Data governance method and device based on large language model, equipment, storage medium and product

Abstract

The invention discloses a data governance method and device based on a large language model, equipment, a storage medium and a product, and relates to the technical field of data governance, and the method comprises the following steps: obtaining natural language information of a large language model input by a target user; identifying key elements according to the natural language information, and performing context supplementation based on the key elements to obtain cue words; generating a data governance task flow according to the cue word, wherein the data governance task flow comprises at least one data governance task generated based on the cue word; according to the method, the high-quality prompt instruction is obtained by identifying the key elements, the stability and reliability of the data governance result are improved, and the data governance tool parameters are generated according to the data governance task and data governance is carried out according to the data governance tool parameters. And according to the feedback of the governance result, the governance tool used by the governance task and the governance tool parameters are self-optimized, so that the data governance efficiency and the decision-making quality are improved, and the problem of result divergence when the complex data governance task is processed is avoided.

Description

Technical Field

[0001] The present application relates to the field of data governance technology, and in particular to data governance methods, devices, equipment, storage media and products based on large language models. Background Art

[0002] Data governance covers a series of management and operational tools aimed at effectively managing an organization's data resources. Tools at the management level include: data architecture management, metadata management, data standards management, master data management, data quality management, data asset management, data security management, data life cycle management, etc.; tools at the operational level include: data storage, data collection, data processing, data sharing and exchange, AI computing support, and data analysis applications.

[0003] Data governance starts with the influx of data sources, cleansing and processing the data, and then provides continuous governance services in data storage, data calculation, data service applications and other links. Its purpose is to integrate scattered, non-standard, and non-uniform data, output standardized and unified data sources, and provide accurate data basis for decision-making. The components of data governance mainly include data source, data integration, data catalog, data architecture, data development, data quality, data security, and data services. In traditional processes, data governance personnel are required to have business knowledge, data management capabilities, and data development technology at the same time. The threshold is high, the process is complex, and the degree of automation is low. This leads to poor performance of large language models when feedback divergence problems. Summary of the invention

[0004] The main purpose of this application is to provide a data governance method, device, equipment, storage medium and product based on a large language model, aiming to solve the technical problem in the prior art that the large language model performs poorly when the feedback diverges.

[0005] To achieve the above objectives, the present application proposes a data governance method based on a large language model, and the data governance method based on a large language model includes: When the natural language information is not missing, identifying key elements according to the natural language information to obtain the conversation topic, user intention and entity; When the conversation topic, the user intention and the entity are verified to be qualified, context supplement is performed according to the conversation topic, the user intention and the entity to obtain background information and detail information; A prompt word is obtained according to the background information and the detail information.

[0006] In one embodiment, the step of generating a data governance task flow according to the prompt word includes: Obtaining data governance complexity based on the prompt word; When the data governance complexity is greater than or equal to a complexity threshold, generating multiple governance tasks based on the prompt word; A data governance task flow is generated according to the plurality of governance tasks.

[0007] In one embodiment, the step of generating a governance tool and a data governance tool parameter according to the data governance task, and performing data governance according to the governance tool and the data governance tool parameter comprises: Obtaining an execution action according to the data governance task, and obtaining a governance tool according to the execution action; A data governance tool parameter is obtained according to the governance tool, and data governance is performed according to the data governance tool parameter.

[0008] In one embodiment, after the steps of generating a governance tool and a data governance tool parameter according to the data governance task and performing data governance according to the governance tool and the data governance tool parameter, the step further includes: Obtain governance results; Performing a governance quality assessment based on the governance results to obtain a governance status; When the governance status is that the data quality has deteriorated, a negative evaluation is generated, and the parameters of the data governance tool are updated based on the negative evaluation; When the governance status is that data quality has improved, a positive evaluation is generated and data governance is completed.

[0009] In one embodiment, when the governance status is data quality degradation, the step of generating a negative evaluation and updating the data governance tool parameters based on the negative evaluation includes: When the governance status is that the data quality has deteriorated, a historical status is obtained according to the negative evaluation, and a historical action is obtained according to the data governance tool parameter and the governance tool; Optimizing a parameter vector of the large language model according to the historical state and the historical action to obtain a new model parameter vector; The data governance tool parameters are updated according to the new model parameter vector.

[0010] In addition, to achieve the above purpose, the present application also proposes a data management device based on a large language model, and the data management device based on a large language model includes: An acquisition module is used to acquire natural language information input by a target user into a large language model; A prompt word generation module, used for identifying key elements according to the natural language information, and performing context supplement based on the key elements to obtain prompt words; A data governance module, configured to generate a data governance task flow according to the prompt word, wherein the data governance task flow includes at least one data governance task generated based on the prompt word; The data governance module is further used to generate a governance tool and data governance tool parameters according to the data governance task, and perform data governance according to the governance tool and the data governance tool parameters.

[0011] In addition, to achieve the above-mentioned objectives, the present application also proposes a data governance device based on a large language model, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the data governance method based on a large language model as described above.

[0012] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the data governance method based on the large language model as described above are implemented.

[0013] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the data governance method based on the large language model as described above.

[0014] One or more technical solutions proposed in this application have at least the following technical effects: By identifying key elements to obtain high-quality prompts and instructions, the stability and reliability of data governance results can be improved, and the governance tools and governance tool parameters used in governance tasks can be self-optimized based on the feedback from the governance results, thereby reducing the need for human intervention, reducing the risks brought by human factors, improving data governance efficiency and decision-making quality, and avoiding the problem of divergent results when dealing with complex data governance tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0017] Figure 1A flowchart of the first embodiment of the data governance method based on a large language model of this application is provided; Figure 2 A schematic diagram of the application architecture of the data management method based on a large language model provided in Embodiment 1 of the data management method based on a large language model of the present application; Figure 3 A schematic diagram of the reasoning and action framework of an intelligent agent provided in Example 1 of the data governance method based on a large language model of this application; Figure 4 A schematic diagram of an intelligent agent framework including a guidance module and a supervision module provided in Example 1 of the data governance method based on a large language model of the present application; Figure 5 A schematic diagram of an automated data governance process provided in Example 1 of the data governance method based on a large language model of this application; Figure 6 A flowchart of the second embodiment of the data governance method based on a large language model of this application is provided; Figure 7 A schematic diagram of a data set structure 1 provided for Example 2 of the data governance method based on a large language model of this application; Figure 8 A schematic diagram of a data set structure 2 provided for Example 2 of a data governance method based on a large language model of this application; Fig. 9 A schematic diagram of the architecture of a data management device for a language model provided in Example 2 of the data management method based on a large language model of the present application; Fig.10 A schematic diagram of a calling flow of a data management device based on a large language model provided in Embodiment 2 of the data management method based on a large language model of the present application; Fig.11 This is a schematic diagram of the module structure of a data management device based on a large language model according to an embodiment of the present application; Fig.12 This is a schematic diagram of the device structure of the hardware operating environment involved in the data governance method based on a large language model in an embodiment of the present application.

[0018] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0019] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0020] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0021] The main solution of the embodiment of the present application is: obtaining natural language information of the target user input into the large language model; identifying key elements based on the natural language information, supplementing the context based on the key elements, and obtaining prompt words; generating a data governance task flow based on the prompt words, the data governance task flow includes generating at least one data governance task based on the prompt words; generating data governance tool parameters based on the data governance tasks, and performing data governance based on the data governance tool parameters.

[0022] Since the existing data governance starts with the import of data sources, the data is cleaned and processed, and then continuous governance services are provided in the data storage, data calculation, data service application and other links. Its purpose is to integrate scattered, non-standard, and non-uniform data, output standardized and unified data sources, and provide accurate data basis for decision-making. The components of data governance mainly include data source, data integration, data catalog, data architecture, data development, data quality, data security, and data services. In the traditional process, data governance personnel are required to have business knowledge, data management capabilities, and data development technology at the same time. The threshold is high, the process is complex, and the degree of automation is low. This leads to poor performance of large language models when feedback divergence problems.

[0023] The present application provides a solution, a data governance method, apparatus, equipment, storage medium and product based on a large language model, which relates to the field of data governance technology, including: obtaining natural language information of a target user input into a large language model; identifying key elements based on the natural language information, supplementing the context based on the key elements, and obtaining prompt words; generating a data governance task flow based on the prompt words, the data governance task flow including generating at least one data governance task based on the prompt words; generating data governance tool parameters based on the data governance tasks, and performing data governance based on the data governance tool parameters. The method obtains high-quality prompt instructions by identifying key elements, improves the stability and reliability of data governance results, and self-optimizes the governance tools and governance tool parameters used in the governance tasks based on the feedback of the governance results, thereby improving data governance efficiency and decision-making quality, and avoiding the problem of result divergence when processing complex data governance tasks.

[0024] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a data governance method based on a large language model, etc. The following takes the data governance method based on a large language model as an example to illustrate this embodiment and the following embodiments.

[0025] Based on this, the embodiment of the present application provides a data governance method based on a large language model, referring to Figure 1 , Figure 1This is a flow chart of the first embodiment of the data governance method based on a large language model of the present application.

[0026] In this embodiment, the data governance method based on the large language model includes steps S10 to S40: Step S10: Obtain natural language information of the target user input into the large language model.

[0027] It is understandable that natural language information can be the content input by the user into the big language model, and the big language model can be a natural language processing (NLP) system based on deep learning, which learns the structure, pattern and semantics of the language by analyzing a large amount of text data, and can be used to realize functions such as text generation and dialogue systems.

[0028] Understandably, users use natural language to converse with the large language model.

[0029] It should be noted that the application architecture of the data governance method based on the large language model described in the present invention can refer to Figure 2 , Figure 2 The invention understands the user's natural language input, and the intelligent agent decomposes the data governance requirements into several data governance tasks, generates an automated data governance pipeline, finds available tools based on the data tasks, generates tool parameters and executes them, observes the results and uses them as part of the subsequent process input. Through the guidance and supervision modules, the actions and data quality of the intelligent agent are evaluated, and the evaluation results are used as part of the subsequent process input to achieve automated data governance.

[0030] In short, the data governance method based on the large language model of the present invention can be implemented by including a guidance module, a supervision module and an intelligent agent module.

[0031] Among them, the guidance module extracts entities from the user's natural language input to supplement the data governance context and the details of the processed data. The intelligent agent generates an automated data governance pipeline to handle data tasks on the data governance pipeline. The guidance module guides the intelligent agent to generate structured outputs so that the intelligent agent can generate correct tool parameters. The simple text entered by the user is automatically converted into high-quality prompt words, and the user no longer needs to conduct multiple rounds of dialogue, which improves efficiency.

[0032] The supervision module monitors the output of the intelligent agent. When the output does not conform to the structure, it rejects the current result and requires the model to re-output, thus avoiding the propagation of errors.

[0033] The design of intelligent agents mainly focuses on solving the problem of fact hallucination in large models, ignoring the importance of data quality in data governance. Through the supervision module, data quality is supervised and evaluated in real time in the data governance pipeline. After each data task is completed, the data quality is supervised, the actions and results of the intelligent agent are evaluated, and the action space of the agent is limited. When the data quality is detected to be degraded, the supervision module sends a feedback signal, requiring the intelligent agent to roll back or adjust the task to ensure that the data quality is always at a high level.

[0034] It should be further explained that the conceptual framework of intelligent agents includes three main components: brain, perception and action, which are similar to the human sensory system, brain and limbs. The perception module perceives changes in the external environment and converts information into understandable representations. The brain module acts as a control center to process information, including thinking, decision-making and storage operations. The action module performs operations and affects the environment through tools.

[0035] Further, the reasoning and action framework of intelligent agents can refer to Figure 3 , Figure 3 The role of the reasoning and action framework is to coordinate the LLM model and external information acquisition, and interact with other functions; it mainly consists of three parts: Thought, Action, and Observation. Thought: Generated by the LLM model, it is the basis for LLM to generate behavior. The biggest advantage of the LLM model is that it can select the API to be used and generate the API parameters that need to be filled in according to the judgment of Thought. Action: LLM determines the specific behavior that needs to be performed this time. Action generally consists of two parts: behavior and object. In programming terms, it is the API name and the corresponding input parameters. Observation: The LLM framework obtains external input. Synchronize external feedback information to the LLM model to assist the LLM model in further analysis or decision-making.

[0036] It should be noted that the reasoning and action process of the intelligent agent can be 1. Input the goal: the starting point of the task. It can be manual input by the user or rely on triggers. 2. The large language model begins to analyze the steps required for the problem (Thought). 3. Perform actions according to the steps (Action). 4. Based on the observed information (Observation). 5. Cycle 2-4 until it is determined that the task goal is achieved. 6. Finish: The task is finally executed successfully and the final result is returned.

[0037] It should be noted that the logic and process of processing data governance tasks in the data governance task flow in the present invention can be fully understood here.

[0038] It should be further explained that the performance of intelligent agents depends to a large extent on the capabilities of large language models. Although high-performance large language models have higher accuracy, these models are often closed source and large in scale, and consume a lot of resources during inference. Compared with large-scale models, when using smaller local large language models to build agents, there will be problems with divergent model output results, leading to wrong decisions. In addition, because the agent relies on natural language to interact with external tools, there is a risk of incorrect tool selection or introduction of conflicting information. This proposal proposes a guidance and supervision module to evaluate the actions of the agent. The guidance and supervision module is composed of LLM. Unlike the agent, the LLM has vertical field and data governance expertise, and can provide more accurate supervision and feedback. The intelligent agent framework proposed in the present invention, which includes a guidance module and a supervision module, can be referred to. Figure 4 , Figure 4 It includes supervision module, guidance module, thinking module, action module and observation module.

[0039] Step S20: identifying key elements according to the natural language information, and performing context supplementation based on the key elements to obtain prompt words.

[0040] It should be noted that identifying key elements based on the natural language information and supplementing the context based on the key elements to obtain prompt words can be a guidance module that understands natural language intent based on training, generates guidance, modifies and expands user input, supplements context and details, and generates high-quality prompt words.

[0041] It is understandable that the key elements may be the key information in the natural language information input by the user, in simple terms, they may be the nouns, verbs and the like in a sentence.

[0042] It should be understood that context supplementation can be expanded based on the keywords input by the user, and the expansion direction can be expansion towards the field background or expansion towards synonyms, and prompt words are extracted based on the expanded content.

[0043] In a feasible implementation, step S20 may include steps A21 to A23: Step A21, when the natural language information is not missing, identify key elements based on the natural language information to obtain the conversation topic, user intention and entity.

[0044] It is understandable that a large language model can be pre-trained to recognize key elements in natural language, which may be conversation topics, user intent, and entities (entity objects).

[0045] It should be noted that the key elements include the topic of the conversation, the user's intention, and related entities. By accurately identifying these elements, we can better understand the user's real needs based on the key elements, and thus provide more accurate and effective responses. Specifically, the topic of the conversation refers to the core content or topic around which the entire conversation revolves, the user's intention refers to the specific goal or need that the user hopes to achieve through the conversation, and the entity refers to the specific thing, person, or concept mentioned in the conversation. Through a comprehensive analysis of these elements, we can more comprehensively grasp the context of the conversation and the user's real intention, and thus make a more appropriate response.

[0046] It should be emphasized that natural language information can be judged to be missing or not through clarity, completeness, contextual coherence, intent clarity, and entity recognition.

[0047] Understandably, clarity checks whether the information provided by the user is clear and specific enough. Completeness assesses whether the information provided by the user covers all the necessary elements required to complete the request or task. Contextual coherence checks whether the user's input is complete in the context of the current conversation. Intent clarity understands the user's specific needs or goals. Entity recognition determines whether necessary entities (such as names of people, places, and organizations) are mentioned by analyzing nouns, proper nouns, etc. in the text.

[0048] Step A22, when the conversation topic, the user intention and the entity are qualified in fact verification, context supplement is performed according to the conversation topic, the user intention and the entity to obtain background information and detail information.

[0049] It should be noted that the fact checking of the conversation topic, the user intention and the entity may be to perform logical judgment and grammar, spelling and other checks on the conversation topic, user intention and entity in the natural language information input by the user to see if they are qualified.

[0050] Specifically, we first need to clarify the topic of the conversation, which helps us grasp the core content of the conversation. Secondly, we need to carefully analyze the user's intentions to understand what purpose or result the user hopes to achieve through the conversation. Finally, we need to pay attention to the entities mentioned in the conversation, which may be names of people, places, events, or other specific things, which provide specific reference points for the conversation.

[0051] In the specific implementation, identify key elements: identify key elements in user input, including topics, intent, and main entities; if it is detected that the user input information is seriously missing and accurate data governance cannot be performed, guide the user to provide detailed information and prompt the user on how to improve the input; perform fact checking: check whether there is a logical conflict in the user input, and perform logic and content correction. Correct user errors: grammar and spelling correction.

[0052] Step A23, obtaining a prompt word according to the background information and the detail information.

[0053] It should be noted that when supplementing the context, if it involves a specific field, add background information of the field. For example, when mentioning "drug" in the medical field, you can supplement the symptoms, treatment plans, etc.

[0054] It should be further explained that if additional details are needed: add specific details for key elements. For example, if the user mentions "project", you can add the project name, associated database, related upstream and downstream data tasks, etc.

[0055] It should be understood that the background and details can be obtained based on the subject content of the conversation, combined with the intention expressed by the user, and taking into account the relevant entities mentioned in the conversation. The background information and detail information can further supplement and enrich the contextual information, thereby better understanding the conversation and providing more accurate and comprehensive information support.

[0056] In a specific implementation, the original text input by the user is "customer satisfaction factor". S1. Define the original text: T="customer satisfaction factor".

[0057] S2. Identify key entities: Extract the key entity set K in the original text, which can be expressed as K={"customer", "satisfaction", "factor"}.

[0058] S3. Supplementary context: The original text is extended by introducing relevant contextual information C. The context includes domain-specific terms, definitions, background information, etc. For example, in the customer satisfaction scenario, relevant words such as "service quality" and "customer feedback" can be introduced. The text Tc after supplementing the context is expressed as: Tc=T+C, where C={"service quality", "customer feedback", "satisfaction index"}.

[0059] S4. Add specific details: Further improve the quality of the prompt word by introducing specific details information D. The details can include common satisfaction measurement methods, influencing factors, related indicators, etc., then D = {"NPS (Net Promoter Score)", "CSAT (Customer Satisfaction Score)", "Customer Experience"}.

[0060] S5. Data retrieval enhancement: Combined with S2-S4, retrieve data in the data warehouse. The following are some possible fields: table name: customer_satisfaction; fields: device_id (device ID), satisfaction_score (satisfaction score), interaction_count (number of interactions), service_quality (service quality), response_time (response time); table name: customer_feedback; fields: customer_id (customer ID), feedback_text (feedback text), sentiment_score (sentiment score), nps_score (net recommendation score), purchase_history (purchase history).

[0061] The detailed information is updated to: D={"Smart home devices","Smart home services","User interaction","Device compatibility","Data privacy","NPS (Net Promoter Score)","CSAT (Customer Satisfaction Score)","customer_satisfaction table in Hive database","customer_feedback table in MySQL database"}.

[0062] S5. Clarify the task goal: Combined with the information of the logged-in user, for example, if the current user belongs to the digital home department, the task goal is set as G="Analyze the key factors of customer satisfaction in the digital home field and generate strategic recommendations to improve customer satisfaction, especially for smart home products and services.".

[0063] S6. Construct a complete prompt word: The final high-quality prompt word can be expressed as P="Please analyze "+T+", pay special attention to the following key factors: "+K_ext+", combine "+D+", and generate a "+G+". ".

[0064] The final complete prompt is "P="Please analyze the factors of customer satisfaction in the digital home field, with special attention to the following key factors: customer satisfaction, digital home, smart home, key factors, satisfaction indicators, customer feedback, service quality, and user experience. Combine smart home devices, smart home services, user interactions, device compatibility, data privacy, and data extracted from the customer_satisfaction table (fields: device_id, satisfaction_score, interaction_count, service_quality, response_time) and customer_feedback table (fields: customer_id, feedback_text, sentiment_score, nps_score, purchase_history) in the Hive database to generate a strategic recommendation for improving customer satisfaction, focusing on improving the user experience of smart home products and services. "” In this implementation, richer background information and detailed information are obtained through context supplementation, which helps to better understand the overall picture of the conversation and enables the large language model to provide users with more targeted answers and suggestions.

[0065] The above are only feasible implementations of step S20 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S20.

[0066] Step S30: Generate a data governance task flow according to the prompt word, wherein the data governance task flow includes at least one data governance task generated based on the prompt word.

[0067] It is understandable that the prompt words generated for the natural language information input by the user may include multiple prompt words after contextual supplementation, and different prompt words may correspond to data governance tasks in different directions.

[0068] It should be understood that the data governance task flow may be a task flow consisting of multiple governance tasks generated according to prompt words.

[0069] In a feasible implementation, step S30 may include steps A31 to A33: Step A31, obtaining data management complexity based on the prompt word.

[0070] It is understandable that data governance complexity can be divided into more than 10 levels, which can be limited according to actual needs.

[0071] It should be noted that the data management complexity obtained based on the prompt words can be determined according to the type of the number of prompt words or the direction involved by the prompt words.

[0072] In a specific implementation, the prompt words may include "Please analyze the customer satisfaction factors in the digital home field, and pay special attention to the following key factors: customer satisfaction, digital home, smart home, key factors, satisfaction indicators, customer feedback, service quality, and user experience. The number of prompt words involved is n, and the category of prompt words involved is m, then the data governance complexity is m+n, etc.; it can also be other data governance complexity judgment methods, which are not limited in this embodiment.

[0073] In the specific implementation, the intelligent agent breaks down the complex data governance tasks according to the prompts and analyzes the required steps.

[0074] Step A32, when the data governance complexity is greater than or equal to the complexity threshold, generate multiple governance tasks based on the prompt word.

[0075] It is understandable that when the complexity of the data governance under discussion reaches or exceeds the set complexity threshold, the system will generate multiple corresponding governance tasks based on the provided prompt words to ensure the integrity and accuracy of the data. The complexity threshold can be limited according to actual conditions, and this embodiment does not limit this.

[0076] It should be noted that the generation of multiple governance tasks according to the prompt words may be a large language model that has been pre-trained to perform task decomposition according to the prompt words, obtain multiple governance tasks and generate a task flow.

[0077] Step A33: Generate a data governance task flow based on the multiple governance tasks.

[0078] It is understandable that the data governance task flow is obtained by arranging the various governance tasks in sequence.

[0079] In this implementation, the characters are decomposed according to the complexity of the prompt words to generate a data governance task flow including multiple tasks, and each data governance task in the data governance task flow is processed separately to guide the completion of all tasks. Data governance can be carried out more quickly and effectively based on the decomposed tasks.

[0080] The above are only feasible implementations of step S30 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S30.

[0081] Step S40: Generate a governance tool and data governance tool parameters according to the data governance task, and perform data governance according to the governance tool and the data governance tool parameters.

[0082] It is understandable that the corresponding governance tool should be selected according to the data governance task, and the tool parameters that should be called by the governance tool should be determined based on the governance tool and the data governance task. Different parameters will produce different effects.

[0083] It should be noted that tools refer to the APIs that agents call. There are two important factors here: first, the agent can call the right tools, and second, the agent can use the tools correctly. Tools can expand the action space of large language model agents.

[0084] In practical applications, tools are usually made into APIs and provided to users for calling. However, in the field of data governance, some dedicated environments and data tools are often required, such as data lake warehouses, Spark, Flink, etc., and general large models lack field adaptability; at the same time, due to the diversity and inconsistency of tool types and parameters, users need to choose tools according to their needs. This proposal targets data governance tools, organizes the tool names, calling functions, input parameters, output types, etc. into a unified format, and through guidance and supervision modules, enables large model intelligent agents to handle complex data governance tasks.

[0085] In specific implementation, the automated data governance process can refer to Figure 5 In the figure, the user inputs natural language instructions, and the guidance module extracts entities according to the natural language instructions, supplements the context details and determines whether the information is missing. If the information is missing, the user's natural language instructions can be retrieved again. If the information is not correct, machine-oriented instructions are generated, so that the large language model outputs structured data, and institutional instructions are generated based on structured data; the tool to be called is inferred based on the machine-oriented instructions, and the supervision module determines whether the governance tool is selected correctly. If not, it is regenerated. If it is correct, the tool parameters are generated, and the tool parameters are further determined. If it is correct, the tool is called for data governance. Then the data task results are obtained to determine whether the data governance quality is improved. If it is improved, the task in the task flow is completed, and all tasks in the task flow are repeated to end.

[0086] This embodiment provides a data governance method based on a large language model, which obtains high-quality prompt instructions by identifying key elements, improves the stability and reliability of data governance results, and self-optimizes the governance tools and governance tool parameters used in governance tasks based on the feedback of the governance results, thereby reducing the need for manual intervention, reducing the risks brought by human factors, improving data governance efficiency and decision-making quality, and avoiding the problem of divergent results when processing complex data governance tasks.

[0087] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 6, step S40 includes steps S41-S42: Step S41: Obtain an execution action according to the data governance task, and obtain a governance tool according to the execution action.

[0088] It is understandable that the execution action is obtained according to the data governance task, and the governance tool is obtained according to the execution action. The intelligent agent determines the specific behavior that needs to be executed this time, searches the tool library and sorts the tools, and decides which tool to call.

[0089] Step S42: obtaining data governance tool parameters according to the governance tool, and performing data governance according to the data governance tool parameters.

[0090] It should be noted that the data governance tool parameters obtained according to the governance tool can be a guidance module to guide the large model, requiring the model to output structured data, generating correct tool calling parameters, and specifying the model output format, such as JSON, XML, etc. Based on providing a structured data template, requiring the model to fill in data, the intelligent agent generates tool parameters, and calls the tool.

[0091] It is understandable that based on structured model data, machines can better interpret and execute.

[0092] It should be added that when generating data governance tools and governance tool parameters, the supervision module checks whether the output of the large model is structured; Verify Schema: Verify whether the model output conforms to the Schema definition. Identify redundant characters: When invalid characters are detected in the output, reject the model results and point out the errors in the model results. The supervision module analyzes the data results, reconciles the data, and evaluates whether they meet the expectations of the data tool: check data consistency, check whether the data volume is consistent before and after governance, check field names and data types; if data quality improves: give a positive evaluation, the tool parameters are correct; if data quality decreases: give a negative evaluation: the tool parameters are wrong, and regenerate the tool parameters.

[0093] In a feasible implementation manner, step S42 may include steps A43 to A46: Step A43, obtain the governance results.

[0094] It is understandable that governance outcomes can be either effective or ineffective.

[0095] It should be noted that the intelligent agent observes the execution results. The supervision module monitors the data quality in the process, measures the consistency of data governance, data quality measurement results, data quality trends, data quality issues, and evaluates the suitability of the output selection tool of the intelligent agent. The supervision module calculates a variety of data quality evaluation indicators, including indicators such as the number of table rows, null values, duplicate values, outliers, field aggregation, and their volatility. The supervision module analyzes whether the data quality of the results of this data task has decreased.

[0096] Step A44, performing governance quality assessment based on the governance results to obtain the governance status.

[0097] It will be appreciated that governance states may include both degraded data quality and improved data quality.

[0098] It should be noted that the governance quality assessment can be an assessment given by the supervision module, and how to evaluate the data quality can be achieved through the reward function in the large language model.

[0099] It should be understood that when constructing a large language model and introducing a guidance module and a supervision module, the large language model including the guidance module and the supervision module is trained and optimized in advance.

[0100] In the specific implementation, the steps for training the guidance and supervision module are as follows: prepare the training data set, including: data governance standards and specifications, data policies and compliance requirements, data privacy, security policies, data sharing and exchange rules, data classification and grading; metadata, data dictionary, structured data classification, definition, structure; data governance process, data lifecycle management, data governance best practices; data quality standards, quality dimensions, evaluation methods, quality monitoring process; data governance tools, including name, function description, input parameters, output response; prompt word template. Train the guidance and supervision module model; jointly optimize the intelligent agent and guidance and supervision module, where the data set structure 1 can be referenced. Figure 7 And the data set structure 2 can refer to Figure 8 .

[0101] Step A45, when the governance status is that the data quality has deteriorated, a negative evaluation is generated, and the parameters of the data governance tool are updated based on the negative evaluation.

[0102] It should be noted that the data quality deteriorates: negative evaluation is output and the tool selection is wrong; the intelligent agent receives the supervised evaluation: negative evaluation: the agent adjusts according to the evaluation results, reselects the tool, or regenerates the tool parameters; a negative evaluation is given again, the process is terminated, and human intervention is required. The intelligent agent repeats the process according to the evaluation results until the subtask is completed.

[0103] It should be noted that when the data quality deteriorates, negative evaluations are generated and new data governance tool parameters are generated, including: when the data quality deteriorates, a historical status is obtained based on the negative evaluations, and historical actions are obtained based on the data governance tool parameters and the governance tool; the parameter vector of the large language model is optimized based on the historical status and the historical actions to obtain a new model parameter vector; and new data governance tool parameters are generated based on the new model parameter vector.

[0104] In the specific implementation, the GPT model is selected and initialized. The history H is defined as the sequence of input i, action a, and reward r: t =i 1 , a 1 , r 1 ,···,i t , a t , r t .

[0105] Among them, i includes user input, model output, action is the prompt word generated by the model, and the reward function is based on the quality of the generated prompt word. The agent will rely on previous history when taking actions, so the overall state S t Can be regarded as history t Function:

[0106] Among them, the entire state space is represented by S.

[0107] The agent's actions follow the stochastic strategy π, and according to the current state S, the probability of the output action is:

[0108] For select tools, the rewards are:

[0109] For the generation parameters, the rewards are:

[0110] Set negative rewards (penalties) to discourage the agent from taking wrong actions.

[0111] Further introduce the reward function:

[0112] Among them, r θ Indicates that in state s, according to strategy π θ (Parameter is θ) The reward obtained by choosing action a, Represents strategy π PPO (a|s) and the benchmark strategy π baseis the KL divergence between (a|s), and λ is a hyperparameter that weighs the reward and KL divergence.

[0113] The KL divergence is used to measure the difference between two distributions and can be obtained by the following formula:

[0114] Among them, the objective function is:

[0115] in, In the strategy π θ Under the condition of , the expectation of state S and action a. The purpose of the guidance and supervision module is to make the intelligent agent take correct actions. Therefore, there is no need to pay attention to the output of the guidance and supervision module, but only to optimize the output of the intelligent agent model.

[0116] Furthermore, the goal of the policy gradient is to maximize the policy π θ The expected return J(θ) has a gradient of:

[0117] Policy update equation:

[0118] Among them, θ is the parameter vector of the model, η is the learning rate, is the gradient of the objective function J with respect to the policy parameters.

[0119] Furthermore, the objective function J(θ) aims to maximize the total reward generated by the strategy, while introducing KL divergence to limit the pace of strategy updates and ensure the stability of updates. By optimizing the objective function, the policy parameters θ are adjusted to balance exploration and exploitation during the learning process and avoid performance fluctuations caused by too large a policy update.

[0120] It should be further explained that the optimizing the parameter vector of the large language model according to the historical state and the historical action to obtain a new model parameter vector includes: obtaining a reward value according to the historical state and a preset reward function; obtaining an optimization strategy according to the reward value, the historical action and the historical state; obtaining a model gradient of the large language model according to the optimization strategy; optimizing the parameter vector of the large language model based on the model gradient and a preset learning rate to obtain a new model parameter vector.

[0121] It should be noted that the optimization step can be to initialize a set of parameters θ 0 ; Perform multiple rounds of iterations: Use the current parameter θ k Interact with the environment to generate t , at};based on , find the value based on the current state {S t , a t The optimal parameter θ of k+1 .

[0122] It should be understood that the optimization method can train the guidance and supervision modules separately, or can jointly train the guidance and supervision modules and the intelligent agent.

[0123] In the framework of the proposed method, the intelligent agent receives observations of the environment and outputs actions; the guidance and supervision module receives actions output by the user and the agent and observations of the environment, and outputs guidance and supervision evaluation information.

[0124] Step A46, when the governance status is that the data quality has improved, a positive evaluation is generated to complete data governance.

[0125] It should be noted that the data quality improves: a positive evaluation is given and the tool is selected correctly; a positive evaluation is given: the agent continues the process; the guidance and supervision process is repeated: a positive evaluation is given and the agent continues the process; until all data governance tasks in the data governance task flow are completed.

[0126] In this embodiment, by pre-training and optimizing the large language model including the supervision module and the guidance module, corresponding data governance tools and tool parameters can be accurately generated for data governance tasks, so that data governance tasks can be processed more efficiently and effectively.

[0127] The above are merely feasible implementations after step S42 provided in this embodiment, and this embodiment does not specifically limit the specific implementations after step S42.

[0128] It should be noted that the data governance method based on the large language model of the present invention can also be applied to the data governance device of the large language model; the device is composed of an artificial intelligence layer, a business platform layer, and a task scheduling layer. The architecture of the data governance device of the language model can refer to Fig. 9 , Fig. 9 The middle artificial intelligence layer provides automated data governance capabilities. Through artificial intelligence technology, it receives user instructions, calls data processing tools, monitors data quality, and realizes automation of the entire data governance process.

[0129] Further, it specifically includes: Guidance and supervision module: responsible for guiding and supervising the intelligent agent, standardizing the output of the intelligent agent, limiting the action space of the intelligent agent, and supervising the data quality in the data governance process. Intelligent agent: used to process natural language requests, generate data governance pipelines, and call tools.

[0130] Furthermore, the business platform layer provides data management and processing, and the supported services include data source management, import and export of structured and unstructured data, system resource management, etc. Specifically, it includes: Data aggregation system: used to manage external data sources, and synchronize heterogeneous data sources such as relational databases, non-relational databases, file / object storage, and real-time messages to the data management system; Data management and control module: used to store and manage data, and preview the collected data Data development module: used to extract, convert, clean, aggregate, filter, etc. of the collected data. Data application module: used to apply the processed data, including data sharing, data analysis, etc. Each module contains relevant data governance tools to execute data tasks in the data governance pipeline. Including data synchronization, data preprocessing, data analysis, etc.

[0131] Furthermore, the task scheduling layer is decoupled from the business platform layer and does not undertake business logic itself, including: Task scheduling center: used to schedule task executors to perform data processing tasks, responsible for managing scheduling information, and issuing scheduling requests according to scheduling configurations. Task executors: used to perform specific data processing tasks, responsible for receiving scheduling requests and executing task logic.

[0132] It should be emphasized that the calling process of the data management device based on the large language model can refer to Fig.10 , the external data in the figure is collected and stored; the user inputs the data governance requirements; high-quality prompts are generated and the automated data governance pipeline is generated; the data tasks of the data governance pipeline are executed and the data quality is monitored; the automated data governance is completed and the results are returned.

[0133] This embodiment provides a data governance method based on a large language model, which jointly trains the guidance and supervision modules through data governance knowledge domains and data quality metrics. Through the joint training method, the guidance and supervision modules can play a role independently or jointly in different scenarios. With the introduction of the guidance and supervision modules, the accuracy and efficiency of the intelligent agent in processing data tasks have been significantly improved. The guidance and supervision modules effectively reduce the probability of manual intervention due to divergence of model results and improve production efficiency.

[0134] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the data governance method based on a large language model of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0135] This application also provides a data management device based on a large language model, please refer to Fig.11 , the data management device based on the large language model includes: An acquisition module 10 is used to acquire natural language information input by a target user into a large language model; A prompt word generation module 20, used to identify key elements according to the natural language information, and perform context supplement based on the key elements to obtain prompt words; A data governance module 30, configured to generate a data governance task flow according to the prompt word, wherein the data governance task flow includes at least one data governance task generated based on the prompt word; The data governance module 30 is further used to generate a governance tool and data governance tool parameters according to the data governance task, and perform data governance according to the governance tool and the data governance tool parameters.

[0136] The data governance device based on a large language model provided by the present application adopts the data governance method based on a large language model in the above-mentioned embodiment, which can solve the technical problem in the prior art that the large language model performs poorly when the feedback diverges. Compared with the prior art, the beneficial effects of the data governance device based on a large language model provided by the present application are the same as the beneficial effects of the data governance method based on a large language model provided by the above-mentioned embodiment, and the other technical features in the data governance device based on a large language model are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0137] The present application provides a data governance device based on a large language model, and the data governance device based on a large language model includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the data governance method based on the large language model in the above-mentioned embodiment one.

[0138] Reference below Fig.12 , which shows a schematic diagram of the structure of a data management device based on a large language model suitable for implementing an embodiment of the present application. The data management device based on a large language model in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.12 The data governance device based on the large language model shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0139] like Fig.12As shown, the data governance device based on the large language model may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage device 1003 to the random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the data governance device based on the large language model are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the data management device based on the large language model to communicate with other devices wirelessly or by wire to exchange data. Although the data management device based on the large language model with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0140] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0141] The data governance device based on a large language model provided by the present application adopts the data governance method based on a large language model in the above-mentioned embodiment, which can solve the technical problem in the prior art that the large language model performs poorly when the feedback diverges. Compared with the prior art, the beneficial effects of the data governance device based on a large language model provided by the present application are the same as the beneficial effects of the data governance method based on a large language model provided by the above-mentioned embodiment, and the other technical features of the data governance device based on a large language model are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0142] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0143] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0144] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the data governance method based on a large language model in the above-mentioned embodiment.

[0145] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0146] The computer-readable storage medium may be included in a data management device based on a large language model; or may exist independently without being assembled into a data management device based on a large language model.

[0147] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by a data governance device based on a large language model, the data governance device based on the large language model: obtains natural language information input by the target user into the large language model; identifies key elements based on the natural language information, performs context supplement based on the key elements, and obtains prompt words; generates a data governance task flow based on the prompt words, and the data governance task flow includes at least one data governance task generated based on the prompt words; generates a governance tool and data governance tool parameters based on the data governance task, and performs data governance based on the governance tool and the data governance tool parameters.

[0148] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0149] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0150] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0151] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned data governance method based on a large language model, and can solve the technical problem in the prior art that a large language model performs poorly when facing the feedback divergence problem. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the data governance method based on a large language model provided in the above-mentioned embodiment, and will not be elaborated here.

[0152] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the data governance method based on a large language model as described above.

[0153] The computer program product provided by the present application can solve the technical problem in the prior art that a large language model performs poorly when facing the feedback divergence problem. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the data governance method based on the large language model provided by the above embodiment, and will not be elaborated here.

[0154] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A data governance method based on a large language model, characterized in that: The method comprises: Obtain natural language information input by the target user into the large language model; Identify key elements according to the natural language information, and perform context supplement based on the key elements to obtain prompt words; Generate a data governance task flow according to the prompt word, wherein the data governance task flow includes at least one data governance task generated based on the prompt word; A governance tool and data governance tool parameters are generated according to the data governance task, and data governance is performed according to the governance tool and the data governance tool parameters.

2. The method according to claim 1, characterized in that The step of identifying key elements according to the natural language information and supplementing the context based on the key elements to obtain prompt words includes: When the natural language information is not missing, identifying key elements according to the natural language information to obtain the conversation topic, user intention and entity; When the conversation topic, the user intention and the entity are verified to be qualified, context supplement is performed according to the conversation topic, the user intention and the entity to obtain background information and detail information; A prompt word is obtained according to the background information and the detail information.

3. The method according to claim 1, characterized in that The step of generating a data governance task flow according to the prompt word comprises: Obtaining data governance complexity based on the prompt word; When the data governance complexity is greater than or equal to a complexity threshold, generating multiple governance tasks based on the prompt word; A data governance task flow is generated according to the plurality of governance tasks.

4. The method according to claim 1, characterized in that The step of generating a governance tool and data governance tool parameters according to the data governance task, and performing data governance according to the governance tool and the data governance tool parameters comprises: Obtaining an execution action according to the data governance task, and obtaining a governance tool according to the execution action; A data governance tool parameter is obtained according to the governance tool, and data governance is performed according to the data governance tool parameter.

5. The method according to claim 4, characterized in that After the steps of generating a governance tool and data governance tool parameters according to the data governance task, and performing data governance according to the governance tool and the data governance tool parameters, the method further includes: Obtain governance results; Performing a governance quality assessment based on the governance results to obtain a governance status; When the governance status is that the data quality has deteriorated, a negative evaluation is generated, and the parameters of the data governance tool are updated based on the negative evaluation; When the governance status is that data quality has improved, a positive evaluation is generated and data governance is completed.

6. The method according to claim 5, characterized in that The step of generating a negative evaluation when the governance status is that the data quality is degraded, and updating the parameters of the data governance tool based on the negative evaluation includes: When the governance status is that the data quality has deteriorated, a historical status is obtained according to the negative evaluation, and a historical action is obtained according to the data governance tool parameter and the governance tool; Optimizing a parameter vector of the large language model according to the historical state and the historical action to obtain a new model parameter vector; The data governance tool parameters are updated according to the new model parameter vector.

7. A data management device based on a large language model, characterized in that: The data management device based on the large language model includes: An acquisition module is used to acquire natural language information input by a target user into a large language model; A prompt word generation module, used for identifying key elements according to the natural language information, and performing context supplement based on the key elements to obtain prompt words; A data governance module, configured to generate a data governance task flow according to the prompt word, wherein the data governance task flow includes at least one data governance task generated based on the prompt word; The data governance module is further used to generate a governance tool and data governance tool parameters according to the data governance task, and perform data governance according to the governance tool and the data governance tool parameters.

8. A data management device based on a large language model, characterized in that: The data management device based on a large language model includes: a memory, a processor, and a data management program based on a large language model stored in the memory and executable on the processor, wherein the data management program based on a large language model is configured to implement the data management method based on a large language model as described in any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium stores a data management program based on a large language model, and when the data management program based on a large language model is executed by a processor, the data management method based on a large language model as described in any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product comprises a data management program based on a large language model, and when the data management program based on a large language model is executed by a processor, the steps of the data management method based on a large language model as described in any one of claims 1 to 6 are implemented.

Images