre-defined reference knowledge base based on artifact conflicts, avoiding the problems of unstable solution quality and difficulties in knowledge transfer caused by relying on expert experience; it determines the adjustment direction for each artifact conflict by combining enterprise architecture standards and industry best practices, ensuring that the solution has a normative basis and practical support; it determines the adjustment direction according to the priority ranking results, ensuring that high-priority conflicts receive solutions first, avoiding resource dispersion and low processing efficiency. The solution determined based on the adjustment direction includes elements such as the modification object, modification content, applicable standards, and implementation steps, clarifying which artifacts need to be modified, how to modify them, what standards to follow, and specific operational steps, reducing the workload of secondary analysis for enterprise architecture managers and enabling the solution to be directly implemented.

[0017] As a preferred embodiment of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph described in this invention, the method further includes: displaying the artifact conflicts and solutions through visual interaction, and collecting user feedback; wherein the step of visual interaction display includes: displaying the nodes and relationships of each artifact conflict through a visual interface, and using different colors to mark the corresponding priorities; collecting user feedback on the accuracy of the artifact conflict detection results and the rationality of the solutions, and using the feedback as the basis for conflict rule optimization.

[0018] The beneficial effects of this preferred technical solution are as follows: A visual interface displays the nodes and relationships of conflicts among various products, intuitively presenting the distribution and impact paths of product conflicts in a multimodal knowledge graph. Different colors are used to indicate corresponding priorities, enabling enterprise architecture managers to quickly identify conflicts that require priority handling. User feedback on the accuracy of product conflict detection results and the rationality of solutions is collected, establishing a feedback channel between the detection system and actual applications, enabling the identification of misjudgments and omissions during the detection process. Feedback evaluations serve as the basis for conflict rule optimization, allowing conflict rules to be adjusted and improved based on problems encountered in actual applications, such as adjusting rule thresholds, adding new conflict rules, and deleting rules with high false alarm rates, thereby improving the accuracy of subsequent conflict detection.

[0019] Secondly, the present invention provides an enterprise architecture artifact conflict detection system based on a multimodal knowledge graph, comprising: Data preprocessing module: used to collect and extract multimodal data from enterprise architecture artifacts, clean and standardize the terminology of the multimodal data to obtain standard data; Knowledge graph construction module: used to extract entity information and architectural relationships from the standard data, add multimodal attributes to each entity information and align them, and construct a multimodal knowledge graph; Conflict detection module: used to define conflict rules, perform conflict detection on the multimodal knowledge graph according to the conflict rules, obtain product conflicts and locate the root cause of the conflict; Conflict assessment and solution module: used to prioritize and rank the conflicts of the products, and generate corresponding solutions in conjunction with the reference knowledge base; Visualization and Feedback Module: This module is used to visualize the product conflicts and solutions, and to collect user feedback and evaluations.

[0020] Thirdly, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph.

[0021] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the enterprise architecture artifact conflict detection method based on a multimodal knowledge graph.

[0022] Compared with existing technologies, the beneficial effects of this invention are as follows: By constructing a multimodal knowledge graph, it achieves unified modeling and association mapping of enterprise architectural artifacts in different formats such as Word documents, Excel spreadsheets, Visio flowcharts, and PPT presentations, breaking the limitation of traditional detection tools that can only handle single text formats. Furthermore, it extracts entity information by combining rule matching and entity recognition models, and guides the language model to mine implicit relationships through prompts, enabling the knowledge graph to not only represent explicit architectural relationships but also capture potential semantic associations between different artifacts. In the conflict detection stage, it integrates rule matching and semantic similarity calculation, enabling simultaneous identification of explicit conflicts such as term repetition and direct contradictions, as well as implicit conflicts such as inconsistent terminology and logical incompatibility, improving detection accuracy compared to traditional keyword matching methods. More importantly, based on the architectural relationships in the multimodal knowledge graph, it can identify cross-domain conflicts between business architecture, application architecture, data architecture, and other domains, solving the problem that existing technologies struggle to cover cross-domain conflicts. In terms of conflict resolution, it can not only pinpoint the root cause of conflicts but also prioritize them based on their scope, severity, and urgency. It then generates specific solutions by combining enterprise architecture standards and industry best practices from a reference knowledge base. These solutions include adjustment directions and implementation suggestions, providing directly actionable guidance for enterprise architecture managers. Furthermore, through visualization and user feedback mechanisms, conflict rules can be continuously optimized to improve the accuracy of subsequent detection, forming a closed-loop detection system. This reduces the conflict detection cycle for a single architecture system from weeks to hours, lowering labor costs and improving the overall efficiency and quality of enterprise architecture management. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of the overall process of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph according to an embodiment of the present invention. Detailed Implementation

[0025] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0026] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for detecting conflicts in enterprise architectural artifacts based on multimodal knowledge graphs is provided, including steps S100 to S500: S100. Collect and extract multimodal data from enterprise architecture artifacts, and perform terminology standardization processing on the multimodal data to obtain standard data.

[0027] S200. Extract entity information from standard data and extract the architectural relationships between each entity. At the same time, add multimodal attributes to each entity and construct a multimodal knowledge graph based on the entity information, architectural relationships, and multimodal attributes.

[0028] S300. Define conflict rules, perform conflict detection on the multimodal knowledge graph according to the conflict rules, obtain product conflicts, trace the source of product conflicts, and locate the root cause of the conflict.

[0029] S400. Prioritize the product conflicts according to their root causes, and generate corresponding solutions for each product conflict based on the priority ranking results.

[0030] The S500 uses a visually interactive approach to showcase product conflicts and solutions, and collects user feedback.

[0031] It should be noted that during the actual compilation and maintenance of enterprise architecture artifacts (including management systems, business process diagrams, organizational job descriptions, technical architecture design documents, etc.), due to the involvement of multiple business domains, various document formats, and multiple compilation departments, issues such as inconsistent terminology, conflicting job definitions, and cross-domain compatibility contradictions can easily arise between artifacts. Traditional manual inspection methods are inefficient, with conflict detection for a single architecture system taking several weeks and heavily relying on expert experience, making it difficult to cover implicit and cross-domain conflicts. Existing automated inspection tools are mostly based on keyword matching and simple rule validation, unable to handle correlation conflicts in multimodal data (such as text, charts, and structured data), and their semantic consistency check accuracy is generally low, failing to meet the actual needs of enterprise architecture management.

[0032] Therefore, to address the aforementioned issues in enterprise architecture artifact conflict detection, steps S100-S500 are used to construct a multimodal knowledge graph that uniformly represents architecture artifacts in different formats, enabling the mapping of text, charts, and structured data. By integrating rule matching and language model reasoning, explicit and implicit conflicts are accurately identified, improving the accuracy of conflict detection. Through the cross-domain association capabilities of the knowledge graph, systematic detection of conflicts across multiple domains, including business architecture, application architecture, and data architecture, is achieved. Priority is evaluated and referenced from the knowledge base to generate feasible solutions. Simultaneously, through visualization and user feedback mechanisms, the accuracy of conflict detection is continuously optimized, reducing the conflict detection cycle for a single architecture system from several weeks to several hours.

[0033] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the above embodiment, a method for detecting conflicts in enterprise architecture artifacts based on multimodal knowledge graphs is provided.

[0034] In the embodiments of this application, step S100 involves collecting and extracting multimodal data from enterprise architecture artifacts, and performing terminology standardization processing on the multimodal data to obtain standard data. The steps for obtaining standard data include A1 to A2: A1. Select the corresponding parsing method according to the format of each enterprise's architecture artifacts to extract information, obtain multimodal data, and clean the multimodal data. In this embodiment, taking the enterprise architecture artifacts of Guangxi Power Grid Co., Ltd. as an example, the company uploaded more than 100 enterprise architecture artifacts, including the "Power Marketing Business Management System" (Word document), the "Power Grid Operation and Maintenance Business Process Flowchart" (Visio file), the "Organizational Job Responsibility Allocation Table" (Excel file), and the "Application Architecture Blueprint" (PPT file). First, the corresponding parsing method was selected according to the file format: for Word, PDF, and PPT files, the open-source tools Apache POI and PDFBox were used to extract the text content; for Visio flowcharts, OCR technology was used to recognize the text information and node structure in the charts, extracting business process node names, node order, and connection relationships; for Excel job responsibility allocation tables, a data parsing engine was used to extract structured data such as field names, field values, and table relationships. After the information extraction was completed, the multimodal data was cleaned to remove invalid data such as duplicate uploaded policy attachments, blank documents, and incorrectly formatted files. Redundant information such as headers, footers, and irrelevant attachments was filtered using regular expressions, ultimately resulting in a multimodal dataset containing text data, chart data, and structured data.

[0035] A2. Based on the preset terminology standards, perform terminology unification processing on the cleaned multimodal data, and map the synonyms of different expressions in the multimodal data to standard terms to obtain standard data. In this embodiment, the terminology of the cleaned multimodal data is standardized according to the terminology standards in the "Enterprise Structure Management Regulations of China Southern Power Grid Co., Ltd." For example, if the terms "marketing control" are detected in the "Power Marketing Business Management System," "marketing management" in the "Customer Service Specification," and "marketing supervision" in the "Business Process Flowchart," all referring to the same business function, they are uniformly mapped to the standard term "marketing control" according to the preset terminology standards. Similarly, "job responsibility," "job duties," and "job description" in Excel spreadsheets are uniformly mapped to the standard term "job responsibility," and "CRM system," "customer relationship management system," and "customer management system" in different documents are uniformly mapped to the standard term "customer relationship management system." Through terminology standardization, the differences in terminology expression in different enterprise structure artifacts are eliminated, resulting in standardized data with consistent terminology and semantics.

[0036] In the embodiments of this application, step S200 involves extracting entity information from standard data and extracting the architectural relationships between each entity information. At the same time, multimodal attributes are added to each entity information, and a multimodal knowledge graph is constructed based on the entity information, architectural relationships, and multimodal attributes. The steps for constructing a multimodal knowledge graph based on entity information, architectural relationships, and multimodal attributes include B1 to B4: B1. Extract entity information from standard data by combining rule matching and entity recognition models; In this embodiment, firstly, for explicit terms in the enterprise architecture artifacts, rule matching is used to extract them from the standard data obtained in step A2. For example, the regular expression "(electricity|grid)(marketing|operation and maintenance|dispatch)" is used to identify business domain entities such as "electricity marketing" and "grid operation and maintenance," and dictionary matching is used to identify application architecture entities such as "customer relationship management system" and "equipment ledger system." Secondly, for complex terms and ambiguous expressions, a Transformer-based entity recognition model is used for extraction. The input of the entity recognition model is a text sentence in the standard data, such as "the customer manager in the marketing department is responsible for customer relationship maintenance." The entity recognition model extracts semantic features through the BERT encoding layer and outputs the entity label sequence "marketing department / organizational entity customer manager / position entity" through the CRF decoding layer, thereby identifying the "marketing department-customer manager" organizational position entity. At the same time, the entity recognition model adopts a parameter-efficient fine-tuning technique and is optimized based on 500 power enterprise architecture label corpora. The fine-tuned entity recognition model can accurately identify entity types unique to the power field. Ultimately, by combining rule matching with entity recognition models, various types of entity information, such as business domain entities, architectural element entities, organizational position entities, and rule standard entities, are extracted from standard data.

[0037] B2. Identify the explicit relationships between the information of each entity, and use prompt words to guide the language model to mine the implicit relationships between the entities. Combine the explicit and implicit relationships to obtain the architectural relationships. In this embodiment, dependency parsing and semantic role labeling techniques are first used to identify explicit relationships between entity information. Dependency parsing is used to identify dependency relationships between words in a sentence. For example, from the sentence "The customer manager belongs to the marketing department" in the "Organizational Job Responsibility Allocation Table," dependency parsing identifies "belongs to" as the core verb, "customer manager" as the subject, and "marketing department" as the object, thus extracting the explicit relationship "customer manager - belongs to - marketing department." Next, a prompt word-guided language model is used to mine implicit relationships, constructing a prompt word template: "According to enterprise architecture knowledge, {entity A} and {entity B} in business flow..." The relationships in the process are: A. Support; B. Adapt; C. Dependency; D. No association. Then, the entity pairs "Customer Relationship Management System" and "Electricity Marketing Business Requirements" are substituted into the template and input into the language model to obtain the output "B. Adapt, because the customer data management function provided by the Customer Relationship Management System meets the customer information requirements of the Electricity Marketing Business". This reveals the implicit relationship "Customer Relationship Management System - Adapt - Electricity Marketing Business Requirements". Finally, the explicit and implicit relationships are combined to obtain 12 types of architectural relationships, including membership, support, adaptation, conflict, duplication, dependency, inclusion, supplement, substitution, compliance, association, and no association.

[0038] B3. Add multimodal attributes to each entity information, use an attention mechanism to align each multimodal attribute, and align the entity information according to the alignment result of the multimodal attributes; In this embodiment, multimodal attributes are first added to the entity information extracted in step B1. For example, for the "Account Manager" entity, text attributes (job description "responsible for customer relationship maintenance and business development"), chart attributes (the "Customer Registration" node in the business process diagram), and structured attributes (5 positions) are added. For the "Customer Relationship Management System" entity, text attributes (system function description), chart attributes (system hierarchy in the application architecture blueprint), and structured attributes (system version number, number of deployment servers) are added. After the attributes are added, an attention mechanism is used to construct a modal association matrix to align the multimodal attributes. This attention mechanism identifies semantically related attributes and establishes alignment relationships by calculating the association weights between different modal attributes. For example, the "Customer Registration" node in the Visio flowchart is associated with the "Customer Registration Process" clause in the Word document, and the "Customer ID" field in the Excel spreadsheet is associated with the "Customer Coding Rules" in the data standard document. Finally, based on the alignment results of the multimodal attributes, the entity information is further aligned. For example, the entity "Customer Relationship Management System" extracted from the Word document and the entity "Customer Relationship Management System" extracted from the Excel spreadsheet have the same terminology, but by comparing their multimodal attributes (both system version number is V3.2 and deployment server is 10.20.30.40), it is confirmed that they are the same entity and are merged to eliminate duplicate entities.

[0039] B4. Construct a multimodal knowledge graph based on the aligned entity information, architectural relationships, and multimodal attributes; In this embodiment, a multimodal knowledge graph is constructed based on the aligned entity information obtained in step B3, the architectural relationships obtained in step B2, and the multimodal attributes obtained in step B3. The multimodal knowledge graph uses entity information as nodes, architectural relationships as edges, and multimodal attributes as attribute labels for nodes and edges. For example, the knowledge graph contains a "Customer Relationship Management System" node (carrying text attributes, chart attributes, and structured attributes). This node is connected to the "Electricity Marketing" node through a "Support" architectural relationship, to the "Electricity Marketing Business Needs" node through an "Adapt" relationship, and to the "Q / CSG2041002-2023 Enterprise Architecture Management Regulations" node through a "Compliance" relationship. After construction, human feedback reinforcement learning technology is used to optimize the multimodal knowledge graph. This technology collects conflict cases annotated by experts as training data. For example, if an expert annotates "There is no support relationship between the Customer Relationship Management System and the Equipment Ledger System," the incorrect relationship is deleted or corrected based on expert feedback, reducing the false positive rate of relationship identification and improving the relationship accuracy of the knowledge graph. The final optimized multimodal knowledge graph covers entity information, architectural relationships, and multimodal attributes from more than 100 enterprise architecture artifacts of Guangxi Power Grid Company, forming a unified knowledge representation system.

[0040] In this embodiment of the application, step S300 involves defining conflict rules, performing conflict detection on the multimodal knowledge graph according to the conflict rules, obtaining product conflicts, tracing the product conflicts, and locating the root cause of the conflicts. In this embodiment, based on the multimodal knowledge graph constructed in step S200, conflict rules are defined and multi-level conflict detection is performed. Explicit conflicts, implicit conflicts, and cross-domain conflicts are identified through rule matching, semantic similarity calculation, and graph reasoning algorithms, respectively. The resulting product conflicts are then summarized, and the detected product conflicts are analyzed for their source. The root cause of the conflict is located by tracing back through the architectural relationship paths of the multimodal knowledge graph. For example, regarding the conflict in the processing time limit (24 hours vs. 48 hours) of the "Customer Complaint Handling Process" in the "Electricity Marketing Business Management System" and the "Customer Service Specification," the knowledge graph backtracking reveals that the conflict stems from different definitions of the same business process in the two products. Further tracing back to Article 3.2 of the "Electricity Marketing Business Management System" and Article 5.1 of the "Customer Service Specification" reveals that the conflict originates from different definitions of the same business process in the two products.

[0041] The steps for obtaining the product conflict include C1 to C4: C1. Match the multimodal knowledge graph using conflict rules to detect explicit conflicts; In this embodiment, conflict rules include rules such as terminology duplication, direct contradiction, and overlapping responsibilities. For example, the terminology duplication rule is defined as "the same architectural element is defined multiple times in different artifacts," the direct contradiction rule is defined as "the same business rule has obvious conflicts in different artifacts," and the overlapping responsibilities rule is defined as "the core responsibilities of different organizational positions are completely consistent." The multimodal knowledge graph is traversed to detect explicit conflicts. For example, if both the "Electricity Marketing Business Management System" and the "Customer Service Specification" define a "customer complaint handling process," but their processing time limits are "24 hours" and "48 hours" respectively, this is considered a direct contradiction conflict. If the core responsibilities of "Account Manager" and "Customer Service Specialist" are completely consistent, this is considered an overlapping responsibilities conflict. If the "Customer Relationship Management System" is repeatedly defined in the application architecture blueprint and IT asset list, this is considered a terminology duplication conflict. Finally, the detected explicit conflicts are recorded as conflict instances, including conflict type, involved entities, and conflict description information.

[0042] C2. Calculate the semantic similarity of attributes between the information of each entity to detect implicit conflicts; In this embodiment, a semantic consistency check algorithm is used to calculate the similarity of the text attributes of entities. This algorithm, based on lexical matching, synonym recognition, and semantic alignment techniques, measures the degree of semantic similarity between entities. For example, it detects that both the "Organizational Structure Diagram" and the "Business Process Diagram" use the term "Customer Service Center," which are not defined in the terminology standard of step A2. The semantic similarity calculation yields a similarity of 0.93, exceeding the preset threshold of 0.85, and both refer to the same architectural element, thus constituting a terminology inconsistency conflict. It also detects that the "Equipment Inspection Analysis" step in the "Power Grid Operation and Maintenance Business Process Diagram" requires "Equipment Data Analysis Function," but the "Equipment Ledger System" in the "Application Architecture Blueprint" lacks this function, constituting a logical incompatibility conflict. Finally, the detected implicit conflicts are recorded as conflict instances, labeled with similarity scores and the relevant entity information.

[0043] C3. By reasoning and analyzing the information of each entity through the architectural relationships in the multimodal knowledge graph, cross-domain conflicts are obtained; In this embodiment, a graph reasoning algorithm is used to perform multi-hop path traversal of the knowledge graph to identify potential conflicts across business domains and architectural levels. For example, by using multi-hop reasoning to detect cross-domain relationships in the knowledge graph, traversing along the cross-domain relationship path between "data architecture - data standards" and "business architecture - business requirements," it is detected that the "customer ID coding rules" in the "Data Standards Manual" stipulate that the customer ID is an 8-digit number, but the "customer coding requirements" in the "Electricity Marketing Business Management System" stipulate that it is a 10-digit number, which is determined to be a cross-domain adaptation conflict. Traversing along the cross-domain relationship path between "application architecture - data management" and "business architecture - business rules," it is detected that the "customer data retention period" is specified as 3 years in the data standards, but 5 years is required in the business requirements, which is determined to be a compliance conflict. Finally, the detected cross-domain conflicts are recorded as conflict instances, and the reasoning path and the involved architectural relationships are labeled.

[0044] C4. Combining explicit conflicts, implicit conflicts, and cross-domain conflicts yields product conflicts; In this embodiment, the explicit conflicts detected in step C1, the implicit conflicts detected in step C2, and the cross-domain conflicts detected in step C3 are summarized to obtain artifact conflicts. For example, the summarized artifact conflicts include direct contradictions in the "customer complaint handling process," inconsistencies in terminology between "customer service center" and "customer service manager," cross-domain adaptation conflicts in "customer ID coding rules," and compliance conflicts in "customer data retention period." Finally, each artifact conflict is recorded as a conflict instance, including information such as conflict type, involved entities, conflict description, similarity score, and reasoning path, forming a complete artifact conflict detection result.

[0045] In this embodiment of the application, step S400 involves prioritizing the product conflicts according to their root causes and generating corresponding solutions for each product conflict based on the priority ranking results. The steps for prioritizing product conflicts based on their root causes include D1 to D4: D1. Analyze the multimodal knowledge graph based on the root causes of the conflict to obtain the number of entity information and enterprise architecture artifacts involved in the artifact conflict, and determine the scope of the impact of the artifact conflict based on the number of entity information and enterprise architecture artifacts. First, the architectural relationship paths of the multimodal knowledge graph are traversed to count the number of entity information and enterprise architectural artifacts involved in the artifact conflict. For example, for the processing time limit conflict of the "customer complaint handling process", the knowledge graph traversal reveals that the conflict involves 8 entity information such as "customer complaint handling system", "customer service position", and "customer service assessment standard", and 4 enterprise architectural artifacts such as "electricity marketing business management system", "customer service specification", "application architecture blueprint" and "organizational job responsibility allocation table". In this embodiment, the impact scope level is determined according to the number of entity information and the number of enterprise architectural artifacts. The impact scope assessment rules are set as follows: ≥10 entity information or ≥5 enterprise architectural artifacts are judged as high impact scope; 5-9 entity information or 2-4 enterprise architectural artifacts are judged as medium impact scope; and <5 entity information and <2 enterprise architectural artifacts are judged as low impact scope. This artifact conflict involves 8 entity information and 4 enterprise architectural artifacts, and is judged as medium impact scope.

[0046] D2. Determine whether the product conflict violates the preset compliance requirements and determine the severity of the product conflict; In this embodiment, based on the compliance requirements in the "Enterprise Architecture Management Regulations of China Southern Power Grid Co., Ltd.", it is determined whether artifact conflicts violate preset compliance requirements. Compliance requirements include terminology standardization requirements, business process consistency requirements, data standard compliance requirements, and system architecture rationality requirements. For example, regarding the conflict in the processing time limit of the "Customer Complaint Handling Process," the inspection found that Article 4.3 of the "Enterprise Architecture Management Regulations of China Southern Power Grid Co., Ltd." clearly stipulates that "the definition of the same business process should be consistent in different artifacts." This artifact conflict violates the business process consistency requirement and is judged as high severity. Regarding the terminology inconsistency between "Customer Service Center" and "Customer Service Center," the inspection found that this conflict does not violate mandatory compliance requirements, but is only a terminology standardization suggestion, and is judged as low severity. The severity level is determined based on the compliance violation: violations of mandatory compliance requirements are judged as high severity, violations of suggested compliance requirements are judged as medium severity, and no violations of compliance requirements are judged as low severity.

[0047] D3. Determine whether the product conflict affects the normal operation of the business and determine the urgency of the product conflict; In this embodiment, the impact of artifact conflicts on business operations is assessed by analyzing the operational status of entities involved in the conflict, such as business processes, application systems, and organizational positions. For example, regarding the conflict over the processing time limit for the "customer complaint handling process," the analysis revealed that this artifact conflict led to inconsistent standards between the customer service and marketing departments in actual execution, resulting in three escalation incidents of customer complaints and affecting normal business operations, thus classifying it as high urgency. Regarding the cross-domain adaptation conflict of the "customer ID coding rules," the analysis found that this conflict has not yet been exposed in actual business operations, and the system is still operating normally, classifying it as low urgency. Finally, the urgency level is determined based on the impact on business operations: those affecting normal business operations are classified as high urgency, those potentially affecting business operations are classified as medium urgency, and those not affecting business operations are classified as low urgency.

[0048] D4. Use the analytic hierarchy process (AHP) to calculate priority scores for the scope of impact, severity, and urgency, and rank the product conflicts according to the priority scores. In this embodiment, a judgment matrix is ​​constructed, and the relative importance of indicators is determined through expert scoring. The weight vector is calculated as follows: urgency weight 0.5, severity weight 0.3, and impact range weight 0.2. Then, the three indicators for each product conflict are quantitatively scored: high level is assigned 3 points, medium level 2 points, and low level 1 point. For example, the impact range of "customer complaint handling process" is medium (2 points), severity is high (3 points), and urgency is high (3 points), so the priority score is calculated as 2×0.2+3×0.3+3×0.5=2.8 points; the impact range of "customer ID coding rules" is low (1 point), severity is medium (2 points), and urgency is low (1 point), so the priority score is calculated as 1×0.2+2×0.3+1×0.5=1.3 points. Finally, the product conflicts are sorted in descending order according to the priority scores, with higher-scoring conflicts having higher priority.

[0049] In an optional implementation, step S400, which assesses the priority of product conflicts based on the root cause of the conflict and preset evaluation indicators, can also be analyzed and evaluated based on the conflict propagation path. Specifically, through the architectural relationship path of a multimodal knowledge graph, a diffusion traversal is performed starting from the root cause node to calculate the length of the conflict propagation path and the number of stakeholders covered. For example, for a time-limit conflict in the "customer complaint handling process," starting from the root cause, Clause 3.2 of the "Electricity Marketing Business Management System," a traversal is performed along the "definition-support-dependency" relationship path. It is found that the propagation path length of this product conflict is 5 hops, covering 3 stakeholders: the marketing department, the customer service department, and the information technology department. Then, the priority evaluation score is determined based on the propagation path length and the number of stakeholders; the longer the propagation path and the more stakeholders, the higher the priority score.

[0050] The steps for generating corresponding solutions to conflicts of each product according to the priority ranking result include E1~E2: E1. Retrieve relevant standards and specifications from the preset reference knowledge base based on product conflicts; In this embodiment, the pre-set reference knowledge base includes enterprise architecture standards such as the "Enterprise Architecture Management Regulations of China Southern Power Grid Co., Ltd.", the "TOGAF Enterprise Architecture Framework", and the "Power Industry Information Standards", as well as enterprise architecture artifact cases and conflict resolution experiences from other provincial power grid companies within the power industry. Semantic retrieval technology is used, with the description of artifact conflicts as query input, to retrieve relevant content from the reference knowledge base. For example, for a conflict regarding the processing time limit of a "customer complaint handling process", the query "customer complaint handling process time limit standard" is entered, retrieving the "Southern Power Grid Customer Service Management Measures" which stipulates that "customer complaints should be processed within 24 hours", and the best practice case from Guangdong Power Grid Company, "customer complaint handling processes are unified as a 24-hour response mechanism". Finally, the retrieved enterprise architecture standards and industry best practices are used as references for solution generation.

[0051] E2. Based on the standard specifications and the priority ranking results, determine the adjustment direction for each product conflict, and determine the solution based on the adjustment direction; In this embodiment, different resolution strategies are adopted for different artifact conflicts: for terminology duplication conflicts, it is recommended to merge duplicate entities and unify the descriptions in the architectural artifacts; for terminology inconsistency conflicts, it is recommended to replace with standard terms and update the relevant architectural artifacts; for cross-domain adaptation conflicts, it is recommended to optimize the technical architecture or adjust business requirements, and specific adjustment directions are given. For example, for the time limit conflict of "customer complaint handling process" with a priority score of 2.8, the generated solution includes modifying Clause 5.1 of the "Customer Service Specification", modifying the content to "unify the processing time limit from 48 hours to 24 hours", based on the "Southern Power Grid Customer Service Management Measures", and the implementation steps are "revise the relevant clauses of the "Customer Service Specification", synchronously update the functional requirements of the "Customer Complaint Handling System", and adjust the assessment standards of the "Customer Service Positions". For the functional deficiency conflict of the "Equipment Ledger System", the generated solution is "it is recommended to expand the data analysis module of the equipment ledger system, with core requirements including equipment operation data statistics, fault trend analysis, equipment health assessment, etc., and technical indicators requiring support for real-time analysis of millions of equipment data".

[0052] In an optional implementation, step S400, which generates corresponding solutions for each product conflict according to the priority ranking results, can also be based on matching with historical cases. Specifically, the currently detected product conflict is matched with historical conflict cases stored in the reference knowledge base for similarity, and similarity scores are calculated for features such as conflict type, involved entities, and scope of impact. The historical case with the highest similarity is then retrieved. For example, for a time-limit conflict in the "customer complaint handling process," a similar time-limit conflict in the "customer repair response process" handled by Guangdong Power Grid Company is found, with a similarity score of 0.89. The solution from this historical case is extracted as a reference template and adjusted according to the specific circumstances of the current conflict. According to the priority ranking results, historical cases are matched and solutions are generated for high-priority conflicts first. For conflicts without similar historical cases in the reference knowledge base, a rule generation method based on enterprise architecture standards is adopted.

[0053] In another optional implementation, the generation of corresponding solutions for each product conflict in step S400 according to the priority ranking result can also be batch generated based on conflict association analysis. First, the correlation between product conflicts is analyzed through the architectural relationship path analysis of the multimodal knowledge graph to identify conflict sets with common root causes or mutual influence. For example, if a time limit conflict in the "customer complaint handling process", a functional requirement conflict in the "customer complaint handling system", and a performance evaluation standard conflict in the "customer service position" are detected, these three form a chain of association through "support" and "dependency" relationships, and are determined to be a set of related conflicts; then, a unified solution is generated for the set of related conflicts to ensure that the adjustment direction of each conflict is consistent and to avoid the exacerbation of other related conflicts by solving one conflict alone. According to the priority ranking result, batch solutions are generated first for the set of related conflicts containing high-priority conflicts.

[0054] In this embodiment of the application, step S500 involves visually and interactively displaying product conflicts and solutions, and collecting user feedback; The steps for the visual interactive display include F1~F2: F1. Display the conflict nodes and relationships of each product through a visual interface, and use different colors to mark the corresponding priorities; In this embodiment, the visualization interface adopts a knowledge graph format, using nodes to represent entity information, edges to represent architectural relationships, and highlighted markers to indicate conflict locations. For example, the visualization interface displays a conflict related to the "Customer Complaint Handling Process" document. This conflict involves nodes such as the "Electricity Marketing Business Management System" node, the "Customer Service Specification" node, the "Customer Complaint Handling Process" node, the "Customer Complaint Handling System" node, and the "Customer Service Position" node. These nodes are connected through architectural relationships such as "definition," "support," and "dependency." Different colors are used to indicate the priority of the conflict: high-priority conflicts are marked in red, medium-priority conflicts in orange, and low-priority conflicts in yellow. After a user clicks on a conflict node, the visualization interface supports drill-down viewing of the conflict details, displaying detailed information such as conflict type, tracing path, scope of impact, severity, urgency, priority score, and solutions, helping users to fully understand the conflict situation.

[0055] F2. Collect user feedback on the accuracy of product conflict detection results and the rationality of solutions, and use the feedback as the basis for optimizing conflict rules; In this embodiment, users can rate the accuracy of conflict detection and the reasonableness of the solution for each conflict. Rating options include "Accurate / Reasonable," "Partially Accurate / Reasonable," and "Inaccurate / Inappropriate." For example, a user's feedback on a conflict in the "Customer Complaint Handling Process" product, rated as "Accurate conflict detection, reasonable solution," is recorded as positive feedback; a user's feedback on a conflict in a certain product, rated as "Inaccurate conflict detection," is recorded as negative feedback. Finally, based on the user's positive and negative feedback, the matching conditions of the conflict detection rules, the weight parameters of priority evaluation, and the strategy template for solution generation are adjusted to continuously improve the accuracy of conflict detection and the reasonableness of the solutions.

[0056] In one optional implementation, step S500 involves visually displaying artifact conflicts and solutions, and collecting user feedback. It can also be presented in comparison to artifact documents. Specifically, the original documents of multiple enterprise architecture artifacts involved in the conflict are displayed side-by-side in the visualization interface, with the conflict location and content highlighted in the documents. For example, for a time-limit conflict in the "Customer Complaint Handling Process," Clause 3.2 of the "Electricity Marketing Business Management System" is displayed on the left, and Clause 5.1 of the "Customer Service Standard" is displayed on the right. The conflict content of "24 hours" and "48 hours" is highlighted in both documents, and a suggested solution is displayed in the middle area. Users can directly annotate the documents online, marking the accuracy of conflict detection, supplementing missing conflict information, and modifying the direction of solution adjustments. After collecting user feedback on online annotations, the conflict detection rules and solution strategies are updated to improve the accuracy of subsequent detection.

[0057] In summary, by constructing a multimodal knowledge graph, unified modeling and association mapping of enterprise architectural artifacts in different formats, such as Word documents, Excel spreadsheets, Visio flowcharts, and PowerPoint presentations, is achieved, breaking the limitation of traditional detection tools that can only handle single text formats. Based on this, entity information is extracted using a combination of rule matching and entity recognition models, and implicit relationships are mined using prompt words to guide the language model. This allows the knowledge graph to not only represent explicit architectural relationships but also capture potential semantic connections between different artifacts. In the conflict detection stage, rule matching and semantic similarity calculation are integrated, enabling simultaneous identification of explicit conflicts such as terminology repetition and direct contradictions, as well as implicit conflicts such as inconsistent terminology and logical incompatibility. Compared to traditional keyword matching methods, this improves detection accuracy. More importantly, based on the architectural relationships in the multimodal knowledge graph, cross-domain conflicts between business architecture, application architecture, and data architecture can be identified, solving the problem that existing technologies struggle to cover cross-domain conflicts. In terms of conflict resolution, it can not only pinpoint the root cause of conflicts but also prioritize them based on their scope, severity, and urgency. It then generates specific solutions by combining enterprise architecture standards and industry best practices from a reference knowledge base. These solutions include adjustment directions and implementation suggestions, providing directly actionable guidance for enterprise architecture managers. Furthermore, through visualization and user feedback mechanisms, conflict rules can be continuously optimized to improve the accuracy of subsequent detection, forming a closed-loop detection system. This reduces the conflict detection cycle for a single architecture system from weeks to hours, lowering labor costs and improving the overall efficiency and quality of enterprise architecture management.

[0058] Example 3 illustrates a schematic scheme for a conflict detection method for enterprise architectural artifacts based on multimodal knowledge graphs. It should be noted that the technical solution of this system for conflict detection of enterprise architectural artifacts based on multimodal knowledge graphs is based on the same concept as the technical solution of the aforementioned conflict detection method for enterprise architectural artifacts based on multimodal knowledge graphs. Details not described in detail in this embodiment can be found in the description of the technical solution of the aforementioned conflict detection method for enterprise architectural artifacts based on multimodal knowledge graphs.

[0059] This embodiment also provides an enterprise architecture artifact conflict detection system based on multimodal knowledge graphs, including: Data preprocessing module: used to collect and extract multimodal data from enterprise architecture artifacts, clean the multimodal data and standardize the terminology to obtain standard data; Knowledge graph construction module: used to extract entity information and architectural relationships from standard data, add multimodal attributes to each entity information and align them, and build a multimodal knowledge graph; Conflict detection module: used to define conflict rules, perform conflict detection on the multimodal knowledge graph according to the conflict rules, obtain product conflicts and locate the root cause of the conflict; Conflict Assessment and Solution Module: This module is used to prioritize and rank product conflicts and generate corresponding solutions by combining them with a reference knowledge base. Visualization and Feedback Module: Used to visualize product conflicts and solutions, and collect user feedback and evaluations.

[0060] This embodiment also provides an electronic device applicable to enterprise architecture artifact conflict detection based on multimodal knowledge graphs, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the enterprise architecture artifact conflict detection method based on multimodal knowledge graphs as proposed in the above embodiment.

[0061] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the enterprise architecture artifact conflict detection method based on multimodal knowledge graphs as proposed in the above embodiments.

[0062] The storage medium proposed in this embodiment and the enterprise architecture artifact conflict detection method based on multimodal knowledge graph proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0063] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. 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 can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0064] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting conflicts in enterprise architectural artifacts based on multimodal knowledge graphs, characterized in that, include: Collect and extract multimodal data from enterprise architecture artifacts, and perform terminology standardization processing on the multimodal data to obtain standard data; Entity information is extracted from the standard data, and the architectural relationships between the entity information are extracted. At the same time, multimodal attributes are added to each entity information. A multimodal knowledge graph is constructed based on the entity information, architectural relationships, and multimodal attributes. Define conflict rules, perform conflict detection on the multimodal knowledge graph according to the conflict rules, obtain product conflicts, trace the product conflicts, and locate the root cause of the conflict; The product conflicts are prioritized based on the root causes of the conflicts, and corresponding solutions are generated for each product conflict according to the priority ranking results.

2. The enterprise architecture artifact conflict detection method based on multimodal knowledge graph as described in claim 1, characterized in that, The steps to obtain standard data include: Based on the format of each enterprise architecture artifact, the corresponding parsing method is selected to extract information, resulting in multimodal data, which is then cleaned. The terminology of the cleaned multimodal data is standardized according to the preset terminology standard, and the synonyms of different expressions in the multimodal data are mapped to standard terms to obtain standard data.

3. The enterprise architecture artifact conflict detection method based on multimodal knowledge graph as described in claim 2, characterized in that, The steps for constructing a multimodal knowledge graph based on the entity information, architectural relationships, and multimodal attributes include: Entity information is extracted from the standard data by combining rule matching with an entity recognition model; The explicit relationships between the entity information are identified, and the implicit relationships between the entities are mined by using prompt words to guide the language model. The explicit and implicit relationships are combined to obtain the architectural relationships. Multimodal attributes are added to each of the entity information, an attention mechanism is used to align the multimodal attributes, and the entity information is aligned according to the alignment results of the multimodal attributes. A multimodal knowledge graph is constructed based on the aligned entity information, architectural relationships, and multimodal attributes.

4. The enterprise architecture artifact conflict detection method based on multimodal knowledge graph as described in claim 3, characterized in that, The steps to obtain a conflict of articles include: The multimodal knowledge graph is matched using the conflict rules to detect explicit conflicts. Calculate the semantic similarity of attributes among the entity information to detect implicit conflicts; By reasoning and analyzing the entity information through the architectural relationships in the multimodal knowledge graph, cross-domain conflicts are obtained; The explicit conflict, implicit conflict, and cross-domain conflict are combined to obtain the product conflict.

5. The enterprise architecture artifact conflict detection method based on multimodal knowledge graph as described in claim 4, characterized in that, The steps for prioritizing product conflicts based on the root causes of the conflicts include: The multimodal knowledge graph is analyzed based on the root causes of the conflict to obtain the number of entity information and enterprise architecture artifacts involved in the artifact conflict, and the scope of the artifact conflict is determined based on the number of entity information and enterprise architecture artifacts. Determine whether the product conflict violates preset compliance requirements and determine the severity of the product conflict; Determine whether the product conflict affects normal business operations and determine the urgency of the product conflict; The Analytic Hierarchy Process (AHP) is used to calculate priority scores for the scope of impact, severity, and urgency, and the product conflicts are ranked according to the priority scores.

6. The enterprise architecture artifact conflict detection method based on multimodal knowledge graph as described in claim 5, characterized in that, The steps for generating corresponding solutions for each product conflict based on the priority ranking results include: Based on the product conflict, relevant standards and specifications are retrieved from a pre-defined reference knowledge base; Based on the aforementioned standard specifications, the adjustment direction for each product conflict is determined according to the priority ranking result, and a solution is determined based on the adjustment direction.

7. The enterprise architecture artifact conflict detection method based on multimodal knowledge graph as described in claim 1 or 6, characterized in that, Also includes: The product conflicts and solutions are presented through visual interaction, and user feedback is collected. The steps involved in the visual interactive display include: The conflict nodes and relationships of each product are displayed through a visual interface, and different priorities are marked with different colors. Collect user feedback on the accuracy of the product conflict detection results and the rationality of the solutions, and use the feedback as the basis for optimizing the conflict rules.

8. A conflict detection system for enterprise architectural artifacts based on multimodal knowledge graphs, employing the method described in any one of claims 1-7, characterized in that, include: Data preprocessing module: used to collect and extract multimodal data from enterprise architecture artifacts, clean and standardize the terminology of the multimodal data to obtain standard data; Knowledge graph construction module: used to extract entity information and architectural relationships from the standard data, add multimodal attributes to each entity information and align them, and construct a multimodal knowledge graph; Conflict detection module: used to define conflict rules, perform conflict detection on the multimodal knowledge graph according to the conflict rules, obtain product conflicts and locate the root cause of the conflict; Conflict assessment and solution module: used to prioritize and rank the conflicts of the products, and generate corresponding solutions in conjunction with the reference knowledge base; Visualization and Feedback Module: This module is used to visualize the product conflicts and solutions, and to collect user feedback and evaluations.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the enterprise architecture artifact conflict detection method based on multimodal knowledge graph as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the enterprise architecture artifact conflict detection method based on a multimodal knowledge graph as described in any one of claims 1 to 7.

Images