Processing method, device, program product, and electronic device for system elements

Abstract

The application discloses a processing method and device of a system element, a program product and an electronic device, relates to the technical field of artificial intelligence, and the method comprises the following steps: performing semantic coding on standard words in a standard dictionary to obtain standard vectors, and constructing a standard vector knowledge base based on the standard vectors and standard word information corresponding to each standard word; performing semantic coding on original elements in an information system; inputting the original vectors obtained through coding into the standard vector knowledge base for semantic matching; performing compliance verification on the original elements based on M similar standard words obtained through matching, wherein the compliance verification comprises at least one of name semantic sub-verification, description text semantic sub-verification, enumerated value semantic sub-verification and enumerated value sequence sub-verification; in the case that the original elements are not compliant, generating a rectification scheme of the original elements based on the compliance verification result, and processing the original elements based on the rectification scheme. The application solves the technical problem of non-standard element setting in an enterprise information system.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a method, apparatus, program product, and electronic device for processing system elements. Background Technology

[0002] As enterprises deepen their digital transformation, various information systems are widely used within enterprises, forming a digital ecosystem with multiple systems operating collaboratively. However, due to differences in the construction cycle, development teams, and business needs of different information systems, there is a lack of unified standards for naming data elements between systems, resulting in a large number of non-standard data governance phenomena. For example, data elements with the same business meaning have multiple naming methods (such as "Customer Number", "Customer ID", "Customer Code"); the enumeration values ​​corresponding to the same enumeration class data elements in different systems are set differently and expressed inconsistently (such as the "Status" field of the enumeration class, which is set to "Enabled / Disabled" in some systems and "Activated / Inactive" in others), thus bringing serious challenges to enterprise knowledge data management and governance.

[0003] Non-standard element naming directly exacerbates data silos, increases the difficulty of data integration, and reduces knowledge reuse rates. This phenomenon not only increases the maintenance costs of enterprise information systems but may also lead to business decision-making errors due to semantic ambiguity in data elements. Currently, enterprises often rely on manual verification or simple keyword mapping to address this issue, which is inefficient, error-prone, and unable to meet the needs of large-scale, multi-system digital governance. Therefore, there is an urgent need for an intelligent and standardized digital governance solution to solve the problem of non-standard element naming and achieve efficient collaboration between enterprise knowledge management and data governance.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides a method, apparatus, program product, and electronic device for processing system elements, to at least solve the technical problem of non-standard element settings in enterprise information systems.

[0006] According to one aspect of this application, a method for processing system elements is provided, comprising: semantically encoding standard words in a standard dictionary to obtain standard vectors; constructing a standard vector knowledge base based on the standard vectors and standard word information corresponding to each standard word, wherein the standard dictionary is a set of fields generated based on the enterprise's preset business specifications; semantically encoding original elements in an information system to obtain original vectors, wherein the original elements are fields used to describe business data; inputting the original vectors into the standard vector knowledge base for semantic matching to obtain M similar standard words, wherein M is a positive integer, and the vector similarity between the original vectors and the standard vectors corresponding to each similar standard word is greater than a preset similarity; performing compliance verification on the original elements based on the M similar standard words to obtain verification results, wherein the compliance verification includes at least one of the following: name semantic sub-verification, description text semantic sub-verification, enumeration value semantic sub-verification, and enumeration value order sub-verification; and generating a rectification plan for the original elements based on the verification results when the original elements are non-compliant, and processing the original elements based on the rectification plan.

[0007] Optionally, semantic encoding is performed on the standard words in the standard dictionary to obtain standard vectors, including: obtaining standard word information for each standard word in the standard dictionary, wherein the standard word information includes at least one of the following: standard word name, description text, field type, enumeration value list, business scenario, and related words, and there is a business relationship between each standard word and its corresponding related words; generating semantically enhanced text for each standard word based on the standard word information of each standard word; and semantically encoding the semantically enhanced text of each standard word through a vector model to obtain a standard vector for each standard word, wherein the vector model is a large model trained based on the historical encoding records of business fields in the information system.

[0008] Optionally, before semantically encoding the original elements in the information system, the system element processing method further includes: collecting the original information of each original element in the information system through a collection module, wherein the collection module is used to simulate browser behavior; performing data cleaning, standardization processing and data filtering on the original information of each original element to obtain the element information of each original element, wherein the element information of each original element is used to generate semantically enhanced text for each original element, and the element information includes at least the element name, description text, field type, enumeration value list and business scenario.

[0009] Optionally, the original vector is input into a standard vector knowledge base for semantic matching to obtain M similar standard words, including: detecting the vector similarity between the original vector and the standard vector corresponding to each standard word in the standard vector knowledge base; taking the standard words with vector similarity greater than a preset similarity as the first standard words to obtain L first standard words corresponding to the original vector, where L is a positive integer greater than or equal to M; and performing optimization operations on the L first standard words to obtain M similar standard words, wherein the optimization operations are at least used to filter duplicate standard words and standard words that have business semantic conflicts with the element information of the original element.

[0010] Optionally, compliance verification of the original element is performed based on M similar standard words, including: determining the verification rules for matching the original element based on the M similar standard words, wherein the verification rules include at least one of the following: semantic verification rules and sequential verification rules; generating target prompt words based on the preset prompt word template corresponding to the verification rules, the standard word information of the M similar standard words, and the element information of the original element; inputting the target prompt words into the intelligent agent model, wherein the intelligent agent model is a large model trained based on the historical verification records of the business fields of the information system; and performing compliance verification of the original element based on the target prompt words through the intelligent agent model.

[0011] Optionally, compliance verification of the original element is performed using an intelligent agent model based on the target prompt word, including: performing a name semantic sub-verification based on the element name of the original element in the target prompt word and the standard word name of each similar standard word; if there is a similar standard word among the M similar standard words whose standard word name is the same as the element name, the original element is determined to pass the name semantic sub-verification; if there is no similar standard word among the M similar standard words whose standard word name is the same as the element name, performing a description text semantic sub-verification based on the description text of the original element in the target prompt word and the description text of each similar standard word; if there is a similar standard word among the M similar standard words whose semantic similarity is greater than a preset similarity, the original element is determined to pass the description text semantic sub-verification, where the semantic similarity is the similarity between the description text of the original element and the description text of each similar standard word.

[0012] Optionally, compliance verification of the original element is performed by an intelligent agent model based on the target prompt word, including: if the field type of the original element is an enumeration type, after performing name semantic sub-verification and / or description text semantic sub-verification, the intelligent agent model performs enumeration value semantic sub-verification based on the enumeration value list of the original element in the target prompt word and the enumeration value list of each similar standard word; if a first enumeration value list exists in the M enumeration value lists corresponding to the M similar standard words, the original element is determined to pass the enumeration value semantic sub-verification, wherein the enumeration values ​​in the first enumeration value list are the same as the enumeration values ​​in the enumeration value list of the original element; furthermore, the intelligent agent model performs enumeration value order sub-verification based on the enumeration value list of the original element in the target prompt word and the enumeration value list of each similar standard word; if a second enumeration value list exists in the M enumeration value lists corresponding to the M similar standard words, the original element is determined to pass the enumeration value semantic sub-verification, wherein the order of the enumeration values ​​in the second enumeration value list is consistent with the order of the enumeration values ​​in the enumeration value list of the original element.

[0013] Optionally, after inputting the original vector into the standard vector knowledge base for semantic matching to obtain M similar standard words, the system element processing method further includes: obtaining the recall hit rate of each standard word in the standard dictionary, wherein the recall hit rate is used to characterize the frequency of each standard word being identified as a similar standard word of the original element; storing the standard words with a recall hit rate greater than the preset hit rate, the standard vector of the standard word, and the standard word information into the preset cache area.

[0014] According to another aspect of this application, a system element processing apparatus is also provided, comprising: a knowledge base construction unit, used to construct a standard vector knowledge base based on the standard vector and standard word information corresponding to each standard word after semantically encoding standard words in a standard dictionary to obtain standard vectors, wherein the standard dictionary is a set of fields generated based on the enterprise's preset business specifications; an original element encoding unit, used to semantically encode original elements in the information system to obtain original vectors, wherein the original elements are fields used to describe business data; a similar standard word determination unit, used to input the original vectors into the standard vector knowledge base for semantic matching to obtain M similar standard words, wherein M is a positive integer, and the vector similarity between the original vector and the standard vector corresponding to each similar standard word is greater than a preset similarity; an original element verification unit, used to perform compliance verification on the original elements based on the M similar standard words to obtain verification results, wherein the compliance verification includes at least one of the following: name semantic sub-verification, description text semantic sub-verification, enumeration value semantic sub-verification, and enumeration value order sub-verification; and an original element processing unit, used to generate a rectification plan for the original elements based on the verification results when the original elements are non-compliant, and to process the original elements based on the rectification plan.

[0015] According to another aspect of this application, a computer program product is also provided, which stores a computer program, wherein a processing method for controlling the computer program product to execute any of the above-mentioned system elements is provided when the computer program is running.

[0016] According to another aspect of this application, an electronic device is also provided, wherein the electronic device includes one or more processors and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the processing method of the system element of any of the above.

[0017] In this application, after semantically encoding the standard words in the standard dictionary to obtain standard vectors, a standard vector knowledge base is first constructed based on the standard vector and standard word information corresponding to each standard word. The standard dictionary is a set of fields generated based on the enterprise's preset business specifications. Then, the original elements in the information system are semantically encoded to obtain original vectors, where the original elements are fields used to describe business data. Subsequently, the original vectors are input into the standard vector knowledge base for semantic matching to obtain M similar standard words, where M is a positive integer. The vector similarity between the original vector and the standard vector corresponding to each similar standard word is greater than the preset similarity. Then, the original elements are subjected to compliance verification based on the M similar standard words to obtain verification results. The compliance verification includes at least one of the following: name semantic sub-verification, description text semantic sub-verification, enumeration value semantic sub-verification, and enumeration value order sub-verification. In the case of non-compliance of the original elements, a rectification plan for the original elements is generated based on the verification results, and the original elements are processed based on the rectification plan.

[0018] As described above, this application employs a combination of vector semantic encoding and compliance verification. By constructing a standard vector knowledge base and combining it with semantic similarity retrieval technology to quickly recall similar standard words, and then using multi-dimensional compliance verification to intelligently detect the compliance of original elements in the information system, it achieves the goal of accurately identifying non-standard elements that are semantically identical to standard words but inconsistent in expression. This application does not rely on traditional manual verification methods or hard-coding rules, but instead uses semantic vectors to capture the business meaning of elements and combines enumerated value order sub-verification to achieve automated and high-precision identification of non-standard named elements. This achieves the technical effect of leaping from element literal matching to semantic governance, improving the efficiency of rectifying non-standard elements, and thus solving the technical problem of non-standard element settings in enterprise information systems. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 This is a flowchart of an optional system element processing method according to an embodiment of this application;

[0021] Figure 2 This is a flowchart of an optional system element compliance verification method according to an embodiment of this application;

[0022] Figure 3 This is a flowchart of an optional system element governance method according to an embodiment of this application;

[0023] Figure 4 This is a schematic diagram of a processing apparatus for an optional system element according to an embodiment of this application;

[0024] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] It should also be noted that all information and data (including but not limited to information used for display and analysis) involved in this application are authorized by the user or fully authorized by all parties. For example, if there is an interface between this system and the relevant user or organization, before obtaining the relevant information, it is necessary to send a request to the aforementioned user or organization through the interface, and obtain the relevant information only after receiving consent from the aforementioned user or organization.

[0028] Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of relevant information and data involved in this application all comply with the relevant laws, regulations, and standards of the relevant regions, and necessary security measures have been taken. They do not violate public order and good morals. In addition, this application provides corresponding operation entry points for users to choose to agree to authorization or refuse authorization. If the user chooses to refuse authorization, the corresponding expert decision-making process will be initiated.

[0029] In one optional related technical embodiment, a method for manually verifying and standardizing system data elements is provided. This method involves organizing technical and business personnel within the enterprise to manually review the element naming rules of each system, develop a unified standard dictionary, and then manually verify and modify the elements of each system one by one to ensure naming standardization. This solution relies on manual operation and can solve the problem of non-standard naming to a certain extent, but it has obvious limitations: First, it is extremely inefficient. For multi-system scenarios containing thousands of elements, manual verification requires a lot of manpower and resources and is time-consuming. Second, it has a low fault tolerance rate. Manual operation is prone to omissions and misjudgments, making it difficult to guarantee the quality of governance. Third, it has high maintenance costs. When system iterations and business updates lead to the addition or change of elements, manual verification needs to be reorganized, making dynamic adaptation impossible.

[0030] In one optional related technical embodiment, a keyword mapping matching method for system data elements is provided. This method pre-establishes a keyword mapping relationship between a standard dictionary and the naming of each system element (e.g., mapping "customer code" and "customer ID" to the standard term "customer number"). Through a keyword matching algorithm, it automatically identifies non-compliant elements in each system and performs preliminary corrections. This solution improves efficiency compared to manual verification, but it has several core drawbacks: First, the matching accuracy is low, only achieving literal keyword matching and failing to understand the semantic meaning of elements, easily leading to mismatches (e.g., "customer address" and "customer contact information" are misjudged as the same type because they contain the keyword "customer"); second, it has poor scalability, requiring manual maintenance of the mapping relationship when new elements are added or the standard dictionary is updated, making it unable to adapt to business changes; third, it cannot handle the problem of disordered enumeration values ​​for enumeration-type elements, only verifying the naming itself and failing to check the compliance of the enumeration values.

[0031] In one optional related technical embodiment, a rule-based compliance verification system is provided. Some enterprises have introduced a rule-based compliance verification system to automatically verify system elements based on preset regular expressions and naming rules (such as "fields should be named in camelCase" and "enumerated values ​​must conform to a preset list"). This system can perform basic compliance checks, but it suffers from insufficient flexibility: First, rule formulation relies on technical personnel, requiring the writing of numerous rules for different business scenarios and element types, resulting in high maintenance costs; second, the rules are rigid and cannot handle complex semantic ambiguities, failing to identify elements with the same semantics but different expressions; third, it lacks rectification recommendation functionality, only pointing out non-compliance of elements without providing users with specific rectification directions, increasing the difficulty of rectification.

[0032] In one optional related technical embodiment, a dictionary matching system is provided. This system stores an enterprise standard dictionary in a relational database and uses a string matching algorithm to compare the consistency of each system element with the standard dictionary to achieve compliance verification. This solution is simple in structure and easy to deploy, but it has significant shortcomings: First, the retrieval efficiency is low. When the standard dictionary is large (e.g., containing tens of thousands of standard words), string matching requires traversing the entire dictionary, resulting in high response latency. Second, it lacks semantic understanding capabilities and cannot distinguish between elements with the same meaning but different literal meanings. Third, it does not support the rapid retrieval of high-frequency words; frequent dictionary searches further degrade system performance, making it unsuitable for high-concurrency governance scenarios.

[0033] In summary, the above embodiments have the following technical defects:

[0034] (1) Lack of semantic recognition capability and low matching accuracy:

[0035] Related technologies (such as keyword mapping and simple dictionary matching) are all based on literal-level matching, which cannot understand the semantic meaning of elements and cannot solve the problem of non-standard naming where "the meaning is the same but the literal meaning is different". This easily leads to mismatches and missed matches, resulting in poor governance quality. For example, it cannot identify elements where "user age" and "customer age" have the same semantic meaning, nor can it distinguish the semantic difference between "order status - enabled" and "order status - activated", thus failing to meet the needs of enterprises for refined governance.

[0036] The purpose of this application is to introduce vector model and intelligent agent technology, to semantically encode standard words and system elements through vector model to achieve semantic-level matching; to accurately identify elements with the same meaning but different literal meaning by utilizing the natural language understanding ability of intelligent agents, to solve the problem of semantic ambiguity, and at the same time to realize semantic compliance checks of enumeration values, thereby significantly improving matching accuracy.

[0037] (2) Low retrieval efficiency, unable to adapt to high-concurrency scenarios:

[0038] In related technologies, standard dictionaries are mostly stored in relational databases and retrieved using string traversal matching. When the standard dictionary is large and there are many retrieval requests, the response latency is high, making it unsuitable for enterprise digital governance scenarios with multiple systems and high concurrency. Furthermore, no special processing is done for high-frequency words; frequent retrieval of the same words further consumes system resources and reduces governance efficiency.

[0039] The purpose of this application is to utilize a vector model knowledge base to store a standard dictionary, enabling fast semantic retrieval of standard words; and to introduce Redis caching technology to cache high-frequency standard words and high-frequency retrieval results, thereby shortening retrieval response time, improving the system's concurrent processing capabilities, and ensuring efficient operation even in large-scale, high-concurrency scenarios.

[0040] (3) The data collection method is singular and has poor adaptability:

[0041] Data collection for related technologies often relies on fixed interfaces or manual import, which cannot be adapted to different types and architectures of enterprise information systems (such as outdated systems lacking standardized interfaces and heterogeneous systems with significantly different data formats). This results in the ineffective collection of raw data from some systems, affecting the comprehensiveness of governance. Furthermore, the collection process lacks automation mechanisms, requiring manual intervention and leading to low efficiency.

[0042] The purpose of this application is to design a self-developed data acquisition method that supports multiple acquisition modes (interface acquisition, direct database connection acquisition, file import acquisition, and web crawling acquisition), adapts to different architectures and types of enterprise information systems, and achieves automated and comprehensive acquisition of raw data without manual intervention, thereby improving data acquisition efficiency and coverage.

[0043] (4) Standard dictionaries are inconvenient to maintain and have poor scalability:

[0044] In related technologies, the maintenance of standard dictionaries often relies on technical personnel manually modifying the database or configuration files, which is complex and cannot be handled by non-technical personnel. Furthermore, it does not support batch import or batch modification, resulting in a huge workload when the standard dictionary needs to be updated on a large scale, making it impossible to quickly adapt to business changes.

[0045] The purpose of this application is to provide a visual maintenance interface and batch import function, enabling non-technical personnel to quickly add, modify, and delete standard words through an intuitive interface, supporting batch import and export in preset formats, simplifying the maintenance process of the standard dictionary, improving maintenance efficiency, and realizing the dynamic updating and expansion of the standard dictionary.

[0046] (5) Lack of a rectification and recommendation mechanism, and an imperfect governance closed loop:

[0047] The relevant technologies can only identify non-compliant elements, but cannot provide users with specific rectification suggestions. Users need to manually refer to the standard dictionary to find rectification solutions, which is inefficient. Furthermore, the lack of a rectification feedback mechanism makes it impossible to track and verify the rectification effect and form a closed loop for governance. This results in some non-compliant elements not being rectified thoroughly, affecting the overall governance effect.

[0048] The purpose of this application is to generate a Top 10 rectification recommendation list for non-compliant elements based on intelligent agents and RAG (Retrieval-Augmented Generation) technology, clarifying the rectification direction and priority; to establish a governance closed loop of "collection-verification-rectification-feedback" to track the rectification effect, ensure that all non-compliant elements are rectified in place, and improve the integrity and effectiveness of digital governance.

[0049] As can be seen from the above, this application provides a method for digital governance of elements in information systems, aiming to solve the technical defects existing in the above-mentioned prior art. This application focuses on the pain point of non-standard element naming between enterprise systems, and uses "vector + intelligent agent" as the core to build an intelligent and standardized element digital governance system. The specific technical problems to be solved include:

[0050] (1) How to achieve efficient storage and semantic retrieval of the standard dictionary and solve the problems of low matching accuracy and lack of semantic recognition capability in traditional keyword mapping: Related technologies use relational databases to store the standard dictionary and rely on keyword matching to verify elements. They cannot understand the semantics of elements, resulting in low matching accuracy and high mismatch rate. Therefore, it is necessary to design a knowledge base based on a vector model to semantically encode the standard words and store them in vector form to achieve semantic-level retrieval of standard words. At the same time, combined with RAG technology, standard words with similar semantics to system elements can be quickly recalled to support subsequent compliance verification and completely solve the semantic ambiguity problem.

[0051] (2) How to design an adaptive data collection method to achieve comprehensive and automated collection of raw data from different types of enterprise information systems: Enterprise information systems are diverse in type and heterogeneous in architecture. Some older systems lack standardized interfaces, and existing collection methods cannot adapt to them, resulting in incomplete and inefficient data collection. Therefore, it is necessary to develop an adaptive data collection method that supports multiple collection modes, can automatically identify system types and adapt data formats, and achieve automated collection, cleaning, and standardization of raw data to ensure the integrity and accuracy of the collected data and provide a high-quality data foundation for subsequent compliance verification.

[0052] (3) How to use intelligent agent technology to achieve intelligent compliance verification of elements, especially the compliance check of enumeration values ​​of enumeration-type elements, and solve the problems of disordered enumeration value order and inconsistent expression: Related technologies cannot achieve fine-grained compliance checks of enumeration-type elements and cannot solve the problems of disordered enumeration value order and inconsistent expression. Therefore, it is necessary to use the natural language understanding and reasoning ability of intelligent agents, combined with the enumeration rules in the standard dictionary, to perform intelligent compliance verification of system elements. This can not only determine whether the element naming is compliant, but also perform semantic verification and order checks on the enumeration values ​​of enumeration-type elements to ensure the standardization and consistency of enumeration values.

[0053] (4) How to improve the system's retrieval and verification efficiency to adapt to high-concurrency governance scenarios: When the standard dictionary is large and there are many governance requests, the traditional retrieval method has a high response latency and cannot adapt to high-concurrency scenarios. Therefore, it is necessary to introduce Redis caching technology to cache high-frequency standard words, high-frequency retrieval results and high-frequency verification results, reduce repeated retrieval and calculation, shorten response time, improve the system's concurrent processing capability, and ensure that the system can still run efficiently in large-scale, high-concurrency scenarios.

[0054] (5) How to design a convenient standard dictionary maintenance method that supports non-technical personnel and enables dynamic updates and expansion of the standard dictionary: The maintenance of related technical standard dictionaries is complex, and non-technical personnel cannot participate, resulting in low maintenance efficiency. Therefore, it is necessary to design a visual maintenance interface that allows non-technical personnel to quickly maintain the standard dictionary through an intuitive interface. At the same time, it should support batch import and export functions to achieve batch updates of the standard dictionary, adapt to rapid changes in business, and reduce maintenance costs.

[0055] (6) How to achieve rectification recommendation and governance closed loop for non-compliant elements to improve the integrity and effectiveness of digital governance: The relevant technologies lack rectification recommendation mechanisms and governance closed loops, resulting in low rectification efficiency and poor results. Therefore, it is necessary to generate a Top 10 rectification recommendation list for non-compliant elements based on intelligent agents and RAG technology, combined with a standard dictionary, to clarify the rectification priority and specific solutions; at the same time, a rectification feedback mechanism should be established to track the rectification effect and form a complete governance closed loop of "collection-verification-rectification-feedback" to ensure that the governance work is implemented effectively.

[0056] The present application will now be described in detail with reference to various embodiments.

[0057] Example 1

[0058] According to an embodiment of this application, an embodiment of a method for processing system elements is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0059] This application provides a system for processing system elements (hereinafter referred to as the processing system) for executing the system element processing method of this application. Figure 1 This is a flowchart of an optional system element processing method according to an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0060] Step S101: After semantically encoding the standard words in the standard dictionary to obtain standard vectors, a standard vector knowledge base is constructed based on the standard vectors and standard word information corresponding to each standard word. The standard dictionary is a set of fields generated based on the enterprise's preset business specifications.

[0061] Optionally, to address the issues of low storage efficiency and lack of semantic recognition capabilities in traditional standard dictionaries, the processing system constructs a knowledge base based on a vector model (i.e., a standard vector knowledge base), thereby achieving semantic storage and efficient management of the standard dictionary.

[0062] Optionally, semantic encoding refers to the process of converting the natural language text corresponding to standard words into numerical vectors through a vector model; the standard vector knowledge base refers to a structured database that stores at least standard word vectors and their meta-information (i.e., standard word information, such as enumeration values), supports efficient semantic retrieval and indexing mechanisms, and supports filtering by business scenarios, field types, and other dimensions to improve retrieval efficiency.

[0063] Optionally, the standard dictionary is jointly developed by the enterprise's business and technical personnel, and contains standard element information for various business areas of the enterprise, including: standard word name, semantic description, field type (ordinary field, enumerated field), enumerated value list (for enumerated fields), business scenario, associated standard words, etc. The standard dictionary can be manually entered through a visual maintenance interface, or it can be imported in batches through preset formats to support subsequent dynamic updates and maintenance. Then, for each standard word in the standard dictionary, the processing system uses a lightweight, high-precision vector model (such as Qwen2.5-7B-Embedding, Llama3-8B-Embedding) to perform semantic encoding on each standard word in the standard dictionary.

[0064] Optionally, compared to the traditional method of storing standard words in a relational database, the processing system obtains standard vectors by semantically encoding the standard words in the standard dictionary, thereby digitizing the semantic information of standard words and avoiding the inefficiency and subjective omissions of manually building database mapping tables. Furthermore, the standard vector knowledge base constructed in this application supports dynamic updates. After the standard dictionary is modified, it can automatically perform incremental encoding and synchronize it to the standard vector knowledge base, thus realizing the automatic and continuous evolution of standard word knowledge.

[0065] Step S102: Semantically encode the original elements in the information system to obtain the original vector, where the original elements are fields used to describe business data.

[0066] Optionally, the original element refers to the actual, non-standard named field in the information system, which is a digital object to be governed.

[0067] Optionally, before semantically encoding the original elements in the information system, the processing system first simulates browser behavior through the acquisition module to collect the original information (such as form fields and table column names) of each original element in the information system. The acquisition module supports custom acquisition rules to filter irrelevant information and ensure the accuracy of the collected data. Afterward, the acquisition module performs data cleaning, standardization, and data filtering on the original information of each original element to remove invalid data and complete the standard formatting of element names and enumeration values, thereby extracting the core information of the original element (e.g., field names, field types, enumeration value lists, and business scenario descriptions) and obtaining the standardized element information of the original element.

[0068] Optionally, the processing system performs semantic encoding on the original elements in the information system, which is consistent with the semantic encoding step in step S101. That is, the processing system performs semantic encoding on the original elements through the same vector model to ensure that the original vector and the standard vector are in the same semantic space and can be directly compared. This avoids the semantic misalignment problem caused by the inconsistency of the semantic space, thereby improving the accuracy of subsequent element matching.

[0069] Step S103: Input the original vector into the standard vector knowledge base for semantic matching to obtain M similar standard words, where M is a positive integer, and the vector similarity between the original vector and the standard vector corresponding to each similar standard word is greater than the preset similarity.

[0070] Optionally, semantic matching refers to judging the degree of similarity between two semantic expressions based on distance metrics (such as cosine similarity) in vector space.

[0071] Optionally, after inputting the original vector into the standard vector knowledge base, the processing system first uses the cosine similarity method to detect the vector similarity between the original vector and the standard vector corresponding to each standard word, and then selects L first standard words whose vector similarity is greater than the preset similarity. After that, the processing system performs optimization operations on the L first standard words, that is, filters out the duplicate standard words and standard words that conflict with the business semantics of the element information of the original element, and finally obtains M similar standard words.

[0072] Optionally, to achieve semantic-level matching between system elements and standard words, the processing system introduces Search Enhanced Generation (RAG) technology, which uses a vector knowledge base to quickly recall standard words that are semantically similar to the original elements in the information system, providing a data foundation for subsequent compliance verification.

[0073] Step S104: Perform compliance verification on the original element based on M similar standard words to obtain the verification result. The compliance verification includes at least one of the following: name semantic sub-verification, description text semantic sub-verification, enumeration value semantic sub-verification, and enumeration value order sub-verification.

[0074] Optionally, the name semantic sub-validation is used to detect whether the element name of the original element is the same as the element name of the similar standard words; the description text semantic sub-validation is used to detect whether the semantic similarity between the description text of the original element and the description text of the similar standard words is greater than a preset similarity; the enumeration value semantic sub-validation is used to detect whether the set of enumeration values ​​of the original element of the enumeration type is the same as the set of enumeration values ​​of the similar standard words; and the enumeration value order sub-validation is used to detect whether the order of the enumeration values ​​of the original element of the enumeration type is the same as the order of the enumeration values ​​of the similar standard words.

[0075] Optionally, the processing system, in conjunction with the actual business scenario corresponding to the information system, filters the original element matching verification rules based on M similar standard words. Then, based on the preset prompt word template corresponding to the verification rules, the standard word information of the M similar standard words, and the element information of the original element, the system generates the target prompt word. Subsequently, the processing system selects an intelligent agent model with high semantic understanding capabilities (such as Qwen2.5-256B-Instruct, Llama3-128B-Instruct). The intelligent agent model is a large model trained based on the historical verification records of the business fields of the information system. The target prompt word is input into the intelligent agent model, and the intelligent agent model performs compliance verification on the original element based on the target prompt word. Thus, by utilizing the natural language understanding and reasoning capabilities of the intelligent agent, combined with the similar standard word library, the system intelligently performs compliance verification on the original element of the information system, thereby solving the problems of semantic inconsistency between the original element and the standard word and disordered enumeration value order.

[0076] Step S105: If the original element is non-compliant, generate a rectification plan for the original element based on the verification result, and process the original element based on the rectification plan.

[0077] Optionally, addressing the technical issues of low rectification efficiency and incomplete governance loop in traditional technologies for non-compliant original elements, the processing system, based on an intelligent agent model and RAG technology, rapidly generates rectification plans for non-compliant original elements and establishes a rectification feedback mechanism to form a complete governance loop. The specific process is as follows:

[0078] (1) Generation of the Top 10 Rectification Recommendation List:

[0079] The intelligent agent model combines M similar standard words, verification rules, and enterprise business scenarios to generate a Top 10 rectification recommendation list for each non-compliant element. The Top 10 recommendation list is sorted by rectification priority (priority is determined by similarity, business importance, and rectification difficulty). Each of the 10 recommended items includes: recommended similar standard words, rectification suggestions (such as "change 'customer code' to 'customer number'"), rectification basis, rectification difficulty, and estimated rectification time, so as to facilitate users to quickly select the appropriate rectification plan. For enumerated elements, the intelligent agent model also provides suggestions for adjusting the enumerated values ​​to clarify the correct enumerated value order and expression.

[0080] (2) Implementation and feedback of rectification:

[0081] Users can view the original non-compliant elements and the corresponding Top 10 rectification recommendations through a visual interface. They can then select an appropriate rectification plan, implement the rectification operation, and submit rectification feedback on the interface after completion. The processing system automatically re-collects the element information and re-verifies compliance to verify the rectification effect. If the rectified element passes the compliance verification, it is marked as "rectification completed"; if the rectified element is still non-compliant, the processing system regenerates the rectification recommendation list, prompting the user to adjust the rectification plan until the element is rectified.

[0082] (3) Optimization of governance closed loop:

[0083] The processing system records the rectification records of all non-compliant original elements (including rectification time, rectification plan, and rectification effect), forming a governance log for users to query and analyze. At the same time, the processing system regularly performs statistical analysis on the governance log data to identify the original types of high-frequency non-compliant elements and key rectification issues, thereby providing data support for enterprises to optimize the standard dictionary and improve governance rules. This continuously improves the quality and efficiency of digital governance of elements in the information system, forming a complete governance closed loop of "collection-verification-rectification-feedback-optimization".

[0084] As described above, this application employs a combination of vector semantic encoding and compliance verification. By constructing a standard vector knowledge base and combining it with semantic similarity retrieval technology to quickly recall similar standard words, and then using multi-dimensional compliance verification to intelligently detect the compliance of original elements in the information system, it achieves the goal of accurately identifying non-standard elements that are semantically identical to standard words but inconsistent in expression. This application does not rely on traditional manual verification methods or hard-coding rules, but instead uses semantic vectors to capture the business meaning of elements and combines enumerated value order sub-verification to achieve automated and high-precision identification of non-standard named elements. This achieves the technical effect of leaping from element literal matching to semantic governance, improving the efficiency of rectifying non-standard elements, and thus solving the technical problem of non-standard element settings in enterprise information systems.

[0085] In one optional embodiment, the processing system first obtains the standard word information for each standard word in the standard dictionary. The standard word information includes at least one of the following: standard word name, description text, field type, enumeration value list, business scenario, and related words. There is a business relationship between each standard word and its corresponding related words. Then, the processing system generates semantically enhanced text for each standard word based on the standard word information. Then, the processing system performs semantic encoding on the semantically enhanced text of each standard word through a vector model to obtain the standard vector of each standard word. The vector model is a large model trained based on the historical encoding records of the business fields of the information system.

[0086] Optionally, the processing system uses a lightweight, high-precision vector model (such as Qwen2.5-7B-Embedding or Llama3-8B-Embedding) to semantically encode standard words in the standard dictionary. The steps include: first, preprocessing the standard word information, such as the semantic description text and business scenarios, to generate semantically enhanced text; then, using the vector model to transform the semantically enhanced text into fixed-dimensional standard vectors (such as 768-dimensional vectors) to capture the semantic features of the standard words; finally, the processing system fine-tunes the vector model using historical data from within the enterprise, thereby improving the accuracy of the vector model in semantically recognizing enterprise business terms and ensuring the accuracy of vector encoding.

[0087] Optionally, the storage and management steps of the standard vector knowledge base by the processing system include: storing the standard vectors and standard word information (such as names, semantic descriptions, enumeration values, etc.) of standard words into a vector database to construct a standard vector knowledge base. This standard vector knowledge base supports efficient semantic retrieval and can quickly recall semantically similar standard words based on element vectors. At the same time, the processing system establishes an indexing mechanism for standard words in the standard vector knowledge base, supporting filtering by business scenarios, field types, and other dimensions to improve retrieval efficiency. In addition, the standard vector knowledge base can be linked with a visual maintenance module. When the standard dictionary is updated, vector re-encoding and knowledge base updates are automatically triggered to ensure that the content of the standard vector knowledge base is consistent with that of the standard dictionary.

[0088] Optionally, to address the issues of inconvenient maintenance of the standard dictionary and the inability of non-technical personnel to participate, the processing system is designed with a visual maintenance module, supporting convenient maintenance and dynamic updates of the standard dictionary. The functional settings of the visual maintenance module are as follows:

[0089] (1) Visualized operation interface: The interface is designed to be simple and intuitive, supporting operation by non-technical personnel (such as business personnel). The interface includes functions such as standard dictionary list, adding standard words, modifying standard words, deleting standard words, batch import and export, and query filtering. Users can quickly view detailed information of standard words (semantic description, enumeration values, business scenarios, etc.) through the interface without writing code or operating the database.

[0090] (2) Batch import and export function: Supports batch import and export of standard dictionaries in preset formats (such as CSV (Comma-Separated Values, a format for storing tabular data in plain text) and JSON (JavaScript Object Notation, a lightweight data exchange format)). When importing in batches, the system automatically parses the file content and verifies the legality of the data format. If there is erroneous data, the system prompts the user to correct it before importing again. When exporting in batches, standard words can be filtered according to business scenarios, field types and other dimensions, and exported as files in the specified format, which facilitates the backup and sharing of standard dictionaries.

[0091] (3) Dynamic Updates and Version Management: The system supports dynamic updates of the standard dictionary. Users can add, modify, and delete standard words at any time. After an update, the vector knowledge base is automatically updated and the Redis cache is invalidated to ensure data consistency. At the same time, the system records the update history of the standard dictionary and supports version rollback. If a problem occurs after an update, it can be quickly rolled back to the previous version to reduce maintenance risks.

[0092] Optionally, by using the steps in the above embodiments to generate standard vectors for each standard word through a vector model, the processing system can achieve the following technical effects:

[0093] (1) Improve the quality of semantic encoding: Traditional technology only encodes field names and ignores contextual information such as description and scenario. This may result in elements with similar semantics being encoded as vectors with large differences. The processing system constructs semantically enhanced text based on standard word information such as standard word name, description text, field type, enumeration value list, business scenario and related words. This enables the final semantic encoding result to fully reflect the complete business meaning of the field, improves the quality of semantic encoding of standard words, and reduces mismatches.

[0094] (2) Improve the encoding accuracy of standard words by training vector models based on historical encoding data: General models (such as BERT, which stands for Bidirectional Encoder Representations from Transformers, refers to a bidirectional encoder based on transformers) often cannot recognize internal specialized terms in enterprise scenarios. The processing system fine-tunes the large model by using enterprise historical encoding data, enabling it to learn enterprise-specific semantic mapping knowledge, thereby improving the domain adaptability of the model encoding.

[0095] In one optional embodiment, before semantically encoding the original elements in the information system, the processing system first collects the original information of each original element in the information system through a collection module, wherein the collection module is used to simulate browser behavior. Then, the processing system performs data cleaning, standardization processing and data filtering on the original information of each original element to obtain the element information of each original element. The element information of each original element is used to generate semantically enhanced text for each original element. The element information includes at least the element name, description text, field type, enumeration value list and business scenario.

[0096] Optionally, the processing system provides a data acquisition module for obtaining element information for each raw element, in order to solve the problems of poor adaptability and low efficiency of traditional element information acquisition methods. The processing system uses the acquisition module to automate the acquisition of raw data from different types of enterprise information systems. The specific process is as follows:

[0097] (1) Automatic data collection: The collection module collects the original information of the original elements in the webpage (such as form fields and table column names) by simulating browser behavior. The collection module supports custom collection rules to filter irrelevant information and ensure the accuracy of the collected original information.

[0098] (2) Data cleaning and standardization: After collecting the original information, the collection module cleans and standardizes the original information data. First, it removes invalid data (such as null values, duplicate data, and garbled data); second, it standardizes and formats the names and enumeration values ​​of elements (such as unifying uppercase and lowercase and removing special characters); third, it extracts the core information of the elements (field names, field types, enumeration value lists, and business scenario descriptions) to generate standardized element information and store it in the system database, providing a high-quality data foundation for subsequent semantic retrieval and compliance verification.

[0099] Optionally, the processing system collects and preprocesses the raw information of each raw element in the information system through the acquisition module, achieving the following technical effects:

[0100] (1) Realize the collection of original elements of information systems without interfaces: Traditional collection methods rely on direct database connection or API interface. Due to the lack of interfaces, old systems cannot collect original elements. The processing system simulates browser behavior and can directly obtain field information from the UI (User Interface) layer without system cooperation, which solves the technical problems of some old systems having no standardized interfaces and difficulty in collecting original information.

[0101] (2) Improve element information to provide reliable input for semantic modeling: The collected raw information contains a lot of noise, which will reduce the matching accuracy if used directly for semantic encoding. The processing system avoids noise information from interfering with subsequent semantic vector modeling through three processing steps: data cleaning, standardization and data filtering, thereby ensuring the reliability of semantic modeling results.

[0102] (3) Supports complete extraction of enumeration values ​​and business scenarios to support subsequent compliance verification: The collection module not only extracts field names, but also completely captures the drop-down enumeration values ​​of elements (such as "Enable, Disable") and the business modules to which they belong (such as "Customer Management"), thereby providing a reliable basis for subsequent semantic sub-verification of enumeration values.

[0103] In summary, the steps in the above embodiments constitute an automated element information collection process for heterogeneous information systems. The collection module achieves full coverage collection of interface-less information systems by simulating browser behavior. Through data cleaning, standardization, and data filtering, the quality of the processed element information is improved, and finally, structured and standardized element information is output, thereby providing high-quality input data for subsequent semantically enhanced text generation and semantic matching.

[0104] In one optional embodiment, the processing system first detects the vector similarity between the original vector and the standard vector corresponding to each standard word in the standard vector knowledge base. Then, the processing system takes the standard words with vector similarity greater than a preset similarity as the first standard words, and obtains L first standard words corresponding to the original vector, where L is a positive integer greater than or equal to M. Then, the processing system performs optimization operation on the L first standard words to obtain M similar standard words, wherein the optimization operation is at least used to filter duplicate standard words and standard words that have business semantic conflicts with the element information of the original element.

[0105] Optionally, the processing system semantically encodes the element information of the original elements output by the acquisition module using a vector model, transforming it into an original vector with the same dimension as the standard vector, thereby capturing the semantic features of the original elements in the information system. Then, the processing system performs semantic similarity retrieval on the original vector based on RAG technology. The steps include: inputting the original vector into a vector knowledge base; calculating the similarity between the original vector and the standard vector using a cosine similarity algorithm; filtering out standard words corresponding to standard vectors with similarity higher than a preset similarity (e.g., 0.8) to form a similar standard word library (this similar standard word library includes L first standard words); then, using RAG technology and combining standard word information such as business scenarios and field types, the standard words in the similar standard word library are sorted, prioritizing the recall of similar standard words that match the business scenarios and field types of the original elements, thereby improving retrieval accuracy.

[0106] Optionally, the processing system optimizes the standard words in the similar standard word library by: first, removing duplicate standard words; second, filtering similar standard words that are semantically unrelated to the original system elements but have high similarity; and third, supplementing the association information of the standard words (such as the enumeration value list and the generated historical rectification plan), thereby outputting the optimized similar standard word library (which includes M similar standard words) to be passed to the compliance verification module in the intelligent agent model for subsequent compliance verification.

[0107] Optionally, the processing system recalls M similar standard words corresponding to the original element through the steps in the above embodiments, which can achieve the following technical effects:

[0108] (1) Control the scope of recalled standard words by setting similarity threshold: Without setting similarity threshold, a large number of semantically irrelevant standard words may be recalled. For example, "customer address" and "customer phone" are mismatched by meta-elements because they contain the word "customer". Setting a preset similarity can ensure that only semantically highly related standard words are recalled, thereby reducing the false recall rate and improving the accuracy of subsequent verification.

[0109] (2) Remove duplicate standard words by optimizing the operation to ensure the simplicity and consistency of the results: The standard dictionary may have duplicate entries during long-term maintenance. If deduplication optimization is not performed, the same semantic standard words will be repeatedly recommended, interfering with the accurate generation of rectification plan. The processing system improves the standardization of similarity standard deviation recalled by the processing system by performing deduplication operation.

[0110] (3) Introduce business semantic conflict filtering to reduce the probability of mismatch: Traditional technology only detects the similarity between vectors. The processing system filters out standard words that are literally similar but actually have business semantic conflicts by using the "business scenario" field of the original elements to be matched and the standard words, so as to avoid the problem of "semantic separation" in the original elements of the final output M similar standard words.

[0111] In one optional embodiment, the processing system first determines the verification rules for matching the original element based on M similar standard words. The verification rules include at least one of the following: semantic verification rules and sequential verification rules. Then, the processing system generates target prompt words based on the preset prompt word template corresponding to the verification rules, the standard word information of the M similar standard words, and the element information of the original element. Then, the processing system inputs the target prompt words into the intelligent agent model, wherein the intelligent agent model is a large model trained based on the historical verification records of the business fields of the information system. Subsequently, the processing system performs compliance verification on the original element based on the target prompt words through the intelligent agent model.

[0112] Optionally, by using the steps in the above embodiments, the processing system performs compliance verification on the original elements based on the verification rules in the target prompt words through an intelligent agent model, thereby achieving the following technical effects:

[0113] (1) Automatic matching of verification rules to achieve differentiated compliance judgment: The processing system automatically selects semantic verification rules or sequential verification rules based on the field type of M similar standard words. For example, sequential verification is automatically enabled for "order status" and semantic verification is only enabled for "customer number". This achieves accurate verification that varies depending on the field type and avoids misuse of verification rules.

[0114] (2) The generation efficiency of target prompt words is improved by using prompt word templates: the preset prompt word templates are structured and reusable, avoiding the subjectivity and inefficiency of manually writing instructions. Then, the processing system fills the standard word information of M similar standard words and the element information of the original elements into the preset prompt word templates, ensuring the consistency and interpretability of the generated target prompt words.

[0115] (3) Improve the efficiency of multi-dimensional compliance verification by using intelligent agent model: The intelligent agent model performs compliance verification on the original elements based on the multi-dimensional verification rules (such as semantic verification rules and sequential verification rules) in the target prompt words, avoiding manual verification in traditional technology, thereby improving the efficiency of compliance verification.

[0116] In one optional embodiment, the processing system performs a name semantic sub-verification based on the element name of the original element in the target prompt word and the standard word name of each similar standard word using an intelligent agent model. If there is a similar standard word among the M similar standard words whose standard word name is the same as the element name, the original element is determined to pass the name semantic sub-verification. If there is no similar standard word among the M similar standard words whose standard word name is the same as the element name, the system performs a description text semantic sub-verification based on the description text of the original element in the target prompt word and the description text of each similar standard word using an intelligent agent model. If there is a similar standard word among the M similar standard words whose semantic similarity is greater than a preset similarity, the original element is determined to pass the description text semantic sub-verification, where the semantic similarity is the similarity between the description text of the original element and the description text of each similar standard word.

[0117] Optionally, the processing system can achieve the following technical effects through the steps in the above embodiments:

[0118] (1) Layered verification logic to improve judgment accuracy: The step-by-step verification mechanism of "name verification first, semantic verification later" is adopted. It prioritizes the identification of similar standard words with the same name, avoids unnecessary semantic analysis of similar standard words with different names, reduces the invalid calculation of the agent model, and thus improves the verification efficiency.

[0119] (2) Semantic understanding is achieved by describing the semantic sub-verification of the text: Traditional technology only verifies the field name of the original element, and there may be misjudgment of the field name of the original element. The processing system detects the similarity between the description text corresponding to the original element and the similar standard word through the intelligent agent model. It can accurately detect whether the original element and the similar standard word refer to the same business entity semantically, making up for the shortcomings of single-dimensional detection of field name.

[0120] In an optional embodiment, when the field type of the original element is an enumeration type, after performing name semantic sub-validation and / or description text semantic sub-validation, the processing system performs enumeration value semantic sub-validation based on the enumeration value list of the original element in the target prompt word and the enumeration value list of each similar standard word through an intelligent agent model; if a first enumeration value list exists in the M enumeration value lists corresponding to the M similar standard words, the original element is determined to pass the enumeration value semantic sub-validation, wherein the enumeration values ​​in the first enumeration value list are the same as the enumeration values ​​in the enumeration value list of the original element; and, based on the enumeration value list of the original element in the target prompt word and the enumeration value list of each similar standard word through an intelligent agent model, the enumeration value order sub-validation is performed; if a second enumeration value list exists in the M enumeration value lists corresponding to the M similar standard words, the original element is determined to pass the enumeration value semantic sub-validation, wherein the order of the enumeration values ​​in the second enumeration value list is consistent with the order of the enumeration values ​​in the enumeration value list of the original element.

[0121] Optionally, the processing system, through the steps in the above embodiments, first performs a semantic sub-verification of the enumeration value for the original element of the enumeration type. After the semantic sub-verification of the enumeration value of the original element passes, it performs a sequential sub-verification of the enumeration value for the original element. This compliance verification mechanism can achieve the following technical effects:

[0122] This application achieves the following technical effects by performing semantic sub-verification and order sub-verification on the original elements:

[0123] (1) Solving the business ambiguity problem caused by inconsistent semantics of enumeration values: In the heterogeneous information system environment of enterprises, enumeration values ​​with the same business meaning are often expressed in different ways, such as "enabled / disabled", "activated / inactive", "active / inactive", etc. Traditional methods can only perform string matching and cannot identify the semantic equivalence of these expressions, leading to the system misjudging them as non-compliant. This application uses an intelligent agent model to perform semantic comparison between the enumeration values ​​of the original elements and similar standard words, which can identify enumeration values ​​such as "inactive" and "disabled", "active" and "enabled" that have the same semantics but different values, ensuring that semantically consistent enumeration values ​​can be correctly recognized, thereby eliminating the misjudgment of the original elements caused by differences in terminology.

[0124] (2) Correcting system logic errors caused by disordered enumeration values: Some business processes (such as state machines and approval flows) rely on the order of enumeration values ​​to define the legal flow path. If the order of "disabled, enabled" in the system is reversed from the standard "enabled, disabled", it may cause the process engine to misjudge. On the basis of confirming the semantic consistency of the enumeration values, the processing system further performs order sub-checks on the original elements and similar standard words to avoid business process interruption or abnormality caused by the different order settings of the enumeration values ​​of the same original element in different business modules in heterogeneous information systems.

[0125] In summary, by performing semantic and sequential sub-validations on the original elements, the processing system achieves dual protection of the semantic consistency and sequential standardization of enumeration fields in enterprise information systems. This solves the technical problems of chaotic enumeration value representation, disordered order, and misuse by downstream systems in heterogeneous information systems, thereby improving the standardization, accuracy, and system stability of element data governance.

[0126] Optionally, Figure 2 This is a flowchart of an optional system element compliance verification method according to an embodiment of this application, such as... Figure 2 As shown, the processing system utilizes the natural language understanding and reasoning capabilities of the intelligent agent model, combined with a similar standard lexicon, to intelligently perform compliance checks on ordinary elements (i.e., non-enumerated primitive elements) and enumerated elements (i.e., enumerated primitive elements) in the system, obtaining the check results. This addresses the technical problems of semantic inconsistency and disordered enumeration value order in traditional systems. The specific compliance check process is set as follows:

[0127] (1) Selection of agent model and prompt word engineering: Select an agent model with high semantic understanding ability (such as Qwen2.5-256B-Instruct, Llama3-128B-Instruct), and design targeted preset prompt word templates in combination with enterprise business scenarios. The preset prompt word templates should include at least the system role (compliance verification expert), verification rules (matching similar standard words, enumeration value compliance requirements), input data (system element information, similar standard word library), and the format of model output data (compliance results, non-compliance reasons, enumeration value verification results) to ensure that the agent model can accurately understand the verification requirements and output standardized verification results.

[0128] (2) Compliance verification of ordinary elements: The intelligent agent model combines the similar standard word library to perform semantic comparison between the name and semantic description of the original element of non-enumerated type and similar standard words to determine whether the naming of the original element is compliant: if there is a similar standard word that completely matches the semantics of the original element, it is judged as compliant; if there is no completely matching similar standard word, but there is a similar standard word (similarity is higher than the threshold), it is judged as non-compliant, and the reason for non-compliance is recorded (such as "the element name is semantically consistent with the standard word 'customer number', but the expression is inconsistent"); if there is no similar standard word, it is judged as non-compliant and marked as "no standard word matched".

[0129] (3) Compliance verification of enumeration elements: For enumeration elements, the agent model not only needs to verify the compliance of the element naming, but also needs to perform double verification of the enumeration value: First, semantic verification, to determine whether the expression of the enumeration value is consistent with the semantics of the enumeration value of the standard word (such as comparing the semantics of "activate" with the standard enumeration value "enable"); Second, sequence verification, to determine whether the order of the enumeration value is consistent with the order of the enumeration value of the standard word. If the order is inconsistent, it is determined to be non-compliant and the order deviation information is recorded. In addition, the agent model will detect whether there are invalid values ​​or duplicate values ​​in the enumeration value. If they exist, they are marked as non-compliant.

[0130] (4) Output of verification results: The agent outputs standardized compliance verification results, which include at least element ID, element name, field type, compliance status (compliant / non-compliant), non-compliant reason, similar standard words, and enumeration value verification details, and stores the compliance verification results in the system database.

[0131] In one optional embodiment, after inputting the original vector into a standard vector knowledge base for semantic matching to obtain M similar standard words, the processing system first obtains the recall hit rate of each standard word in the standard dictionary, wherein the recall hit rate is used to characterize the frequency of each standard word being identified as a similar standard word of the original element. Then, the processing system stores the standard words with a recall hit rate greater than a preset hit rate, the standard vector of the standard word, and the standard word information into a preset cache area.

[0132] Optionally, the processing system also incorporates Redis (Remote Dictionary Server) caching technology. A Redis caching module is configured to cache frequently accessed data, shortening response time and improving system performance. This enhances system retrieval and verification efficiency, addressing the technical issue of low retrieval and verification efficiency in high-concurrency scenarios. The specific process for setting up the Redis caching module is as follows:

[0133] (1) Definition of cached content: The core content of the cache includes: similar standard words that are frequently recalled (such as "customer number", "order status" and other similar standard words with high recall frequency) and their standard vectors, high-frequency search results (such as the first standard word that is frequently recalled), high-frequency verification results (such as the compliance verification results of the original elements used in frequent compliance verification), and system configuration parameters (such as verification threshold and cache expiration time).

[0134] (2) Cache strategy design: The “LRU (Least Recently Used)” cache eviction strategy is adopted. When the cache capacity reaches the upper limit, the cache data that has been used the least in the most recent preset time period in the Redis cache module will be automatically evicted. At the same time, the cache expiration time (such as 24 hours) is set, and the expired cache data is updated regularly to avoid the cache data being inconsistent with the data in the actual dictionary. For scenarios such as standard dictionary updates and verification rule adjustments, the cache will be automatically invalidated and the latest data will be cached again.

[0135] (3) Cache call process: When the processing system performs semantic retrieval or compliance verification, it first queries the Redis cache module. If the corresponding data exists in the cache module (such as high-frequency recall similar standard words, high-frequency search results), it is directly obtained from the cache module without accessing the vector knowledge base or intelligent agent, thus shortening the response time. If the corresponding data does not exist in the cache, the normal retrieval and verification process is executed. After completion, the retrieval results are stored in the Redis cache for subsequent calls. Actual tests show that this caching strategy can improve the system's retrieval and verification efficiency and effectively adapt to high-concurrency scenarios.

[0136] In one alternative embodiment, Figure 3 This is a flowchart of an optional system element governance method according to an embodiment of this application, such as... Figure 3 As shown, firstly, the original information of the original elements in the information system is collected by the acquisition module. Then, the element information of the preprocessed original elements is input into a pre-built standard vector knowledge base. Through RAG semantic retrieval technology, similar standard words corresponding to the original elements are recalled. Subsequently, the intelligent agent module performs compliance verification on the original elements and the recalled similar standard words. At the same time, the data that is frequently accessed, used, and output during the compliance verification process is cached in the Redis cache module. If the original element fails the compliance verification, the intelligent agent model intelligently generates a rectification plan and collects user feedback data. After the current element is rectified, the next original element is collected and the compliance verification is carried out again.

[0137] Optionally, through the close coordination of the above steps, this application achieves full-process automation and intelligence in the digital governance of elements in heterogeneous enterprise information systems, solving the technical pain points of non-standard element naming, semantic inconsistency, and chaotic enumeration values ​​in traditional technologies. It provides a brand-new solution for enterprise knowledge management and data governance, and this application has the following beneficial effects:

[0138] (1) Significantly improved semantic matching accuracy, solving the problem of non-standard naming: Traditional technologies rely on keyword mapping or string matching, which cannot understand the semantics of elements, resulting in low matching accuracy and high mismatch rate. This invention integrates vector model and LLM (Large Language Model) technology. Through vector model, it realizes the semantic encoding of standard words and system elements. Combined with RAG technology, it recalls similar standard words. Utilizing the semantic understanding capability of LLM, it realizes semantic-level compliance verification, which can accurately identify elements that are "semantically the same but literally different" (such as "user age" and "customer age"), thus solving the problem of semantic ambiguity. At the same time, for enumeration-type elements, it realizes semantic verification and order checking of enumeration values, thereby solving the problems of disordered enumeration value order and inconsistent expression.

[0139] (2) Improve retrieval and verification efficiency and adapt to high concurrency management scenarios: Traditional technology uses relational databases to store standard dictionaries, which has low retrieval efficiency and cannot adapt to high concurrency scenarios. This application realizes efficient semantic retrieval of standard words through vector model knowledge base, and combines Redis caching technology to cache high-frequency data, shortening the retrieval and verification response time.

[0140] (3) Fully automated data collection reduces labor costs: The self-developed data collection method provided in this application supports four modes: interface collection, direct database connection collection, file import collection, and web crawling collection. It is compatible with enterprise information systems of different types and architectures and can achieve full and automated collection of raw data without manual intervention.

[0141] (4) The standard dictionary is easy to maintain and can be handled by non-technical personnel: The maintenance of traditional technical standard dictionaries is complicated, relies on technical personnel, and has high maintenance costs. This application provides a visual maintenance interface and batch import and export functions. Non-technical personnel (such as business personnel) can quickly add, modify, and delete standard words through the intuitive interface. It supports batch updates, simplifies the maintenance process, improves the maintenance efficiency of the standard dictionary, reduces maintenance costs, realizes the dynamic updating and expansion of the standard dictionary, and can quickly adapt to changes in enterprise business.

[0142] (5) The rectification recommendations are accurate and efficient, forming a complete governance closed loop: Traditional technologies can only identify non-compliant elements and cannot provide rectification suggestions, resulting in an imperfect governance closed loop. This application, based on LLM and RAG technologies, generates a Top 10 rectification recommendation list for non-compliant elements, clarifies the rectification priority and specific solutions, and allows users to quickly implement rectification. At the same time, a rectification feedback mechanism is established to track the rectification effect, forming a complete governance closed loop, thereby improving the standardization level of enterprise knowledge management and data governance.

[0143] (6) High versatility and adaptability to various enterprise information scenarios: This application adopts a modular design, supports multiple system types and multiple data formats, and can be widely used in digital governance scenarios of information systems of various enterprises (such as operation and maintenance management platforms, data middleware, etc.). There is no need for customized development for specific systems, and it has high versatility. At the same time, the system supports flexible configuration and can adjust parameters such as verification rules, caching strategies, and rectification priorities according to the business needs of enterprises, adapting to the personalized governance needs of different enterprises and having a wide range of applications.

[0144] Example 2

[0145] This application embodiment can also provide a system element processing device. It should be noted that the system element processing device of this application embodiment can be used to execute the system element processing method provided in this application embodiment. The system element processing device provided in this application embodiment will be described below.

[0146] According to an embodiment of this application, an apparatus for implementing the above-described system element processing method is also provided. Figure 4 This is a schematic diagram of a processing apparatus for an optional system element according to an embodiment of this application, such as... Figure 4 As shown, the device includes: a knowledge base construction unit 401, an original element encoding unit 402, a similar standard word determination unit 403, an original element verification unit 404, and an original element processing unit 405.

[0147] Optionally, the knowledge base construction unit 401 is used to construct a standard vector knowledge base based on the standard vector and standard word information corresponding to each standard word after semantically encoding the standard words in the standard dictionary to obtain standard vectors. The standard dictionary is a set of fields generated based on the enterprise's preset business specifications. The original element encoding unit 402 is used to semantically encode the original elements in the information system to obtain original vectors. The original elements are fields used to describe business data. The similar standard word determination unit 403 is used to input the original vectors into the standard vector knowledge base for semantic matching to obtain M similar standard words. Similar standard words, where M is a positive integer, and the vector similarity between the original vector and the standard vector corresponding to each similar standard word is greater than a preset similarity; Original element verification unit 404 is used to perform compliance verification on the original element based on M similar standard words to obtain the verification result, wherein the compliance verification includes at least one of the following: name semantic sub-verification, description text semantic sub-verification, enumeration value semantic sub-verification, and enumeration value order sub-verification; Original element processing unit 405 is used to generate a rectification plan for the original element based on the verification result when the original element is non-compliant, and to process the original element based on the rectification plan.

[0148] In one optional embodiment, the system element processing device further includes: a standard word information acquisition unit, a semantic enhancement unit, and a standard word encoding unit.

[0149] Optionally, the standard word information acquisition unit is used to acquire the standard word information of each standard word in the standard dictionary, wherein the standard word information includes at least one of the following: standard word name, description text, field type, enumeration value list, business scenario, and related words, and there is a business relationship between each standard word and its corresponding related words; the semantic enhancement unit is used to generate semantic enhancement text for each standard word based on the standard word information of each standard word; the standard word encoding unit is used to perform semantic encoding on the semantic enhancement text of each standard word through a vector model to obtain the standard vector of each standard word, wherein the vector model is a large model trained based on the historical encoding records of the business fields of the information system.

[0150] In one optional embodiment, the system element processing device further includes: a raw information acquisition unit and a raw information preprocessing unit.

[0151] Optionally, the raw information acquisition unit is used to acquire the raw information of each raw element in the information system through the acquisition module before semantic encoding of the raw elements in the information system. The acquisition module is used to simulate browser behavior. The raw information preprocessing unit is used to perform data cleaning, standardization and data filtering on the raw information of each raw element to obtain the element information of each raw element. The element information of each raw element is used to generate semantically enhanced text for each raw element. The element information includes at least the element name, description text, field type, enumeration value list and business scenario.

[0152] In one optional embodiment, the similarity standard word determination unit 403 includes: a similarity detection subunit, an initial recall subunit, and a standard word optimization subunit.

[0153] Optionally, the similarity detection subunit is used to detect the vector similarity between the original vector and the standard vector corresponding to each standard word in the standard vector knowledge base; the initial recall subunit is used to take the standard words with vector similarity greater than a preset similarity as the first standard words to obtain L first standard words corresponding to the original vector, where L is a positive integer greater than or equal to M; the standard word optimization subunit is used to optimize the L first standard words to obtain M similar standard words, wherein the optimization operation is at least used to filter duplicate standard words and standard words that have business semantic conflicts with the element information of the original element.

[0154] In one optional embodiment, the original element verification unit 404 includes: a verification rule determination subunit, a prompt word generation subunit, a prompt word input subunit, and a model verification subunit.

[0155] Optionally, the verification rule determination subunit is used to determine the verification rules for matching the original element based on M similar standard words. The verification rules include at least one of the following: semantic verification rules and sequential verification rules; the prompt word generation subunit is used to generate target prompt words based on the preset prompt word template corresponding to the verification rules, the standard word information of the M similar standard words, and the element information of the original element; the prompt word input subunit is used to input the target prompt word into the intelligent agent model, wherein the intelligent agent model is a large model trained based on the historical verification records of the business fields of the information system; and the model verification subunit is used to perform compliance verification on the original element based on the target prompt word through the intelligent agent model.

[0156] In one optional embodiment, the model verification subunit includes: a semantic sub-verification module, a first determination module, a text semantic sub-verification module, and a second determination module.

[0157] Optionally, the semantic sub-verification module is used to perform name semantic sub-verification based on the element name of the original element in the target prompt word and the standard word name of each similar standard word through an intelligent agent model; the first determination module is used to determine that the original element passes the name semantic sub-verification when there is a similar standard word with the same standard word name as the element name among the M similar standard words; the text semantic sub-verification module is used to perform description text semantic sub-verification based on the description text of the original element in the target prompt word and the description text of each similar standard word through an intelligent agent model when there is no similar standard word with the same standard word name as the element name among the M similar standard words; the second determination module is used to determine that the original element passes the description text semantic sub-verification when there is a similar standard word with a semantic similarity greater than a preset similarity among the M similar standard words, wherein the semantic similarity is the similarity between the description text of the original element and the description text of each similar standard word.

[0158] In one optional embodiment, the model verification subunit further includes: an enumeration value semantic sub-verification module, an enumeration value order sub-verification module, and a third determination module.

[0159] Optionally, the enumeration value semantic sub-validation module is used to perform enumeration value semantic sub-validation based on the enumeration value list of the original element in the target prompt word and the enumeration value list of each similar standard word after performing name semantic sub-validation and / or description text semantic sub-validation when the field type of the original element is an enumeration type. The enumeration value order sub-validation module is used to determine that the original element passes the enumeration value semantic sub-validation when a first enumeration value list exists in the M enumeration value lists corresponding to the M similar standard words, wherein the enumeration values ​​in the first enumeration value list are the same as the enumeration values ​​in the enumeration value list of the original element. Furthermore, the enumeration value order sub-validation is performed by the intelligent agent model based on the enumeration value list of the original element in the target prompt word and the enumeration value list of each similar standard word. The third determination module is used to determine that the original element passes the enumeration value semantic sub-validation when a second enumeration value list exists in the M enumeration value lists corresponding to the M similar standard words, wherein the order of the enumeration values ​​in the second enumeration value list is consistent with the order of the enumeration values ​​in the enumeration value list of the original element.

[0160] In an optional embodiment, the system element processing apparatus further includes a recall hit rate acquisition unit and a caching unit.

[0161] Optionally, the recall hit rate acquisition unit is used to obtain the recall hit rate of each standard word in the standard dictionary after inputting the original vector into the standard vector knowledge base for semantic matching and obtaining M similar standard words. The recall hit rate is used to characterize the frequency of each standard word being identified as a similar standard word of the original element. The caching unit is used to store the standard words with a recall hit rate greater than the preset hit rate, the standard vector of the standard word, and the standard word information into the preset cache area.

[0162] It should be noted that the knowledge base construction unit 401, the original element encoding unit 402, the similar standard word determination unit 403, the original element verification unit 404, and the original element processing unit 405 mentioned above correspond to steps S101 to S105 in the method embodiment. The instances and application scenarios implemented by the above units and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment.

[0163] Example 3

[0164] Embodiments of this application can also provide an electronic device. Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application, such as... Figure 5 As shown, the electronic device includes: one or more ( Figure 5 (Only one is shown) processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0165] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and devices in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, to realize the above-mentioned system element processing method.

[0166] The memory may include high-speed random access memory (RAM), and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, which can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0167] The processor can invoke information and applications stored in the memory via a transmission device to execute the following steps: After semantically encoding standard words in the standard dictionary to obtain standard vectors, a standard vector knowledge base is constructed based on the standard vectors and standard word information corresponding to each standard word, wherein the standard dictionary is a set of fields generated based on the enterprise's preset business specifications; semantically encoding the original elements in the information system to obtain original vectors, wherein the original elements are fields used to describe business data; inputting the original vectors into the standard vector knowledge base for semantic matching to obtain M similar standard words, wherein M is a positive integer, and the vector similarity between the original vector and the standard vector corresponding to each similar standard word is greater than the preset similarity; performing compliance verification on the original elements based on the M similar standard words to obtain verification results, wherein the compliance verification includes at least one of the following: name semantic sub-verification, description text semantic sub-verification, enumeration value semantic sub-verification, and enumeration value order sub-verification; in the case that the original element is non-compliant, a rectification plan for the original element is generated based on the verification results, and the original element is processed based on the rectification plan.

[0168] This application provides a solution for processing system elements. It employs a combination of vector semantic encoding and compliance verification. By constructing a standard vector knowledge base and combining it with semantic similarity retrieval technology to quickly recall similar standard words, and then using multi-dimensional compliance verification to intelligently detect the compliance of original elements in the information system, it achieves the goal of accurately identifying non-standard elements that are semantically identical to standard words but inconsistent in expression. This application does not rely on traditional manual verification methods or hard-coding rules; instead, it uses semantic vectors to capture the business meaning of elements and combines enumerated value order sub-verification to achieve automated and high-precision identification of non-standard named elements. This achieves a technical effect of leaping from element literal-level matching to semantic-level governance, improving the efficiency of rectifying non-standard elements, and thus solving the technical problem of non-standard element settings in enterprise information systems.

[0169] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, PDAs, mobile internet devices, PADs, and other terminal devices. Figure 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0170] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0171] Example 4

[0172] Embodiments of this application may also provide a storage medium.

[0173] Optionally, in this embodiment of the application, the storage medium can be used to store the program code executed by the processing method of the system elements provided in the above method embodiment.

[0174] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0175] This application also provides a computer program product, which, when executed on a data processing device, is adapted to perform processing method steps of system elements.

[0176] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0177] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0178] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0179] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0180] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0181] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part 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 described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0182] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for processing system elements, characterized in that, include: After semantically encoding the standard words in the standard dictionary to obtain standard vectors, a standard vector knowledge base is constructed based on the standard vector and standard word information corresponding to each standard word. The standard dictionary is a set of fields generated based on the enterprise's preset business specifications. Semantic encoding is performed on the original elements in the information system to obtain the original vector, wherein the original elements are fields used to describe business data; The original vector is input into the standard vector knowledge base for semantic matching to obtain M similar standard words, where M is a positive integer, and the vector similarity between the original vector and the standard vector corresponding to each similar standard word is greater than a preset similarity. Based on the M similar standard words, the original element is subjected to compliance verification to obtain the verification result. The compliance verification includes at least one of the following: name semantic sub-verification, description text semantic sub-verification, enumeration value semantic sub-verification, and enumeration value order sub-verification. If the original element is non-compliant, a rectification plan for the original element is generated based on the verification result, and the original element is processed based on the rectification plan.

2. The method for processing system elements according to claim 1, characterized in that, Semantic encoding is performed on standard words in the standard dictionary to obtain standard vectors, including: Obtain the standard word information for each standard word in the standard dictionary, wherein the standard word information includes at least one of the following: standard word name, description text, field type, enumeration value list, business scenario, and related words, and there is a business relationship between each standard word and its corresponding related words; Based on the standard word information of each standard word, semantically enhanced text for each standard word is generated; The semantically enhanced text of each standard word is semantically encoded using a vector model to obtain a standard vector for each standard word. The vector model is a large model trained based on historical encoding records of business fields in the information system.

3. The method for processing system elements according to claim 1, characterized in that, Before semantically encoding the original elements in the information system, the processing method for the system elements further includes: The original information of each original element in the information system is collected by the acquisition module, wherein the acquisition module is used to simulate browser behavior; The original information of each original element is cleaned, standardized, and filtered to obtain the element information of each original element. The element information of each original element is used to generate the semantically enhanced text of each original element. The element information includes at least the element name, description text, field type, enumeration value list, and business scenario.

4. The method for processing system elements according to claim 1, characterized in that, The original vector is input into the standard vector knowledge base for semantic matching to obtain M similar standard words, including: Detect the vector similarity between the original vector and the standard vector corresponding to each standard word in the standard vector knowledge base; The standard words whose vector similarity is greater than a preset similarity are used as the first standard words to obtain L first standard words corresponding to the original vector, where L is a positive integer greater than or equal to M; The L first standard words are optimized to obtain M similar standard words. The optimization operation is used to filter out duplicate standard words and standard words that have business semantic conflicts with the element information of the original elements.

5. The method for processing system elements according to claim 1, characterized in that, The original element is subjected to compliance verification based on the M similar standard words, including: The verification rules for matching the original element are determined based on the M similar standard words, and the verification rules include at least one of the following: semantic verification rules and sequence verification rules; Target prompt words are generated based on the preset prompt word template corresponding to the verification rule, the standard word information of the M similar standard words, and the element information of the original element; The target prompt word is input into the intelligent agent model, wherein the intelligent agent model is a large model trained based on the historical verification records of the business fields of the information system; The intelligent agent model performs compliance verification on the original element based on the target prompt word.

6. The method for processing system elements according to claim 5, characterized in that, The intelligent agent model performs compliance verification on the original element based on the target prompt word, including: The intelligent agent model performs the name semantic sub-verification based on the element name of the original element in the target prompt word and the standard word name of each similar standard word; If among the M similar standard words there exists a similar standard word whose name is the same as the element name, the original element is determined to pass the name semantic sub-verification; If no similar standard word with the same name as the element name exists among the M similar standard words, the agent model performs a semantic sub-verification of the description text based on the description text of the original element in the target prompt word and the description text of each similar standard word. If among the M similar standard words there is a similar standard word with a semantic similarity greater than a preset similarity, the original element is determined to pass the semantic sub-verification of the description text, wherein the semantic similarity is the similarity between the description text of the original element and the description text of each similar standard word.

7. The method for processing system elements according to claim 5, characterized in that, The intelligent agent model performs compliance verification on the original element based on the target prompt word, including: When the field type of the original element is an enumeration type, after performing the name semantic sub-validation and / or the description text semantic sub-validation, the agent model performs the enumeration value semantic sub-validation based on the enumeration value list of the original element in the target prompt word and the enumeration value list of each similar standard word. If a first enumeration value list exists in the M enumeration value lists corresponding to the M similar standard words, the original element is determined to pass the semantic sub-verification of the enumeration value, wherein the enumeration value in the first enumeration value list is the same as the enumeration value in the enumeration value list of the original element. Furthermore, the intelligent agent model performs a sub-verification of the enumeration value order based on the enumeration value list of the original elements in the target prompt word and the enumeration value list of each similar standard word; If a second enumeration value list exists in the M enumeration value list corresponding to the M similar standard words, the original element is determined to pass the semantic sub-verification of the enumeration value, wherein the order of the enumeration values ​​corresponding to the second enumeration value list is consistent with that of the enumeration value list of the original element.

8. The method for processing system elements according to claim 1, characterized in that, After inputting the original vector into the standard vector knowledge base for semantic matching to obtain M similar standard words, the processing method for the system elements further includes: Obtain the recall rate of each standard word in the standard dictionary, wherein the recall rate is used to characterize the frequency with which each standard word is identified as a similar standard word of the original element; The standard words with a recall hit rate greater than a preset hit rate, the standard vector of the standard words, and the standard word information are stored in a preset cache area.

9. A processing apparatus for system elements, characterized in that, include: The knowledge base construction unit is used to construct a standard vector knowledge base based on the standard vector and standard word information corresponding to each standard word after semantically encoding the standard words in the standard dictionary to obtain standard vectors. The standard dictionary is a set of fields generated based on the enterprise's preset business specifications. The original element encoding unit is used to perform semantic encoding on the original elements in the information system to obtain the original vector, wherein the original elements are fields used to describe business data; The similar standard word determination unit is used to input the original vector into the standard vector knowledge base for semantic matching to obtain M similar standard words, where M is a positive integer, and the vector similarity between the original vector and the standard vector corresponding to each similar standard word is greater than a preset similarity. The original element verification unit is used to perform compliance verification on the original element based on the M similar standard words to obtain the verification result. The compliance verification includes at least one of the following: name semantic sub-verification, description text semantic sub-verification, enumeration value semantic sub-verification, and enumeration value order sub-verification. The original element processing unit is used to generate a rectification plan for the original element based on the verification result when the original element is non-compliant, and to process the original element based on the rectification plan.

10. A computer program product, characterized in that, The computer program product includes a computer program, wherein, when the computer program is executed, it controls the computer program product to perform the processing method of any one of the system elements according to claims 1 to 8.

11. An electronic device, characterized in that, It includes one or more processors and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the processing method of the system element according to any one of claims 1 to 8.

Images