Data processing method, device, storage medium and program product based on large model

Automatically analyze exception problems through large language models and call adapter tools according to standard operating procedures, solving the problem of time-consuming and labor-intensive and inconsistent operation standards of existing exception analysis methods, improving analysis efficiency and accuracy.

CN118428480BActive Publication Date: 2024-11-22ALIBABA CLOUD FEITIAN (HANGZHOU) CLOUD COMPUTING TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410872393.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-01
Publication Date
2024-11-22
Estimated Expiration
2044-07-01

AI Technical Summary

Technical Problem

The existing abnormality analysis methods rely on manual verification, which is time-consuming and labor-intensive, and the operating standards are inconsistent, making it difficult to accurately diagnose the causes of abnormalities.

Method used

Using a data processing method based on a large language model, we use the problem to be analyzed and the target standard operating process, determine the adapted analysis tools, generate prompt information, and call the large language model for automatic analysis, and generate analysis results.

Benefits of technology

It improves the efficiency of abnormal analysis, reduces labor costs, and ensures the accuracy and consistency of analysis results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118428480B_ABST
    Figure CN118428480B_ABST
Patent Text Reader

Abstract

The embodiment of the present application provides a data processing method, device, storage medium and program product based on a large model. The method includes the following steps: obtaining a problem to be analyzed and a target standard operating procedure for analyzing the problem to be analyzed; the target standard operating procedure involves multiple steps; determining the target analysis tools adapted to each of the multiple steps; generating a first prompt message according to the problem to be analyzed, the target standard operating procedure and the target analysis tools adapted to each of the multiple steps; inputting the first prompt message into the large language model so that the large language model calls the target analysis tools adapted to each of the multiple steps according to the target standard operating procedure to obtain the analysis result of the problem to be analyzed. The technical solution provided by the embodiment of the present application can improve the efficiency of abnormal analysis and reduce labor costs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data processing method, device, storage medium and program product based on a large model. Background Art

[0002] In some fields, such as discrete manufacturing and supply chain, it is required to strictly follow the Standard Operating Procedure (SOP) to perform exception analysis, that is, it is necessary to analyze the abnormal situations that occur according to the pre-specified SOP to determine the cause of the abnormality.

[0003] The existing abnormal analysis methods basically rely on professional analysts, who manually verify according to the SOP, which is time-consuming and labor-intensive. In addition, the operating standards of different analysts are inconsistent, making it difficult to accurately and effectively diagnose the real cause. Summary of the invention

[0004] In view of the above problems, the present application is proposed to provide a large model-based data processing method, device, storage medium and program product that solve the above problems or at least partially solve the above problems.

[0005] In a first aspect of the present application, a data processing method based on a large language model is provided, comprising:

[0006] Obtaining a problem to be analyzed and a target standard operating procedure for analyzing the problem to be analyzed; the target standard operating procedure involves multiple steps;

[0007] Determining target analysis tools suitable for each of the multiple steps;

[0008] Generate first prompt information according to the problem to be analyzed, the target standard operating procedure, and the target analysis tools adapted to each of the multiple steps; the first prompt information is used to prompt the large language model to call the target analysis tools adapted to each of the multiple steps according to the target standard operating procedure to analyze the problem to be analyzed;

[0009] The first prompt information is input into the large language model, so that the large language model calls the target analysis tools adapted to each of the multiple steps according to the target standard operating process to obtain the analysis result of the problem to be analyzed.

[0010] In an implementation manner provided by the first aspect of the present application, generating first prompt information according to the problem to be analyzed, the target standard operating procedure, and the target analysis tool adapted to each of the multiple steps includes:

[0011] Acquire a first prompt word template; the first prompt word template includes a task instruction related to the problem analysis task;

[0012] The problem to be analyzed, the target standard operating procedure, and the target analysis tools adapted to each of the multiple steps are filled in the corresponding positions in the first prompt word template to obtain the first prompt information.

[0013] In an implementation manner provided in the first aspect of the present application, determining the target analysis tools that are suitable for each of the multiple steps includes:

[0014] According to the problem to be analyzed, the target standard operating procedure and the functional description information of multiple candidate analysis tools, suitable target analysis tools are determined for the multiple steps respectively.

[0015] In an implementation manner provided by the first aspect of the present application, according to the problem to be analyzed, the target standard operating procedure, and the functional description information of multiple candidate analysis tools, determining the adapted target analysis tools for the multiple steps respectively includes:

[0016] Generate second prompt information according to the problem to be analyzed, the target standard operating process, and functional description information of multiple candidate analysis tools; the second prompt information is used to prompt the large language model to determine adapted target analysis tools for the multiple steps respectively from the multiple candidate analysis tools;

[0017] The second prompt information is input into the large language model, so that the large language model can determine adapted target analysis tools for the multiple steps respectively from the multiple candidate analysis tools.

[0018] In an implementation manner provided by the first aspect of the present application, second prompt information is generated according to the problem to be analyzed, the target standard operating procedure, and functional description information of multiple candidate analysis tools, including:

[0019] Acquire a second prompt word template; the second prompt word template includes a task instruction related to the tool screening task;

[0020] The problem to be analyzed, the target standard operating procedure, and functional description information of multiple candidate analysis tools are filled in the corresponding position in the second prompt word template to obtain second prompt information.

[0021] In an implementation manner provided in the first aspect of the present application, the second prompt information is used to prompt the large language model to determine adapted target analysis tools for the multiple steps respectively from the multiple candidate analysis tools and to determine input parameters for the target analysis tools.

[0022] In an implementation manner provided by the first aspect of the present application, generating first prompt information according to the problem to be analyzed, the target standard operating procedure, and the target analysis tool adapted to each of the multiple steps includes:

[0023] A first prompt message is generated according to the problem to be analyzed, the target standard operating procedure, the target analysis tools adapted by each of the multiple steps, and the input parameters of the target analysis tool.

[0024] In one implementation manner provided in the first aspect of the present application, the method further includes:

[0025] When the large language model cannot complete the task indicated by the second prompt information, re-determine the target standard operating procedure for analyzing the problem to be analyzed;

[0026] Based on the re-determined target standard operating procedure for analyzing the problem to be analyzed, new second prompt information is generated for the large language model.

[0027] In an implementation manner provided in the first aspect of the present application, the first prompt information includes historical question and answer content, and / or the second prompt information includes historical question and answer content.

[0028] In an implementation manner provided by the first aspect of the present application, obtaining a target standard operating procedure for analyzing the problem to be analyzed includes:

[0029] Vectorizing the problem to be analyzed to obtain a problem vector;

[0030] Acquire function description vectors of multiple candidate standard operating procedures; the function description vectors of the multiple candidate standard operating procedures are obtained by vectorizing the process descriptions of the multiple candidate standard operating procedures respectively;

[0031] According to the vector similarity between the problem vector and the function description vectors of the multiple candidate standard operating procedures, a target standard operating procedure for analyzing the problem to be analyzed is determined from the multiple candidate standard operating procedures.

[0032] In an implementation manner provided by the first aspect of the present application, determining a target standard operating procedure for analyzing the problem to be analyzed from the multiple candidate standard operating procedures according to the vector similarity between the problem vector and the function description vectors of the multiple candidate standard operating procedures includes:

[0033] The standard operating procedure with the largest vector similarity is determined as the target standard operating procedure for analyzing the problem to be analyzed.

[0034] A second aspect of the present application provides a data processing method based on a large language model, comprising:

[0035] Obtaining a problem to be analyzed and a target standard operating procedure for analyzing the problem to be analyzed; the target standard operating procedure involves multiple steps;

[0036] Generate second prompt information according to the problem to be analyzed, the target standard operating process, and functional description information of multiple candidate analysis tools; the second prompt information is used to prompt the large language model to determine adapted target analysis tools for the multiple steps respectively from the multiple candidate analysis tools;

[0037] The second prompt information is input into the large language model, so that the large language model can determine adapted target analysis tools for the multiple steps respectively from the multiple candidate analysis tools.

[0038] In an implementation manner provided by the second aspect of the present application, obtaining a target standard operating procedure for analyzing the problem to be analyzed includes:

[0039] Vectorizing the problem to be analyzed to obtain a problem vector;

[0040] Acquire function description vectors of multiple candidate standard operating procedures; the function description vectors of the multiple candidate standard operating procedures are obtained by vectorizing the process descriptions of the multiple candidate standard operating procedures respectively;

[0041] According to the vector similarity between the problem vector and the function description vectors of the multiple candidate standard operating procedures, a target standard operating procedure for analyzing the problem to be analyzed is determined.

[0042] In an implementation manner provided in the second aspect of the present application, the method further includes:

[0043] When the large language model cannot complete the task indicated by the second prompt information, re-determining a target standard operating procedure for analyzing the problem to be analyzed according to the vector similarity between the problem vector and the function description vectors of the multiple candidate standard operating procedures;

[0044] Based on the re-determined target standard operating procedure for analyzing the problem to be analyzed, new second prompt information is generated for the large language model.

[0045] In a third aspect of the present application, an electronic device is provided. The electronic device comprises: a memory and a processor, wherein:

[0046] The memory is used to store programs;

[0047] The processor is coupled to the memory, and is used to execute the program stored in the memory to implement any of the methods described above.

[0048] According to a fourth aspect of the present application, a computer-readable storage medium storing a computer program is provided, wherein the computer program can implement any of the above-described methods when executed by a computer.

[0049] According to a fifth aspect of the present application, a computer program product is provided, comprising a computer program, which implements any of the above-described methods when executed by a processor.

[0050] In the technical solution provided in the embodiment of the present application, a prompt message is generated according to the problem to be analyzed, the target standard operating process for analyzing the problem to be analyzed, and the target analysis tools adapted to each of the multiple steps involved in the process; the prompt message is input into the large language model to prompt the large language model to call the target analysis tools adapted to each of the multiple steps according to the target standard operating process to obtain the analysis result of the problem to be analyzed. In other words, the technical solution provided in the embodiment of the present application utilizes the task planning capability of the large language model, and the large language model automatically analyzes the problem according to the standard operating process to obtain the analysis result. It can be seen that the technical solution provided in the embodiment of the present application can improve the efficiency of abnormal analysis and reduce labor costs.

[0051] In the technical solution provided in the embodiment of the present application, prompt information is generated based on the problem to be analyzed, the target standard operating process for analyzing the problem to be analyzed, and the functional description information of multiple alternative analysis tools; the prompt information is input into the large language model to prompt the large language model to determine the adapted target analysis tools for multiple steps from multiple alternative analysis tools. In other words, the technical solution provided in the embodiment of the present application utilizes the semantic understanding ability of the large language model to automatically match appropriate analysis tools for the processes in the standard operating process. In this way, analysts do not need to spend time and experience looking for analysis tools that are adapted to each step in the standard operating process, which can improve the efficiency of abnormal analysis and reduce labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 A flowchart of a data processing method provided in one embodiment of the present application;

[0054] Figure 2 A schematic diagram of an interface provided in an embodiment of the present application;

[0055] Figure 3 A flowchart of a data processing method provided by another embodiment of the present application;

[0056] Figure 4 A flowchart of a data processing method provided in yet another embodiment of the present application;

[0057] Figure 5 A structural block diagram of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0058] In order 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 according to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0059] In addition, some of the processes described in the specification, claims and the above-mentioned figures of the present application include multiple operations that appear in a specific order, and these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to different types.

[0060] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0061] First, the vocabulary involved in the embodiments of the present application is explained. It can be understood that this explanation is for a clearer understanding of the embodiments of the present application, and does not necessarily constitute a limitation on the embodiments of the present application.

[0062] Large Language Model (LLM) is a deep learning model trained on massive text data. It can not only generate natural language text, but also deeply understand the meaning of the text and handle various natural language tasks such as text summarization, question answering, translation, etc. Large language model can also be referred to as large model.

[0063] Fine-tuning: Based on the pre-trained large-scale deep learning model, further training is performed for specific tasks or fields.

[0064] Standard Operating Procedure (SOP) refers to the standard operating procedures and requirements of a certain event described in a unified format to guide and regulate daily work. Generally speaking, SOP refers to an internal procedure designed to execute complex daily affairs within limited time and resources.

[0065] Vector Embedding: Convert natural language into a vector of fixed dimension to represent the meaning and context information of the natural language.

[0066] In order to solve or partially solve the above technical problems, the embodiment of the present application provides a data processing method based on a large language model. The execution subject of the data processing method can be a server. Of course, when the hardware performance and software performance of the client can support the operation of the large language model, the execution subject of the data processing method can be the client.

[0067] The following will take the execution subject as the server as an example to introduce the data processing method provided by the embodiment of the present application. Figure 1 As shown, the method may specifically include the following steps:

[0068] 101. Receive issues to be analyzed.

[0069] The problem to be analyzed is a problem described in natural language and may be provided / input by a user.

[0070] For example, the question to be analyzed may be: "What is the reason why the order completion rate did not meet the target in May 2023?"

[0071] The user can input the problem to be analyzed through the human-computer interaction interface provided by the client. In response to the user inputting the problem to be analyzed, the problem to be analyzed is sent to the server. Figure 2 As shown, the human-computer interaction interface may be a human-computer chat interface 200, on which is displayed a question to be analyzed 201 input by the user, "What is the reason why the order completion rate in May 2023 did not meet the standard?"

[0072] 102. Perform vector encoding on the problem to be analyzed to obtain a problem vector.

[0073] In this application, the problem to be analyzed can be understood as an anomaly to be analyzed.

[0074] Among them, vector encoding can also be called vectorization or vector embedding.

[0075] Vector embedding technology can be used to map the problem to be analyzed into a problem vector. Specifically, a word vector model can be used to map the problem to be analyzed into a problem vector, wherein the word vector model can be Word2Vec. Word2Vec is a model that learns semantic knowledge from a large amount of text corpus in an unsupervised manner. It represents each word as a high-dimensional vector and represents the semantic relationship between words through the similarity between vectors.

[0076] 103. Obtain process description vectors of multiple standard operating processes in a standard operating process library.

[0077] The multiple standard operating procedures in the standard operating procedure library are the multiple alternative standard operating procedures mentioned above. Each standard operating procedure involves multiple steps.

[0078] The process description vector of each standard operating process is obtained by vector encoding the process description of the standard operating process. Optionally, the process description of the standard operating process can be a functional description, that is, used to describe the functions implemented by the standard operating process. Different standard operating processes are used to implement different functions, or in other words, different standard operating processes are used to analyze different types of problems.

[0079] The standard operating process library may store multiple standard operating processes and a process description vector of each standard operating process.

[0080] In order to shorten the data analysis delay, the process description of the standard operating process can be vector-encoded in advance to obtain the process description vector of the standard operating process, and the standard operating process and its process description vector can be stored in the standard operating process library accordingly. In this way, when the above data processing method is subsequently executed, there is no need to vector-encode the process description of the standard operating process, and it can be directly obtained from the library.

[0081] Of course, in practical applications, if there is no need to consider the data analysis delay, the process description of the standard operating process in the standard operating process library can be vector encoded during the execution of the above data processing method to obtain a process description vector.

[0082] Optionally, the process description vector and the question vector can be obtained using the same word vector model.

[0083] 104. Calculate the vector similarity between the problem vector and the process description vectors of multiple standard operating processes.

[0084] Vector similarity is used to describe the similarity between two vectors. Generally, vector similarity can be calculated using the cosine similarity calculation formula, and of course, it can also be calculated using other methods.

[0085] 105. Determine the target standard operating procedure based on the matching results.

[0086] Select / determine a target standard operating procedure from multiple standard operating procedures based on vector similarity.

[0087] Exemplarily, a standard operating procedure with the greatest vector similarity is selected from multiple standard operating procedures as the target standard operating procedure.

[0088] In the embodiment of the present application, the standard operation process is automatically matched for the problem to be analyzed by vector matching. It can be seen that the technical solution provided by the embodiment of the present application can be applied to a variety of problem analysis scenarios.

[0089] Optionally, the target standard operating procedure may also be specified by the user for the problem to be analyzed. For example, the target standard operating procedure may be selected or input by the user. In one achievable implementation, after the user inputs the problem to be analyzed, a plurality of alternative standard operating procedures are provided to the user; based on the user's selection operation among the plurality of alternative standard operating procedures, the target standard operating procedure for analyzing the problem to be analyzed is determined. In another achievable implementation, the problem to be analyzed input by the user and the target standard operating procedure for analyzing the problem to be analyzed are received. That is, the target standard operating procedure for analyzing the problem to be analyzed is input by the user.

[0090] 106. Generate prompt information A (i.e., the second prompt information mentioned above) based on the problem to be analyzed, the target standard operating process, the functional descriptions of multiple analysis tools, and the historical question and answer content.

[0091] It should be noted that the prompt information in the embodiment of the present application may also be referred to as a prompt word (prompt).

[0092] The prompt information A is used to prompt the large language model to match the analysis tool used to implement each SOP step in the target standard operating process and determine the input parameters for the matched analysis tool. The large language model is pre-trained. It should be noted that the SOP step in the SOP refers to the step described in the SOP.

[0093] In an optional embodiment, prompt information A includes task instructions for tool screening tasks and tool input parameter acquisition tasks, questions to be analyzed, target standard operating procedures, functional descriptions of multiple analysis tools, and historical question and answer content. Among them, task instructions are used to indicate the tasks that need to be performed by the large language model. Optionally, task instructions may include task description information. Exemplarily, the task instruction may be: "Please determine the appropriate analysis tools and their input parameters for multiple steps in the standard operating procedure from multiple alternative analysis tools based on the functional description information of multiple alternative analysis tools."

[0094] In an optional implementation, a prompt word template is obtained; the prompt word template includes task instructions related to tool screening tasks and tool input parameter acquisition tasks; the function description information of the problem to be analyzed, the target standard operating process, and multiple alternative analysis tools are filled in the corresponding position in the prompt word template to obtain prompt information A. It should be noted that the analysis tool used to implement a certain SOP step may be one or more. That is: a certain SOP step may require one analysis tool to implement, or it may require multiple analysis tools to implement.

[0095] The functional description of the analysis tool is used to describe the functions that the analysis tool can achieve.

[0096] Exemplarily, the historical question and answer content includes the most recent N question and answer information between the user and the large language model. Among them, the value of N can be set according to actual needs, and the embodiment of the present application does not specifically limit this. Exemplarily, N is 2. It should be noted that before the user inputs the above-mentioned question to be analyzed, when the number of questions and answers between the user and the large language model is less than N, the historical question and answer content may include all question and answer information between the user and the large language model before the user inputs the above-mentioned question to be analyzed.

[0097] Exemplarily, the historical question and answer content includes: question and answer information between the user and the large language model within a preset time period before the user inputs the above-mentioned question to be analyzed.

[0098] When generating prompt information A, historical question and answer content is taken into account in order to improve the large language model's understanding of the context, which helps to improve the accuracy of the model. For example: the first question asked by the user is "What is the reason why the order completion rate in April 2024 did not meet the standard?" The reply of the large language model is "The completion rate in North China is very low"; the second question asked by the user is "What about May?" Among them, the semantics of the second question is not complete. Then, the large language model can complete the second question according to the context, and the completed question is "What is the reason why the order completion rate in May 2024 did not meet the standard?" The semantics of the completed question is complete.

[0099] Of course, in actual applications, when generating prompt information A, if the semantics of the question to be analyzed input by the user is complete, the historical question and answer content does not need to be considered.

[0100] Exemplarily, after receiving the question to be analyzed input by the user, semantic understanding of the question to be analyzed may be performed to determine whether the semantics of the question to be analyzed is complete. If it is not complete, the user may be prompted to re-enter the question to be analyzed.

[0101] The large language model in the embodiment of the present application is trained according to the training data. Exemplarily, the large language model can be fine-tuned according to the training data to obtain a large language model that meets the requirements. The training data may include: a sample of the problem to be analyzed and the expected analysis results.

[0102] In the above embodiment, after receiving the problem to be analyzed, the semantic understanding ability of the large model is used to match the analysis tool for each step in the target SOP, which belongs to the online dynamic matching of the analysis tool for the SOP.

[0103] As an option, a corresponding relationship between the SOP and the analysis tool adapted to each step in the SOP can be established in advance. Subsequently, the analysis tool adapted to each step in the SOP can be directly determined according to the SOP. The above-mentioned corresponding relationship can be established manually, or it can be established using the semantic understanding ability of a large language model, and the embodiment of the present application does not specifically limit this. This scheme belongs to offline matching analysis tools for SOP.

[0104] 107. Input the prompt information A into the large language model to obtain a first result output by the large language model.

[0105] The first result includes: an analysis tool for implementing each SOP step in the target standard operating process and input parameters of the analysis tool.

[0106] Input parameters refer to the parameters required to call the analysis tool; output parameters refer to the values ​​returned by the analysis tool after the analysis tool is called.

[0107] In the above embodiment, the large language model determines the input parameters of the analysis tool in step 107, so the large language model can directly use the input parameters in step 109. Optionally, the input parameters of the analysis tool can be determined by the large language model in step 109. In this way, the large language model only needs to match the analysis tool for each SOP step in step 107.

[0108] 108. Generate prompt information B (i.e., the first prompt information mentioned above) based on the problem to be analyzed, the target standard operating process, the first result, and the historical question and answer content.

[0109] Among them, prompt information B is used to prompt the large language model to call the analysis tool matched to the SOP step in the target standard operating process according to the target standard operating process and the input parameters of the matched analysis tool to obtain the analysis result of the problem to be analyzed.

[0110] Prompt information B includes task instructions for the problem analysis task, problems to be analyzed, target standard operating procedures, first results, and historical question and answer content.

[0111] The task instruction is used to indicate the task that the large language model needs to perform. Optionally, the task instruction may include task description information. Exemplarily, the task instruction may be: "Please follow the standard operating procedures, call the analysis tools corresponding to the multiple steps in the process, and analyze the problem to be analyzed."

[0112] In an optional implementation, a prompt word template is obtained; the prompt word template includes task instructions related to the problem analysis task; the problem to be analyzed, the target standard operating process and the first result are filled in the corresponding position in the prompt word template to obtain the first prompt information.

[0113] Exemplarily, when generating prompt information B, historical question and answer content may not be considered.

[0114] 109. Input the prompt information B into the large language model to obtain a second result output by the large language model.

[0115] The second result includes the analysis result of the problem to be analyzed.

[0116] Optionally, the second result may also include a structure tree constructed by the large language model in the process of obtaining the analysis result of the above-mentioned problem to be analyzed.

[0117] The nodes in the structure tree correspond to the SOP steps executed by the large language model to obtain the above analysis results, and the connection relationship between the nodes in the structure tree is determined by the relationship between the executed SOP steps in the target SOP. Among them, the executed SOP steps refer to the SOP steps executed by the large language model to obtain the above analysis results. It should be noted that the structure tree does not include the SOP steps in the target SOP that have not been executed by the large language model.

[0118] It should be noted that, generally, there are multiple end conditions in SOP. As long as one of the end conditions is met, the SOP process will end, and the analysis results can be obtained when the process ends. The structure tree only involves one end condition, because once a certain end condition is met, the SOP process will be ended, and the structure tree only contains the SOP steps that have been executed in the SOP process. Therefore, the structure tree only involves one end condition.

[0119] 110. Generate answer information for the question to be analyzed based on the second result.

[0120] In a feasible solution, the analysis result of the problem to be analyzed can be directly used as the reply information.

[0121] In another feasible solution, the analysis result of the problem to be analyzed can be further processed to increase the readability of the analysis result. The specific processing method can be set according to actual needs, and the embodiment of the present application does not specifically limit this.

[0122] 111. Send the reply information to the user.

[0123] Specifically, the reply information is sent to the user's client.

[0124] For example, Figure 2 As shown, after receiving the reply information 202 “The completion rate in North China is very low”, the client displays the reply information 202 on the human-computer chat interface 200 .

[0125] It should be noted that if the large language model cannot complete the matching of the analysis tool for the above-matched target standard operating process, the target standard operating process can be determined from at least one unselected standard operating process based on the vector similarity, for example: the standard operating process with the largest vector similarity is determined from at least one unselected standard operating process as the new target standard operating process, and the above steps 106 to 109 are continued to be executed based on the new target standard operating process until the above second result is obtained. The at least one unselected standard operating process refers to other standard operating processes in multiple standard operating processes except the selected standard operating process.

[0126] In the embodiment of the present application, starting from natural language, vector matching and large model semantic understanding capabilities are used to automatically select SOP, which is applicable to a wide range of scenarios; this solution uses the matching capabilities and task planning capabilities of the large model to automatically perform tool matching and task planning to obtain the final analysis results.

[0127] It should be noted that the technical solution provided by the embodiment of the present application is a dynamic matching SOP, and an analysis tool adapted to each step in the dynamic matching SOP. In this way, when a certain SOP is updated or a certain analysis tool is updated, the execution flow of this solution does not need to be changed, and only the SOP library and the tool library need to be updated, which is convenient for iterative updates. It should be noted that in actual applications, the updating of SOP and the updating of analysis tools often occur.

[0128] Figure 3The flowchart of the data processing method based on the large language model provided by an embodiment of the present application is shown. The execution subject of the data processing method can be the server. Of course, when the hardware performance and software performance of the client can support the operation of the large language model, the execution subject of the data processing method can be the client. Figure 3 As shown, the method includes:

[0129] 301. Obtain a problem to be analyzed and a target standard operating procedure for analyzing the problem to be analyzed.

[0130] Among them, the target standard operating process involves multiple steps.

[0131] 302. Determine target analysis tools that are suitable for each of the multiple steps.

[0132] 303. Generate first prompt information according to the problem to be analyzed, the target standard operating procedure, and the target analysis tools adapted to each of the multiple steps.

[0133] 304. Input the first prompt information into the large language model, so that the large language model calls the target analysis tools adapted to each of the multiple steps according to the target standard operating process to obtain the analysis result of the problem to be analyzed.

[0134] The specific implementation of the above steps 301 to 304 can refer to the corresponding content in the above embodiments, which will not be repeated here.

[0135] In the technical solution provided in the embodiment of the present application, a prompt message is generated according to the problem to be analyzed, the target standard operating process for analyzing the problem to be analyzed, and the target analysis tools adapted to each of the multiple steps involved in the process; the prompt message is input into the large language model to prompt the large language model to call the target analysis tools adapted to each of the multiple steps according to the target standard operating process to obtain the analysis result of the problem to be analyzed. In other words, the technical solution provided in the embodiment of the present application utilizes the task planning capability of the large language model, and the large language model automatically analyzes the problem according to the standard operating process to obtain the analysis result. It can be seen that the technical solution provided in the embodiment of the present application can improve the efficiency of abnormal analysis and reduce labor costs.

[0136] Optionally, the “determining target analysis tools suitable for each of the multiple steps” in 302 above includes:

[0137] 3021. Generate second prompt information based on the problem to be analyzed, the target standard operating procedure, and functional description information of multiple alternative analysis tools.

[0138] 3022. Input the second prompt information into the large language model so that the large language model can determine suitable target analysis tools for the multiple steps respectively from the multiple candidate analysis tools.

[0139] The specific implementation of the above steps 3021 to 3022 can refer to the corresponding content in the above embodiments, which will not be repeated here.

[0140] Optionally, the second prompt information is used to prompt the large language model to determine adapted target analysis tools for the multiple steps respectively from the multiple candidate analysis tools and to determine input parameters for the target analysis tools.

[0141] Optionally, in the above 303, “generating first prompt information according to the problem to be analyzed, the target standard operating process, and the target analysis tools adapted to each of the multiple steps” may include:

[0142] 3031. Generate a first prompt message according to the problem to be analyzed, the target standard operating procedure, the target analysis tools adapted for each of the multiple steps, and the input parameters of the target analysis tool.

[0143] Optionally, the method further includes:

[0144] 305. When the large language model cannot complete the task indicated by the second prompt information, redetermine a target standard operating procedure for analyzing the problem to be analyzed.

[0145] 306. Generate new second prompt information for the large language model based on the re-determined target standard operating procedure for analyzing the problem to be analyzed.

[0146] In one embodiment, the user may be prompted to re-enter or select a target standard operating procedure for analyzing the problem to be analyzed, and subsequent steps may be continued based on the new target standard operating procedure until the second result is obtained.

[0147] In another embodiment, a new target standard operating process may be re-determined based on the vector similarity between the problem vector and the process description vectors of multiple standard operating processes. Optionally, a new target standard operating process may be determined from at least one unselected standard operating process. Exemplarily, a standard operating process with the greatest vector similarity is determined from at least one unselected standard operating process as a new target standard operating process, and subsequent steps are continued based on the new target standard operating process until the second result is obtained.

[0148] It should be noted here that: for the contents of each step not fully described in detail in the method provided in the embodiment of the present application, please refer to the corresponding contents in the above embodiment, which will not be repeated here. In addition, in addition to the above steps, the method provided in the embodiment of the present application may also include other parts or all of the steps in the above embodiments, which can be specifically referred to the corresponding contents in the above embodiments, which will not be repeated here.

[0149] Figure 4 The flowchart of the data processing method based on the large language model provided by an embodiment of the present application is shown. The execution subject of the data processing method can be the server. Of course, when the hardware performance and software performance of the client can support the operation of the large language model, the execution subject of the data processing method can be the client. Figure 4 As shown, the method includes:

[0150] 401. Obtain a problem to be analyzed and a target standard operating procedure for analyzing the problem to be analyzed.

[0151] Among them, the target standard operating process involves multiple steps.

[0152] 402. Generate second prompt information according to the problem to be analyzed, the target standard operating procedure, and functional description information of multiple candidate analysis tools.

[0153] 403. Input the second prompt information into the large language model, so that the large language model can determine adapted target analysis tools for the multiple steps respectively from the multiple candidate analysis tools.

[0154] The specific implementation of the above steps 401 to 403 can refer to the corresponding content in the above embodiments, which will not be repeated here.

[0155] In the technical solution provided in the embodiment of the present application, prompt information is generated based on the problem to be analyzed, the target standard operating process for analyzing the problem to be analyzed, and the functional description information of multiple alternative analysis tools; the prompt information is input into the large language model to prompt the large language model to determine the adapted target analysis tools for multiple steps from multiple alternative analysis tools. In other words, the technical solution provided in the embodiment of the present application utilizes the semantic understanding ability of the large language model to automatically match appropriate analysis tools for the processes in the standard operating process. In this way, analysts do not need to spend time and experience looking for analysis tools that are adapted to each step in the standard operating process, which can improve the efficiency of abnormal analysis and reduce labor costs.

[0156] It should be noted here that: for the contents of each step not fully described in detail in the method provided in the embodiment of the present application, please refer to the corresponding contents in the above embodiment, which will not be repeated here. In addition, in addition to the above steps, the method provided in the embodiment of the present application may also include other parts or all of the steps in the above embodiments, which can be specifically referred to the corresponding contents in the above embodiments, which will not be repeated here.

[0157] It should be noted that the client mentioned above refers to the client program running on the terminal. The terminal can be a mobile terminal, a fixed terminal or a portable terminal, such as a mobile phone, a site, a unit, a device, a multimedia computer, a multimedia tablet, an Internet node, a communicator, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio broadcast receiver, an e-book device, a gaming device or any combination thereof, including accessories, peripherals or any combination thereof of these devices. It is also foreseeable that the terminal device can support any type of interface device for the user (such as a wearable device), etc. The server mentioned above refers to the service program running on the server. The server can be one or more servers. The server can also be a physical server or a virtual server, etc. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, as well as big data and artificial intelligence platforms.

[0158] Figure 5 FIG. 1 is a schematic diagram showing the structure of an electronic device provided by an embodiment of the present application. Figure 5As shown, the electronic device includes a memory 1101 and a processor 1102. The memory 1101 can be configured to store various other data to support operations on the electronic device. Examples of these data include instructions for any application or method for operating on the electronic device. The memory 1101 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read only memory (EEPROM), an erasable programmable read only memory (EPROM), a programmable read only memory (PROM), a read only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0159] The memory 1101 is used to store programs;

[0160] The processor 1102 is coupled to the memory 1101 and is used to execute the program stored in the memory 1101 to implement the methods provided by the above-mentioned method embodiments.

[0161] Further, if Figure 5 As shown, the electronic device also includes: a communication component 1103, a display 1104, a power component 1105, an audio component 1106 and other components. Figure 5 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 5 Components shown.

[0162] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a computer, the steps or functions of the methods provided in the above-mentioned method embodiments can be implemented.

[0163] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, can implement the steps or functions of the methods provided in the above-mentioned method embodiments.

[0164] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0165] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM (Read Only Memory) / RAM (Random Access Memory), a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A data processing method based on a large language model, characterized in that: include: Obtaining a problem to be analyzed and a target standard operating procedure for analyzing the problem to be analyzed; The target standard operating procedure involves multiple steps; The problem to be analyzed is an anomaly to be analyzed; Determine the target analysis tool adapted for each of the plurality of steps; the target analysis tool adapted for each step is an analysis tool used to implement the step; Generate first prompt information according to the problem to be analyzed, the target standard operating process, and the target analysis tools adapted to each of the multiple steps; The first prompt information is used to prompt the large language model to call the target analysis tools adapted to each of the multiple steps according to the target standard operating process to analyze the problem to be analyzed; Inputting the first prompt information into the large language model to obtain an analysis result of the problem to be analyzed; The analysis result includes the cause of the abnormality to be analyzed.

2. The method according to claim 1, characterized in that Generate first prompt information according to the problem to be analyzed, the target standard operating process, and the target analysis tools adapted to each of the multiple steps, including: Acquire a first prompt word template; the first prompt word template includes a task instruction related to the problem analysis task; The problem to be analyzed, the target standard operating procedure, and the target analysis tools adapted to each of the multiple steps are filled in the corresponding positions in the first prompt word template to obtain the first prompt information.

3. The method according to claim 1, characterized in that Determining target analysis tools suitable for each of the multiple steps, including: Generate second prompt information according to the problem to be analyzed, the target standard operating process, and functional description information of multiple candidate analysis tools; the second prompt information is used to prompt the large language model to determine adapted target analysis tools for the multiple steps respectively from the multiple candidate analysis tools; The second prompt information is input into the large language model to determine target analysis tools that are suitable for each of the multiple steps.

4. The method according to claim 3, characterized in that The second prompt information is used to prompt the large language model to determine adapted target analysis tools for the multiple steps from the multiple candidate analysis tools and to determine input parameters for the target analysis tools.

5. The method according to claim 4, characterized in that Generate first prompt information according to the problem to be analyzed, the target standard operating process, and the target analysis tools adapted to each of the multiple steps, including: A first prompt message is generated according to the problem to be analyzed, the target standard operating procedure, the target analysis tools adapted by each of the multiple steps, and the input parameters of the target analysis tool.

6. The method according to claim 3, characterized in that Also includes: When the large language model cannot complete the task indicated by the second prompt information, re-determine the target standard operating procedure for analyzing the problem to be analyzed; Based on the re-determined target standard operating procedure for analyzing the problem to be analyzed, new second prompt information is generated for the large language model.

7. The method according to claim 3, characterized in that The first prompt information includes historical question and answer content, and / or the second prompt information includes historical question and answer content.

8. The method according to claim 3, characterized in that Generating second prompt information according to the problem to be analyzed, the target standard operating procedure, and functional description information of multiple candidate analysis tools, including: Acquire a second prompt word template; the second prompt word template includes a task instruction related to the tool screening task; The problem to be analyzed, the target standard operating procedure, and functional description information of multiple candidate analysis tools are filled in the corresponding position in the second prompt word template to obtain second prompt information.

9. The method according to any one of claims 1 to 8, characterized in that: Obtain a target standard operating procedure for analyzing the problem to be analyzed, including: Vectorizing the problem to be analyzed to obtain a problem vector; Acquire function description vectors of multiple candidate standard operating procedures; the function description vectors of the multiple candidate standard operating procedures are obtained by vectorizing the process descriptions of the multiple candidate standard operating procedures respectively; According to the vector similarity between the problem vector and the function description vectors of the multiple candidate standard operating procedures, a target standard operating procedure for analyzing the problem to be analyzed is determined from the multiple candidate standard operating procedures.

10. The method according to claim 9, characterized in that Determining a target standard operating procedure for analyzing the problem to be analyzed from the multiple candidate standard operating procedures according to the vector similarity between the problem vector and the function description vectors of the multiple candidate standard operating procedures includes: The standard operating procedure with the largest vector similarity is determined as the target standard operating procedure for analyzing the problem to be analyzed.

11. An electronic device, characterized in that: include: A memory and a processor, wherein The memory is used to store programs; The processor is coupled to the memory, and is configured to execute the program stored in the memory to implement the method according to any one of claims 1 to 10.

12. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a computer, the method according to any one of claims 1 to 10 can be implemented.

13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.