Complaint report generation method and device, electronic equipment, storage medium and program product

Abstract

The invention provides a complaint report generation method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring a complaint report demand; calling the large model to analyze the complaint report demand to obtain a task target of the complaint report demand; determining a generative task and a check task based on the task target; calling the large model and / or the professional model to execute the generative task to obtain a complaint report required by the complaint report; calling the large model and / or the professional model to execute a verification task based on the complaint report of the complaint report demand to obtain a verification result; and determining whether to output the complaint report required by the complaint report according to the verification result. According to the embodiment of the invention, the method achieves the automatic generation of the complaint report, reduces the time and cost of manual intervention, and improves the generation efficiency and accuracy of the complaint report.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a complaint report generation method, device, electronic device, storage medium, and program product. Background Art

[0002] In the field of customer service and complaint handling, the traditional complaint handling process typically includes multiple steps, including receiving complaints, recording information, analyzing issues, taking measures, resolving issues, and generating complaint reports. Generating complaint reports is an indispensable part of the complaint handling process. It not only summarizes the entire complaint handling process but also serves as a crucial basis for providing feedback to customers, providing decision-making support to internal management, and reporting to relevant administrative departments.

[0003] However, traditional complaint report generation methods often rely on manual editing and organization, which is affected by human factors, resulting in inaccurate report content and low efficiency.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the Invention

[0005] The present disclosure provides an information sending and service deployment method, device, electronic device and storage medium, which at least to a certain extent overcome the problem of low efficiency in complaint report generation in related technologies.

[0006] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by practice of the present disclosure.

[0007] According to one aspect of the present disclosure, a method for generating a complaint report is provided, comprising:

[0008] Obtain complaint reporting requirements;

[0009] Call the big model to analyze the complaint reporting requirements and obtain the task objectives of the complaint reporting requirements;

[0010] Determine generation tasks and verification tasks based on task objectives;

[0011] Calling large models and / or specialized models to perform generative tasks and obtain complaint reports that meet reporting requirements;

[0012] Calling large models and / or specialized models to perform verification tasks based on complaint reports required by the complaint report and obtain verification results;

[0013] Determine whether to output the complaint report required based on the verification results.

[0014] In an exemplary embodiment, calling a large model and / or a specialized model to perform a generative task to obtain a complaint report meeting the complaint reporting requirements includes:

[0015] Decompose the generative task into a list of subtasks with dependencies;

[0016] For each subtask in the subtask list, call the big model to generate subtask instructions based on the prompt word;

[0017] For each subtask, according to the subtask dependency, call the large model and / or professional model to execute the subtask based on the subtask instruction and obtain the subtask execution result;

[0018] A complaint report that determines the complaint reporting requirements based on the execution results of each subtask.

[0019] In an exemplary embodiment, calling the large model and the specialized model to execute the subtask based on the subtask instruction, and obtaining the subtask execution result includes:

[0020] Call the big model to execute the subtask based on the subtask instruction and obtain the execution result of the first subtask;

[0021] Call the professional model to execute the subtask based on the subtask instruction and obtain the second subtask execution result;

[0022] The fusion model is called to fuse the execution result of the first subtask and the execution result of the second subtask to obtain the subtask execution result.

[0023] In an exemplary embodiment, the large model is called to perform a verification task based on the complaint report required by the complaint report, and the verification results obtained include:

[0024] Calling a large model to analyze the semantic similarity between complaint reports and complaint report requirements;

[0025] If the semantic similarity is greater than the first threshold, determining that the verification result is verification passed;

[0026] If the semantic similarity is less than or equal to the first threshold, the verification result is determined to be verification failure.

[0027] In one exemplary embodiment, when numerical data is included in a complaint report;

[0028] Call the professional model to perform verification tasks based on the complaint report requirements, and obtain the verification results including:

[0029] Call the professional model to calculate the data error rate based on the numerical data in the complaint report and the numerical data of the task target;

[0030] If the data error rate is less than the second threshold, determining that the verification result is verification passed;

[0031] If the data error rate is greater than or equal to the second threshold, the verification result is determined to be verification failure.

[0032] In an exemplary embodiment, determining whether to output a complaint report according to the verification result includes:

[0033] When the verification result is failure, the large model is called to adjust the generative task, and the process returns to the step of calling the large model and / or the specialized model to execute the generative task to obtain the complaint report corresponding to the complaint report requirement;

[0034] Until the verification result is passed, the complaint report required by the complaint report is output.

[0035] In an exemplary embodiment, outputting a complaint report required by a complaint report includes:

[0036] Use the template engine to generate and output complaint report documents that meet the set requirements.

[0037] According to another aspect of the present disclosure, a complaint report generating device is provided, comprising:

[0038] Demand acquisition module, used to obtain complaint report requirements;

[0039] The task target determination module is used to call the large model to analyze the complaint reporting requirements and obtain the task target of the complaint reporting requirements;

[0040] A task determination module is used to determine generation tasks and verification tasks based on task objectives;

[0041] A generative task execution module is used to call a large model and / or a specialized model to execute generative tasks and obtain a complaint report that meets the complaint reporting requirements;

[0042] A verification task execution module is used to call a large model and / or a professional model to execute verification tasks based on complaint reports required by the complaint report and obtain verification results;

[0043] The complaint report output module is used to output the complaint report required based on the verification results.

[0044] According to another aspect of the present disclosure, an electronic device is also provided, which includes: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute any one of the above-mentioned complaint report generation methods by executing the executable instructions.

[0045] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, any one of the above-mentioned complaint report generation methods is implemented.

[0046] According to another aspect of the present disclosure, a computer program product is also provided, including: a computer program or instructions, which implements any of the above-mentioned complaint report generation methods when the computer program or instructions are executed by a processor.

[0047] The complaint report generation method provided in the embodiments of the present disclosure includes: obtaining complaint report requirements; calling a large model to parse the complaint report requirements to obtain the task objectives of the complaint report requirements; determining generative tasks and verification tasks based on the task objectives; calling a large model and / or a professional model to perform generative tasks to obtain a complaint report of the complaint report requirements; calling a large model and / or a professional model to perform verification tasks based on the complaint report of the complaint report requirements to obtain verification results; and determining whether to output the complaint report of the complaint report requirements based on the verification results. In this embodiment, the complaint report requirements are parsed by calling a large model to obtain the task objectives, the task objectives are divided into generative tasks and verification tasks, and after calling a large model and / or a professional model to perform the generative tasks and verification tasks, the complaint report is output. Automatic generation of complaint reports is achieved, the time and cost of manual intervention are reduced, and the generation efficiency and accuracy of complaint reports are improved.

[0048] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0050] Figure 1 A schematic diagram of an exemplary application system architecture is shown in which the complaint report generation method according to the embodiment of the present disclosure can be applied;

[0051] Figure 2 A flowchart of a method for generating a complaint report according to an embodiment of the present disclosure is shown;

[0052] Figure 3 A flowchart showing a generative task execution process in an embodiment of the present disclosure is shown;

[0053] Figure 4A flowchart showing a subtask execution method according to an embodiment of the present disclosure is shown;

[0054] Figure 5 A schematic diagram showing a hierarchical model integration framework in an embodiment of the present disclosure is shown;

[0055] Figure 6 A flowchart showing a verification task execution process in an embodiment of the present disclosure is shown;

[0056] Figure 7 A flowchart showing another verification task execution process in an embodiment of the present disclosure;

[0057] Figure 8 A flowchart showing another method for generating a complaint report according to an embodiment of the present disclosure is shown;

[0058] Figure 9 A schematic diagram showing an intelligent agent for generating a complaint report according to an embodiment of the present disclosure is shown;

[0059] Figure 10 A schematic diagram illustrating a method for generating a complaint report based on an agent in an embodiment of the present disclosure is shown;

[0060] Figure 11 A schematic diagram of a complaint report generating device according to an embodiment of the present disclosure is shown;

[0061] Figure 12 A structural block diagram of an electronic device in an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0062] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0063] In addition, the accompanying drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0064] For ease of understanding, before introducing the embodiments of the present disclosure, several terms involved in the embodiments of the present disclosure are first explained as follows:

[0065] Large Language Models (LLMs) are neural network models trained using deep learning techniques, such as the Generative Pre-trained Transformer (GPT) series. Through multimodal recognition, they can extract intent from input information, perform reasoning based on the extracted information, and generate natural language text.

[0066] Professional models refer to algorithms or models used to handle specific tasks or belong to specific fields. Professional models can be used to connect with external systems, extract key information, submit tasks and / or work orders, etc.

[0067] Natural Language Processing (NLP) is an artificial intelligence technology that uses large or specialized models to enable computers to understand and process human language. Natural Language Processing is primarily used to understand user sentiment and to edit and verify generated reports.

[0068] An agent is a core concept in the field of artificial intelligence. It refers to an entity that can autonomously perceive information in its environment and make decisions based on that information to achieve specific goals or tasks. An agent typically consists of sensors, actuators, decision-making mechanisms, knowledge bases, and learning mechanisms.

[0069] The specific implementation of the embodiment of the present disclosure is described in detail below with reference to the accompanying drawings.

[0070] Figure 1 FIG. 1 shows an exemplary application system architecture diagram to which the complaint report generation method according to the embodiment of the present disclosure can be applied. Figure 1 As shown, the system architecture may include a terminal device 101 , a network 102 and a server 103 .

[0071] The network 102 is a medium for providing a communication link between the terminal device 101 and the server 103 , and can be a wired network or a wireless network.

[0072] Optionally, the above-mentioned wireless network or wired network uses standard communication technologies and / or protocols. The network is typically the Internet, but can also be any network, including but not limited to a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or any combination of a virtual private network). In some embodiments, technologies and / or formats including Hypertext Markup Language (HTML), Extensible Markup Language (XML), etc. are used to represent data exchanged over the network. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPSec) can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.

[0073] The terminal device 101 can be various electronic devices, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, smart speakers, smart watches, wearable devices, augmented reality devices, virtual reality devices, etc.

[0074] Optionally, the client of the application installed in different terminal devices 101 is the same, or the client of the same type of application based on different operating systems. Based on the different terminal platforms, the specific form of the client of the application can also be different, for example, the application client can be a mobile phone client, a PC client, etc.

[0075] The server 103 may be a server that provides various services, such as a background management server that provides support for the devices operated by the user using the terminal device 101. The background management server may analyze and process the received request and other data, and feed back the processing results to the terminal device.

[0076] Optionally, 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 (Content Delivery Network), and big data and artificial intelligence platforms.

[0077] Those skilled in the art will know that Figure 1 The number of terminal devices, networks, and servers in the embodiment is merely illustrative, and any number of terminal devices, networks, and servers may be provided based on actual needs. This embodiment of the present disclosure does not limit this.

[0078] Under the above system architecture, an embodiment of the present disclosure provides a method for generating a complaint report, which can be executed by any electronic device with computing and processing capabilities.

[0079] In some embodiments, the complaint report generation method provided in the embodiments of the present disclosure can be executed by the terminal device of the above-mentioned system architecture; in other embodiments, the complaint report generation method provided in the embodiments of the present disclosure can be executed by the server in the above-mentioned system architecture; in other embodiments, the complaint report generation method provided in the embodiments of the present disclosure can be implemented by the terminal device and server in the above-mentioned system architecture through interaction.

[0080] Figure 2 A flowchart of a method for generating a complaint report according to an embodiment of the present disclosure is shown as follows: Figure 2 As shown, the complaint report generation method provided by the embodiment of the present disclosure includes steps S202-S212.

[0081] The complaint report generation method provided in this embodiment can be executed by a pre-trained intelligent agent for complaint report generation. The above intelligent agent can be deployed in a management system, a work order processing system or other application platforms. By connecting to the required business support system, based on the complaint work order content, processing trajectory, processing results and attachments, a complaint report is automatically generated according to the required report template. Among them, the above business support system includes a work order system, a business system, a file system, etc. The work order system, the business system, and the file system are data collection source systems. The work order system provides work order data, the business system provides business data, and the file system provides the attachment data required for the report; each data is used to provide the corresponding field content of the report.

[0082] S202. Obtain complaint report requirements.

[0083] Complaint reporting requirements can be understood as specific requirements regarding the content, format, and purpose of complaint reports during customer service and complaint handling. These requirements primarily trigger the entire complaint reporting process and determine the direction and quality of complaint reports. For example, these requirements can be described in natural language.

[0084] In one possible implementation, complaint reporting requirements include but are not limited to: work order data, business data, format requirements, usage requirements, etc.

[0085] Work order data can be understood as detailed information about a complaint, including the handling process, results, and analysis conclusions. For example, detailed information includes the complainant's information, the time of the complaint, and the content of the complaint. The handling process includes the measures taken and the person who handled the complaint. Results include whether the issue was resolved and customer feedback. Analysis conclusions include analysis of the causes of the complaint and trend forecasts.

[0086] Business data can be understood as specific business, such as: the products purchased by users in e-commerce, purchase time, price, start time, delivery time, etc.

[0087] Format requirements include report structure, layout specifications, chart requirements, and company-specific template styles. For example, the report structure includes a cover, table of contents, main text, and appendices. Layout specifications include fonts, font sizes, and paragraph spacing. Chart requirements include chart types and data presentation methods.

[0088] Usage requirements are used to clarify the application scenarios of complaint reports, such as: providing feedback on processing results to customers, providing decision-making support to internal management of the company, reporting situations to relevant administrative departments, etc. Different purposes have different requirements for the focus and expression of complaint report content.

[0089] In one possible implementation, when operations personnel or management personnel temporarily or specifically request report generation based on actual business needs, they can manually enter the report requirements through the input terminal of the management system, including specific information such as the purpose of the report, content focus, format requirements, etc. The autonomous task executor in the intelligent agent receives the complaint report requirements input by the user.

[0090] In one possible implementation, when an enterprise needs to regularly generate fixed types of complaint reports, such as daily, weekly, or monthly reports, a scheduled task can be set to automatically trigger the complaint report acquisition process. Specifically, rules for scheduled complaint report output are preconfigured in the management system, including time periods, report types, and template requirements. The management system automatically generates complaint report requests according to the preset timing rules. After detecting the scheduled task trigger signal, the autonomous task executor obtains the complaint report request generated by the management system. This implementation reduces manual intervention and automates the entire process from request acquisition to complaint report output, improving efficiency and reducing the risk of human error.

[0091] In one possible implementation, the work order data, business data, format requirements, etc. included in the complaint requirements can be obtained from the business support system. For example, the work order data is obtained from the work order system, the business data is provided from the business system, and the attachment data or format requirements required for the report are obtained from the file system; each data is used to provide the corresponding field content of the report.

[0092] S204: Call the big model to analyze the complaint report requirement and obtain the task objective of the complaint report requirement.

[0093] Parsing can be understood as the process by which a large model uses natural language processing technology to perform semantic understanding, word segmentation, entity recognition, and logical analysis on the text of complaint report requests. Parsing is used to extract key elements from unstructured complaint report requests. Key elements include but are not limited to: report subject, data scope, format requirements, and application scenarios.

[0094] The task objective can be understood as converting the complaint report requirements into clear task instructions that can be executed by the intelligent agent based on the analysis results. The task objective may include but is not limited to: report type, content scope, format template, output object and other key parameters.

[0095] In one possible implementation, a connection is established with the big model through the interface module of the intelligent agent, a complaint report request is sent to the big model, the parsing result returned by the big model is received, and the task goal is generated based on the parsing result of the big model and combined with historical task data.

[0096] The large model analyzes complaint reporting requirements, including: splitting complaint reporting requirements described in natural language into independent words, understanding the logical relationship in the requirements through grammatical structure analysis, identifying required items and priorities, and combining complaint handling rules to convert vague complaint reporting requirements into specific task instructions.

[0097] In one possible implementation, the autonomous task executor within the agent calls the master model through a hierarchical model integration framework. Specifically, the autonomous task executor pushes complaint report requests through interface modules to the master model, NLP models, and other specialized algorithms, such as data analysis algorithms. The master model parses the semantic information of the complaint report request, the NLP model analyzes sentiment and format requirements, and the specialized algorithms determine the data scope, such as by calling the work order system to extract data for a specific time period. The fusion model then combines the outputs of the master model, NLP model, and specialized algorithm models to generate structured task objectives.

[0098] S206: Determine a generation task and a verification task based on the task goal.

[0099] Generative tasks can be understood as tasks created based on the content generation requirements of the task objectives. They are responsible for producing the substantive content of the complaint report, including but not limited to data collection, text writing, and chart generation. They rely on the text generation capabilities of large models and the data processing capabilities of specialized models to transform the task objectives into specific complaint report content.

[0100] Verification tasks can be understood as verifying the quality of complaint reports generated by generative tasks, ensuring that the content of the complaint reports meets the requirements of the task objectives. Through algorithm verification and model analysis, errors in the generated results can be reduced and the credibility of the reports can be improved.

[0101] According to the task objective parameters, it is divided into generation tasks and verification tasks; for example: if the task objective contains instructions such as "generate XX content" or "collect XX data", a generation task is generated; if the task objective contains requirements such as "verify XX accuracy" or "ensure XX consistency", a verification task is generated.

[0102] S208. Call the large model and / or professional model to perform the generative task and obtain the complaint report required.

[0103] The big model can be understood as a neural network model trained using deep learning technology. It possesses multimodal recognition, intent extraction, logical reasoning, and natural language text generation capabilities. It can perform semantic analysis on input information and generate text content. In complaint report generation, the big model is used to analyze the cause of the complaint, formulate handling suggestions, generate the table of contents, and generate chapter headings.

[0104] Professional models can be understood as algorithmic models used to process specific tasks, such as work order data extraction models, sentiment analysis models, etc., which can achieve docking with external systems, key information extraction and structured data processing.

[0105] A complaint report can be understood as the final document generated based on the complaint reporting requirements, including a summary of the complaint incident, records of the handling process, data analysis conclusions, etc.

[0106] The autonomous task executor calls the corresponding model based on the model identifier in the generative task to execute the generative task. For example, the generative tasks are "call large model A to generate the complaint report body" and "call specialized model B to extract work order data", where "large model A" and "specialized model B" are both model identifiers.

[0107] In one possible implementation, a single model call can be used. When a generative task involves only text generation or single data processing, a large model or specialized model can be called independently. For example, only the large model can be used to generate complaint handling recommendations. The autonomous task executor establishes a connection to the model through the model access component based on the model identifier in the generative task. For example, if the model identifier in the generative task is "large model A," large model A will be called to execute the generative task.

[0108] In one possible implementation, when a generative task does not include a model identifier, the large model and specialized models in the hierarchical model integration framework are simultaneously invoked to collaboratively execute the generative task. Specifically, the large model is invoked to execute the generative task to generate a first complaint report, the specialized models are invoked to execute the generative task to generate a second complaint report, and the fusion model is invoked to fuse the first and second complaint reports to obtain the complaint report that meets the reporting requirements.

[0109] S210: Call the large model and / or professional model to perform verification tasks based on the complaint report required by the complaint report to obtain verification results.

[0110] Verification tasks involve verifying the quality of generated complaint reports. Using algorithms or models, they check whether the reports meet the task objectives and set rules, and output verification results. This ensures that the report content is accurate, formatted, and logically consistent, reducing the risk of human error.

[0111] After the generative task is executed and a complaint report is generated, the autonomous task executor automatically triggers the verification task.

[0112] When the verification task is to verify semantic consistency, the semantic consistency between the complaint report content and the complaint report requirements is verified by calling the large model to ensure that key information is not missed and the conclusion is objective and accurate.

[0113] The large model is used to extract key demand words from the complaint report, calculate cosine similarity with the report text, and determine whether verification has passed based on the similarity. The large model is also used to analyze the sentiment of the report text to ensure that the complaint report meets the requirements of objectivity and neutrality.

[0114] When the verification task is data consistency verification, a professional model is called to verify the consistency between the data in the complaint report and the original work order system data to ensure that the statistics are accurate.

[0115] Call the professional model to use the CRC (Cyclical Redundancy Check) and LRC (Longitudinal Redundancy Check) algorithms to compare the report table data with the original data in the work order system line by line, and confirm whether the verification is passed based on the data error rate.

[0116] S212: Determine whether to output the complaint report required based on the verification result.

[0117] Verification results can be understood as the conclusion output after quality verification of the complaint report generated by the verification task. They are used to determine whether the complaint report meets the task objectives and set rules. Verification results can be either passed or failed. If the verification fails, the verification result also includes detailed error information. Detailed error information includes but is not limited to: text similarity value, data error rate, and location of format errors.

[0118] Outputting a complaint report can be understood as delivering the final, qualified complaint report to a designated destination. This destination includes, but is not limited to, ticket systems, management systems, business platforms, and email. The formats of the complaint report include, but are not limited to, PDF documents, Word documents, structured data, and chart files.

[0119] If the verification result is that the verification is passed, the complaint report will be pushed to the target end through the interface module. If the verification result is that the verification is failed, the complaint report will not be output.

[0120] The complaint report generation method provided in the embodiments of the present disclosure includes: obtaining complaint report requirements; calling a large model to parse the complaint report requirements to obtain the task objectives of the complaint report requirements; determining generative tasks and verification tasks based on the task objectives; calling a large model and / or a professional model to perform generative tasks to obtain a complaint report of the complaint report requirements; calling a large model and / or a professional model to perform verification tasks based on the complaint report of the complaint report requirements to obtain verification results; and determining whether to output the complaint report of the complaint report requirements based on the verification results. In this embodiment, the complaint report requirements are parsed by calling a large model to obtain the task objectives, the task objectives are divided into generative tasks and verification tasks, and after calling a large model and / or a professional model to perform generative tasks and verification tasks, the complaint report is output. Automatic generation of complaint reports is achieved, the time and cost of manual intervention are reduced, and the generation efficiency and accuracy of complaint reports are improved.

[0121] Based on the above embodiment, the process of "calling a large model and / or a professional model to perform a generative task and obtain a complaint report required by the complaint report" is optimized, such as Figure 3As shown, the optimized generative task execution process includes steps S302-S308.

[0122] S302: Decompose the generative task into a list of subtasks with dependencies.

[0123] Decomposition can be understood as splitting a complex generative task into multiple subtasks with finer granularity and more specific logic so that they can be executed and managed step by step.

[0124] Dependencies can be understood as order constraints or data transfer relationships between subtasks. This means that the output of a preceding subtask is the input of a subsequent subtask, or that a subsequent subtask cannot start until the preceding subtask completes. Dependencies include data dependencies and process dependencies. Data dependencies mean that a subsequent subtask requires the output of a preceding subtask, for example, collecting data before generating a chart. Process dependencies mean that a subsequent subtask must execute after the completion of a preceding subtask, for example, generating the report body before generating the table of contents.

[0125] A subtask list can be understood as an ordered set of subtasks formed by decomposing a generative task. Each subtask corresponds to a specific execution step and includes a description of its dependencies with other subtasks. For example, a subtask list for generating a complaint report might include: Subtask A extracts complaint data from the work order system; Subtask B uses a large model to analyze the data and generate a reason paragraph, relying on the data output of Subtask A; Subtask C uses a specialized model to generate a complaint trend chart, relying on the data output of Subtask A; and Subtask D combines text and charts to generate a complaint report, relying on the results of Subtasks B and C.

[0126] In a possible implementation, the generative task is decomposed into a list of subtasks with dependencies according to the stages of complaint report generation. For example, the subtask list includes data collection, data processing, text generation, chart generation, and content integration.

[0127] In one possible implementation, the generative task is decomposed into a list of subtasks with dependencies based on the division of labor between the large model and the specialized models. For example, the specialized model is responsible for data processing, and the large model is responsible for text generation.

[0128] S304: For each subtask in the subtask list, call the large model to generate a subtask instruction based on the prompt word.

[0129] Prompt words can be understood as natural language instructions or keywords used to guide the large model to perform specific tasks, and are used to determine task objectives, input data requirements, and output formats, etc.

[0130] Task instructions can be understood as executable instructions generated by the large model based on prompt words, which are used to guide the large model or professional model to perform subtasks.

[0131] In one possible implementation, a task instruction includes a model identifier and execution parameters. The model identifier specifies the model type and specific instance for executing the subtask. For example, the large model identifier is Large Model A, and the specialized model identifier is Work Order Data Extraction Model B. Execution parameters refer to the specific input data required by the model to execute the subtask, such as data range and format requirements. For example, the time range is April 1, 2025, to June 30, 2025, and the output format is JSON.

[0132] After the autonomous task executor decomposes the subtask list, it calls the large model to generate task instructions for each subtask. The autonomous task executor converts the subtasks into prompt words that the large model can understand, and the large model generates structured task instructions based on the prompt words.

[0133] For example: the prompt word after the subtask is "Collecting complaint data on mobile phone packages in Q2 2025" is converted to "Please generate a task instruction for collecting complaint data on mobile phone packages in the second quarter of 2025. It requires calling a professional model to extract data from the work order system with a business type of 'mobile phone package' from April 1, 2025 to June 30, 2025, and output it in JSON format." In this embodiment, the task instruction for the large model output is no longer shown.

[0134] In one possible implementation, a large model identifier is selected from a preset list based on the task type. For example, a large model identifier is selected from a preset list based on task types such as text generation and sentiment analysis. Specialized model identifiers are then matched to the specialized needs of the subtasks. For example, for a data collection task, the "Work Order Data Extraction" model identifier is selected, while for a chart generation subtask, the "Data Visualization" model identifier is selected.

[0135] Inherit parameters from the generative tasks that the subtasks depend on. In addition, the large model automatically supplements default parameters based on historical task data, for example: the default output format is JSON.

[0136] S306. For each subtask, according to the subtask dependency relationship, call the large model and / or the professional model to execute the subtask based on the subtask instruction to obtain the subtask execution result.

[0137] In one possible implementation, when a subtask instruction includes a model identifier, the model corresponding to the model identifier is called to execute the subtask and obtain the subtask execution result. The model identifier includes a large model identifier or a specialized model identifier. The large model identifier is a symbol that uniquely identifies the type or version of the large model, while the specialized model identifier is a symbol that uniquely identifies the type or version of the specialized model.

[0138] The subtask execution result refers to the output result generated after the model executes the subtask. The subtask execution result includes but is not limited to: structured data, natural language text, chart files, etc., which is an integral part of the input or report of subsequent subtasks.

[0139] The autonomous task executor uses a topological sorting algorithm to analyze the dependencies of the subtask list and generate a queue for execution. For example, if the subtask chain is "Subtask A: Data Collection → Subtask B: Data Cleaning (B) → Subtask C: Text Generation," Subtask B depends on the output of Subtask A, and Subtask C depends on the output of Subtask B. The execution order is Subtask A → Subtask B → Subtask C.

[0140] The execution status of each subtask is recorded through the cache management module, and subsequent subtasks are triggered only when the status of all preceding subtasks is "completed".

[0141] Based on the above embodiment, this embodiment optimizes the process of executing subtasks. When the subtask instruction does not include a model identifier, that is, when the model to be used is not specified, the large model and the professional model are called to work together to generate the subtask execution result. Figure 4 As shown, the optimized subtask execution steps include S402-S406.

[0142] S402: Call the large model to execute the subtask based on the subtask instruction to obtain the first subtask execution result.

[0143] The master model in the multi-layer model integration framework is called, and the execution parameters and input data in the subtask instructions are passed to the master model. The master model processes the input data based on its own algorithms and training data, and generates the execution result of the first subtask through a series of calculations and reasoning.

[0144] Among them, the multi-model integration framework stores large models of different types and different training data, such as language large models trained based on general text, multimodal large models trained based on multimodal data, etc.

[0145] S404: Call the professional model to execute the subtask based on the subtask instruction to obtain a second subtask execution result.

[0146] Invoke the specialized model in the multi-layer model integration framework. Enter the execution parameters and input data from the subtask instructions into the specialized model. Leveraging its domain-specific expertise and algorithms, the specialized model conducts in-depth analysis and processing of the data, generating the second subtask execution result.

[0147] S406 : Call the fusion model to fuse the first subtask execution result and the second subtask execution result to obtain the subtask execution result.

[0148] A fusion model refers to an algorithm or component used to integrate the output results of a large model and a specialized model. It is used to fuse multimodal results such as text, data, and charts into a unified subtask execution result for use in subsequent report generation.

[0149] Select an appropriate fusion model from the multi-layer model integration framework. Fusion models are designed based on different task types and data characteristics. Common fusion methods include weighted averaging, voting mechanisms, and deep learning fusion networks.

[0150] The execution results of the first and second subtasks are fed into the fusion model. The fusion model analyzes and integrates the two results and generates the final subtask execution results based on the set fusion rules.

[0151] For example, in the complaint cause analysis subtask of generating a complaint report, the fusion model analyzes and integrates the different cause judgments provided by the large and specialized models. Overlapping causes are retained. For general reasons proposed by the large model and industry-specific reasons proposed by the specialized models, the fusion model uses a comprehensive weighting to incorporate the more important causes into the final result. This generates a comprehensive and accurate complaint cause analysis as the result of this subtask and is used in the subsequent report writing.

[0152] In one possible implementation, the autonomous task executor generally adopts a hierarchical model integration framework to implement multi-model fusion applications, such as Figure 5 As shown, the hierarchical model integration framework includes large model 1, large model N, NPL model, and other specialized algorithms. The autonomous task manager executes subtasks in sequence and simultaneously pushes subtask instructions to the connected models through the interface module, including: large model 1, large model N, NPL model, and other specialized algorithms. The autonomous task executor will simultaneously collect the output results returned by each model and cache them in separate files; the autonomous task executor will push the cached results to the preferred large model (fusion model) for result merging; the autonomous task executor will collect the merged results (subtask execution results) and determine whether the result logic is met. If so, the result is output; if not, the result is regenerated.

[0153] This embodiment uses a large model and a professional model to collaboratively execute subtasks, and uses a fusion model to integrate the results, which can give full play to the advantages of different models and improve the effect of task execution.

[0154] S308: Determine a complaint report that meets the complaint reporting requirements based on the execution results of each subtask.

[0155] The execution result of a subtask refers to each subtask obtained by decomposing a generative task, and the output result generated after calling a large model, a professional model or a fusion model. It is the smallest component unit of the complaint report content.

[0156] Assemble the scattered sub-task results into a complete complaint report document according to the structure and format requirements of the complaint report.

[0157] In this embodiment, complex generative tasks are broken down into multiple subtasks, each of which focuses on a single goal, avoiding the efficiency bottleneck of a single model handling the entire process; the model execution order is automatically scheduled based on the subtask dependency relationship, reducing manual intervention, improving task execution efficiency, and thereby improving the efficiency of complaint report generation.

[0158] Based on the above embodiment, the method of "calling the large model to perform verification tasks based on the complaint report requirements and obtaining verification results" is optimized, such as Figure 6 As shown, the optimized verification task execution process includes steps S602-S608.

[0159] S602: Calling a large model to analyze the semantic similarity between the complaint report and the complaint report requirement.

[0160] Among them, semantic similarity is a measure of the degree of match between the complaint report content and the complaint report requirements at the semantic level. The higher the semantic similarity value, the more consistent the complaint report content is with the user's intention.

[0161] The autonomous task executor uses a large model to perform semantic understanding on the generated complaint report content. By comparing the keywords, phrases and sentences in the complaint report with the description in the complaint report requirements, it determines the semantic relevance between the two and analyzes whether their content is consistent with the preset demand goals. The semantic relevance calculation algorithm includes but is not limited to: cosine similarity, Jaccard similarity and other text similarity algorithms.

[0162] In one possible implementation, the complaint report request obtained in S202 and the complaint report generated in S208 are input into the large model through an interface module. The large model converts the complaint report request and the complaint report content into high-dimensional vectors. The cosine similarity formula is used to calculate the angle between the vectors to obtain a similarity value, which is the semantic similarity between the complaint report and the complaint report request. For example, a semantic similarity of 0.92 between the complaint report and the complaint report request indicates a 92% match.

[0163] S604: Determine whether the semantic similarity is greater than a first threshold. If so, execute S606; if not, execute S608.

[0164] The first threshold refers to a preset semantic similarity critical value, which is used to determine whether the verification is passed. Exemplarily, the first threshold is set to 90%.

[0165] S606: If the semantic similarity is greater than the first threshold, determine that the verification result is verification passed.

[0166] If the semantic similarity is greater than the first threshold, the verification result is marked as verification passed.

[0167] S608: If the semantic similarity is less than or equal to the first threshold, determine that the verification result is verification failure.

[0168] If the semantic similarity is less than or equal to the first threshold, the verification result is marked as verification failed.

[0169] In this embodiment, semantic automatic verification based on a large model ensures that complaint reports accurately match user needs, reduces manual intervention costs, and improves the efficiency and quality of report generation.

[0170] Based on the above embodiment, the method of "calling the large model to perform verification tasks based on the complaint report requirements and obtaining verification results" is optimized, such as Figure 7 As shown, the optimized verification task execution process includes steps S702-S706.

[0171] S702: Call a professional model to calculate the data error rate based on the numerical data in the complaint report and the numerical data of the task target.

[0172] Numerical data in complaint reports refers to the quantitative information contained in the complaint reports, such as the number of complaints, failure rate, and percentage, such as the failure rate in May being 0.8%. Numerical data of task objectives refers to the quantitative information in user requirements or original data sources.

[0173] The data error rate typically refers to the percentage of numerical data in a complaint report that does not meet expectations or standards. The data error rate is a key metric for measuring data accuracy and is generally calculated as the ratio of the amount of erroneous data to the total amount of data.

[0174] Incorrect data includes, but is not limited to, numerical calculation errors, unit errors, logical contradictions, missing or redundant data, etc. The total data volume refers to all numerical data points involved in the complaint report.

[0175] A specialized model is used to extract all numerical data from complaint reports and standard numerical data from task objectives. The numerical data extracted from the complaint reports are compared with the standard numerical data extracted from task objectives to determine whether the numerical data in the complaint reports contain errors. The number of valid error data is counted, and the data error rate is calculated based on the ratio of the number of valid error data to the total data.

[0176] The autonomous task executor uses CRC and LRC to traverse the table data in the complaint report, determine whether there is a deviation between the collected table data and the table data in the task description, and check whether the collected data is wrong; the main algorithms used include but are not limited to: CRC, LRC, XOR and other data verification algorithms.

[0177] S704: Determine whether the data error rate is less than a second threshold. If so, execute S706; if not, execute S708.

[0178] The second threshold is a preset error rate critical value. Exemplarily, the second threshold is 5%, which is used to determine the error condition of the data.

[0179] S706: If the data error rate is less than the second threshold, determine that the verification result is passed.

[0180] If the data error rate is less than the first threshold, the verification result is marked as verification passed.

[0181] S708: If the data error rate is greater than or equal to the second threshold, determine that the verification result is verification failure.

[0182] If the data error rate is greater than or equal to the second threshold, the verification result is marked as verification failed.

[0183] In this embodiment, the automated verification of numerical data based on professional models improves the accuracy of complaint report data, reduces manual verification costs, and improves report quality.

[0184] Based on the above embodiment, this embodiment optimizes the complaint report generation method. Figure 8 As shown, the optimized complaint report generation method provided in this embodiment mainly includes steps S802-S816.

[0185] S802. Obtain complaint report requirements.

[0186] S804: Call the big model to analyze the complaint report requirement and obtain the task objective of the complaint report requirement.

[0187] S806: Determine a generation task and a verification task based on the task goal.

[0188] S808. Call the large model and / or professional model to perform the generative task and obtain the complaint report required.

[0189] S810. Call the large model and / or professional model to perform a verification task based on the complaint report required by the complaint report to obtain a verification result.

[0190] S812: Determine whether the verification result is passed. If passed, execute S814; if not, execute S816.

[0191] If the semantic similarity is greater than the first threshold and the data error rate is less than the second threshold, the verification result is determined to be verification passed.

[0192] If the semantic similarity is greater than the first threshold or the data error rate is greater than or equal to the second threshold, it is determined that the verification result is verification failure.

[0193] S814. When the verification result is failure, call the large model adjustment generation task and return to the step of executing S808.

[0194] The list of errors that failed verification and the original task instructions are input into the large model. The large model analyzes the error types and generates corrective task instructions for different errors. For example, the corrected task instruction for a data error is "Recalculate the resolution rate, ensuring the formula is "number of resolved complaints / total number of complaints." Task adjustments can include updating subtask parameters, such as increasing the data collection scope or modifying text generation prompts. The new generative task output by the large model is received and the process returns to execute the step of "calling the large model and / or specialized models to execute the generative task and obtain the required complaint report" and subsequent steps.

[0195] The large model generates executable correction instructions, eliminating the need for manual interpretation of error reports and manual modification of task parameters, thus improving task execution efficiency.

[0196] S816: until the verification result is verified to be passed, output the complaint report required. Outputting the complaint report required includes: using a template engine to generate a complaint report document that meets the set requirements and outputting it.

[0197] A template engine is a tool that combines structured data with pre-set templates to generate formatted complaint report documents. The template engine is used to render the final complaint report from validated data according to the formatting requirements of the complaint report. Set requirements refer to user constraints on the report format and content, such as cover style, font specifications, and chapter structure.

[0198] Input the validated structured data and preset templates into the template engine, which then binds the data fields to the template placeholders. The engine generates analysis conclusions based on the template logic, calls the drawing library to generate trend charts and pie charts, sets the font, size, and alignment, and automatically generates a table of contents and chapter pagination. The output is a complaint report document that meets the specified requirements.

[0199] Using the template engine to automatically fill in data can significantly improve the standardization and automation level of complaint reports and increase the efficiency of template generation.

[0200] Based on the above embodiment, this embodiment provides an intelligent agent for generating complaint reports, such as Figure 9 As shown, the intelligent agent mainly includes five modules, namely, an interface module 910, an autonomous task executor 920, a script executor 930, a cache management module 940, and a post-processing module 950.

[0201] Interface module 910 connects the agent to external systems, including the work order system, business system, file system / platform, management system, and data resource pool. Data and file interaction is typically achieved using a RESTful (Representational State Transfer) API (Application Programming Interface) or other applicable interface protocols. Interface module 910 includes an interface management component, an information security component, a data validation component, a protocol adapter component, and a fault-tolerance control component.

[0202] The information security component is used to process interactive data encryption and decryption to ensure the information security of data and file transfer and interaction. Interactive data includes received complaint demand reports, work order data and business data obtained from the business system, task instructions and complaint report requirements input to each model, output results and complaint reports received from each model, and complaint reports sent to the target end. The data validation component verifies the data received by the interface management component to ensure the integrity and validity of the data and prevent erroneous or invalid data from entering the system. The protocol adaptation component supports multiple communication protocols and data exchange standards to facilitate communication with different types of systems or devices. The fault-tolerant control component is used to capture and handle errors during the data exchange process, providing error information and corresponding processing strategies.

[0203] In one possible implementation, the interface module is a data collection and processing module. After acquiring various interaction data, it pushes it directly to the autonomous task executor 920 and simultaneously pushes the data to the cached data module 940. The complaint report request is sent by the management system, and the work order system, business system, and file system are the data collection source systems. The work order system provides work order data, the business system provides business data, and the file system provides the attachment data required for the report. Each data is used to provide the corresponding field content of the complaint report.

[0204] Since the error handling process will affect subsequent steps, data errors will prevent subsequent steps from proceeding. Therefore, a fault-tolerant strategy is required, including re-acquisition of data, data error handling or circuit breaking; when data circuit breaking occurs, it may trigger task suspension or skipping.

[0205] The autonomous task executor 920 independently formulates and executes the task process and objectives for generating complaint reports based on predefined rules or user-defined goals. By integrating built-in executable scripts with large and specialized models, it can autonomously collect the necessary information and attachments for generating complaint reports, executing tasks without human intervention.

[0206] Among them, the specific content of the predefined rules is the rules set according to the complaint report type and the report requirements, for example: the complaint report types include; the specified method is that the operator configures the filling rules of each field according to the report requirements; the predefined rule update mechanism is manually configured by the operator.

[0207] User-defined goals refer to the requirements for the completeness of complaint report content and attachments, which serve as the evaluation criteria for complaint report usability. These goals complement predefined rules and only need to be configured once for each report, and are updated as predefined rules change.

[0208] The autonomous task executor 920 includes: a goal setting component, a model access component, a task and result recovery management component, and an autonomous iteration component.

[0209] The goal setting component is used to receive complaint report requirements input by the user. The model access component is used to achieve docking with large models and professional models through the docking interface module; the task and result recovery management component uses the cache management module 940 to build a temporary database in the server memory to store historical records, task lists, etc., to achieve context storage and make decision improvements based on this. The autonomous iteration component is used to adjust tasks based on the interactive task results and task goals to ensure that the tasks can be executed according to the goals. The self-service iteration component generates complaint reports based on complaint report requirements, and compares the complaint report requirements with the generated complaint reports to verify whether there are any gaps. The main iteration component realizes task optimization through the closed loop of "feedback collection-deviation analysis-strategy generation-execution verification". The algorithms include: deviation analysis algorithm for calculating the deviation between the results and complaint report requirements, strategy generation algorithm for rescheduling tasks, genetic algorithm for adjusting the order of tasks, etc.

[0210] The result of report A field 1 deviates from the result of the preset target. For example, in report field 1, the preset target field is the bill data of the query input parameter (March 2025), but the execution result is the bill field of the current month (May 2025). In this case, the content deviates, triggering a task adjustment.

[0211] Script Executor 930 executes the scripts returned by the autonomous task executor, including temporary database generation, data extraction, and report framework generation. Initial complaint report generation uses a queue-based approach, sending pre-set tasks to the script executor. Dynamic task updates are triggered in real time, and full task updates use a queue-based approach.

[0212] The cache management module 940 uses server memory and storage media to cache task lists, intermediate data, interaction logs, and recycled results generated during task processing. The cache management module independently decides whether to store data in memory or storage media based on the amount of stored data and call frequency.

[0213] Post-processing module 950 proofreads, edits, and formats the generated complaint report to ensure its accuracy and readability. It incorporates natural language processing technology to perform grammar checking, spelling correction, and style consistency. It also utilizes a template engine to format the complaint report, generating a document that meets specified requirements.

[0214] Based on the intelligent agent generating complaint reports, a complaint report generation method is provided, such as Figure 10 As shown, the complaint report generation method provided in this embodiment includes steps S1002-S1020.

[0215] S1002. Input requirements.

[0216] Operations personnel and management personnel may set up timed input of complaint reporting requirements.

[0217] S1004: Task analysis.

[0218] The autonomous task executor calls the large model to perform task analysis based on the input complaint report requirements.

[0219] The autonomous task executor performs parsing based on a hierarchical model integration framework, enabling multi-model applications. When multiple models are called simultaneously, they generate multi-model calculation results. Therefore, all models participate, eliminating duplicate calculations and conflicts. The outputs of each model are ultimately merged and weighted by the fusion model.

[0220] S1006. Goal setting.

[0221] After the large model returns the task analysis results, the task objectives are formed.

[0222] S1008. Task setting.

[0223] The autonomous task executor applies a large model to set tasks based on the task objectives, including generative tasks and verification tasks. Generative tasks are used to generate reports based on requirements, including content writing, data collection, data interpretation, etc. Verification tasks are used to verify the results of generative tasks with the task objectives.

[0224] The large model parses natural language requirements, extracting key elements such as time range and data dimensions and converting them into structured instructions. Tasks are then divided into generative and validation tasks. Generative tasks are then decomposed into dependent subtasks and sorted. Prompt word engineering guides the large model to generate task instructions containing execution parameters and model selection, dynamically combining the basic large model with specialized models to execute tasks. Verification rules are designed or validation logic is generated using the large model. Task parameters are adjusted based on the validation results, forming a closed-loop "parsing-generation-validation-optimization" process to ensure that tasks precisely meet the requirements. Specialized models include text classification models and chart generation models. Task parameters include retry strategies and field expansion.

[0225] S1010, task execution.

[0226] Based on the complaint report of the generative task returned by the large model, the autonomous task executor applies the cache management module and the script executor to write the complaint report into the server's memory and hard disk, including report generation, data interpretation conclusions and descriptive language generation required for other reports.

[0227] S1012. Result collection and inspection.

[0228] The autonomous task executor reads the cached complaint report of the server and applies the verification task to determine the verification result of the complaint report.

[0229] S1014, autonomous task iteration.

[0230] If the complaint report fails to pass the verification, the autonomous task executor will use the large model again to control and correct the existing task; after the correction is completed, the report generation task will be executed again.

[0231] S1016. The results are collected and verified again.

[0232] Repeat steps S1010-S1014 until the complaint report verification passes.

[0233] S1018. Post-report processing.

[0234] The autonomous task executor pushes complaint reports that have passed task verification to the post-processing module. The post-processing module uses the configured report template to conduct report standardization, including report formatting, content standardization revisions, and style standardization revisions.

[0235] S1020. Report output.

[0236] After the processing is completed, the report is output to the corresponding target end through the interface module according to the requirements.

[0237] The technical solution of this embodiment automatically generates complaint reports, reduces the time and cost of manual intervention, and improves the efficiency of complaint report generation. It reduces the risk of human error, and automated processing reduces the impact of human factors on the report generation process, reduces report quality issues caused by human negligence or errors, and improves the accuracy of complaint reports. The intelligent agent supports a variety of communication protocols and data exchange standards, and enhances compatibility and scalability with external systems. The autonomous iterative component can adjust the task process and parameters according to user feedback and task results to adapt to report generation for different subjects or different fields. The intelligent agent is versatile and can provide report generation services to the outside world by being deployed on the enterprise large model platform; it can be deployed internally on the work order platform to generate complaint reports that meet user needs at any time according to the complaint report generation requirements.

[0238] It should be noted that the acquisition, storage, use, and processing of data in the technical solution disclosed herein are in compliance with the relevant provisions of national laws and regulations. Various types of data such as personal identity data, operation data, behavioral data, etc. related to individuals, customers, and groups obtained in the embodiments of the present disclosure have been authorized.

[0239] Based on the same inventive concept, the present disclosure also provides an information sending device, such as the following embodiment. Since the principle of solving the problem in the device embodiment is similar to that in the above method embodiment, the implementation of the device embodiment can refer to the implementation of the above method embodiment, and the repeated parts will not be repeated.

[0240] Figure 11 A schematic diagram of a complaint report generating device according to an embodiment of the present disclosure is shown. The complaint report generating device is configured in an electronic device, such as Figure 11 As shown, the device includes: a demand acquisition module 1110, a task target determination module 1120, a task determination module 1130, a generation task execution module 1140, a verification task execution module 1150 and a complaint report output module 1160.

[0241] Among them, the demand acquisition module 1110 is used to obtain the complaint report demand; the task goal determination module 1120 is used to call the big model to parse the complaint report demand and obtain the task goal of the complaint report demand; the task determination module 1130 is used to determine the generative task and the verification task based on the task goal; the generative task execution module 1140 is used to call the big model and / or professional model to perform the generative task and obtain the complaint report required by the complaint report; the verification task execution module 1150 is used to call the big model and / or professional model to perform the verification task based on the complaint report required by the complaint report and obtain the verification result; the complaint report output module 1160 is used to output the complaint report required by the complaint report according to the verification result.

[0242] In an exemplary embodiment, the generative task execution module 1140 includes: a subtask decomposition unit, which is used to decompose the generative task into a subtask list with dependencies; a subtask instruction generation unit, which is used to call the big model to generate subtask instructions based on prompt words for each subtask in the subtask list; a subtask execution unit, which is used to call the big model and / or professional model to execute the subtask based on the subtask instructions according to the subtask dependencies for each subtask, and obtain the subtask execution result; and a complaint report generation unit, which is used to determine the complaint report required based on the execution results of each subtask.

[0243] In an exemplary embodiment, the subtask execution unit is specifically used to call the large model to execute the subtask based on the subtask instruction to obtain the first subtask execution result; call the professional model to execute the subtask based on the subtask instruction to obtain the second subtask execution result; call the fusion model to fuse the first subtask execution result and the second subtask execution result to obtain the subtask execution result.

[0244] In an exemplary embodiment, the verification task execution module 1150 is specifically used to call the large model to analyze the semantic similarity between the complaint report and the complaint report requirement; if the semantic similarity is greater than the first threshold, the verification result is determined to be verification passed; if the semantic similarity is less than or equal to the first threshold, the verification result is determined to be verification failed.

[0245] In an exemplary embodiment, the verification task execution module 1150 is specifically used to include numerical data in the complaint report; call a professional model to calculate the data error rate based on the numerical data in the complaint report and the numerical data of the task target; if the data error rate is less than a second threshold, the verification result is determined to be a passed verification; if the data error rate is greater than or equal to the second threshold, the verification result is determined to be a failed verification.

[0246] In an exemplary embodiment, the complaint report output module 1160 is specifically used to call the large model to adjust the generative task when the verification result is verification failure, and return to execute the step of calling the large model and / or professional model to execute the generative task to obtain the complaint report corresponding to the complaint report requirement; until the verification result is verification passed, the complaint report corresponding to the complaint report requirement is output.

[0247] In an exemplary embodiment, the complaint report output module 1160 is specifically configured to utilize a template engine to generate and output a complaint report document that meets set requirements.

[0248] It should be noted that the examples and application scenarios implemented by the modules in the above-mentioned apparatus embodiment are the same as those implemented by the corresponding steps in the method embodiment, but are not limited to the contents disclosed in the above-mentioned method embodiment. It should be noted that the above-mentioned modules, as part of the apparatus, can be executed in a computer system, such as a set of computer-executable instructions.

[0249] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation that combines hardware and software aspects, which may be collectively referred to herein as a "circuit," "module," or "system."

[0250] Based on the same inventive concept, an embodiment of the present disclosure further provides an electronic device, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute any one of the above-mentioned information sending methods or any one of the above-mentioned service deployment methods by executing the executable instructions. Since the principles for solving the problems in this electronic device embodiment are similar to those in the above-mentioned method embodiment, the implementation of this electronic device embodiment can refer to the implementation of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0251] Refer to the following Figure 12 12 is a diagram to describe the electronic device 1200 according to this embodiment of the present disclosure. Figure 12 The electronic device 1200 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0252] like Figure 12 As shown, electronic device 1200 is implemented as a general-purpose computing device. Components of electronic device 1200 may include, but are not limited to, the aforementioned at least one processing unit 1210, the aforementioned at least one storage unit 1220, and a bus 1230 connecting various system components (including storage unit 1220 and processing unit 1210).

[0253] Among them, the storage unit stores program code, and the program code can be executed by the processing unit 1210, so that the processing unit 1210 performs the steps described in the above "Exemplary Method" section of this specification according to various exemplary embodiments of the present disclosure. For example, the processing unit 1210 can perform the following steps of the above method embodiment: receiving a network generation request, wherein the network generation request carries network demand information; generating network configuration information based on the network demand information; sending the network configuration information to the distributed network so that the distributed network deploys network services based on the network configuration information. For another example: the processing unit 1210 can perform the following steps of the above method embodiment: receiving network configuration information sent by the management system, wherein the network configuration information is generated based on the network demand information, and the network demand information is carried by the network generation request; and deploying network services based on the network configuration information.

[0254] The storage unit 1220 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 12201 and / or a cache 12202 , and may further include a read-only memory unit (ROM) 12203 .

[0255] The storage unit 1220 may also include a program / utility 12204 having a set (at least one) of program modules 12205, such program modules 12205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0256] The bus 1230 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0257] Electronic device 1200 can also communicate with one or more external devices 1240 (e.g., a keyboard, pointing device, Bluetooth device, etc.), one or more devices that enable a user to interact with electronic device 1200, and / or any device that enables electronic device 1200 to communicate with one or more other computing devices (e.g., a router, modem, etc.). Such communication can occur via input / output (I / O) interface 1250. Furthermore, electronic device 1200 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via network adapter 1260. As shown, network adapter 1260 communicates with other modules of electronic device 1200 via bus 1230. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with electronic device 1200, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0258] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0259] Based on the same inventive concept, the presently disclosed embodiments further provide a computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program implements any of the aforementioned information transmission methods or any of the aforementioned service deployment methods. Because the principles for solving the problems in this computer-readable storage medium embodiment are similar to those in the aforementioned method embodiment, the implementation of this computer-readable storage medium embodiment can be referenced to the implementation of the aforementioned method embodiment, and any repetitions will not be repeated.

[0260] More specific examples of computer-readable storage media in the present disclosure may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0261] In the present disclosure, a computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, which carries readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0262] Alternatively, the program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

[0263] In a specific implementation, the program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, and the like, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0264] Based on the same inventive concept, the present disclosure also provides a computer program product, including a computer program product including: a computer program or instructions, which, when executed by a processor, implements any one of the information sending methods or service deployment methods described in the aforementioned method embodiments. Because the principles for solving the problems described in this computer program product embodiment are similar to those described in the aforementioned method embodiments, the implementation of this computer program product embodiment can refer to the implementation of the aforementioned method embodiments, and any repetitions will not be repeated.

[0265] It should be noted that although several modules or units of the device for action execution are mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0266] Furthermore, although the steps of the method of the present disclosure are described in a particular order in the accompanying drawings, this does not require or imply that the steps must be performed in this particular order, or that all steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0267] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0268] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the appended claims.

Claims

1. A method for generating a complaint report, characterized in that: include: Obtain complaint reporting requirements; Calling the big model to analyze the complaint report requirement and obtain the task objective of the complaint report requirement; Determine a generation task and a verification task based on the task goal; Calling the large model and / or specialized model to execute the generative task and obtain the complaint report required by the complaint report; Calling the large model and / or the professional model to perform the verification task based on the complaint report required by the complaint report, and obtaining a verification result; Determine whether to output the complaint report required based on the verification result.

2. The method for generating a complaint report according to claim 1, wherein: The calling of the large model and / or professional model to perform the generative task and obtain the complaint report required includes: Decomposing the generative task into a list of subtasks with dependencies; For each subtask in the subtask list, calling the large model to generate a subtask instruction based on a prompt word; For each of the subtasks, according to the subtask dependency relationship, calling the large model and / or the professional model to execute the subtask based on the subtask instruction, and obtaining the subtask execution result; Determine the complaint report required based on the execution results of each of the subtasks.

3. The method for generating a complaint report according to claim 2, wherein: The calling of the large model and the professional model to execute the subtask based on the subtask instruction, and obtaining the subtask execution result includes: Calling the large model to execute the subtask based on the subtask instruction to obtain a first subtask execution result; Calling the professional model to execute the subtask based on the subtask instruction to obtain a second subtask execution result; The fusion model is called to fuse the first subtask execution result and the second subtask execution result to obtain the subtask execution result.

4. The method for generating a complaint report according to claim 1, wherein: The calling of the large model to perform the verification task based on the complaint report required by the complaint report, and obtaining the verification result includes: Calling the large model to analyze the semantic similarity between the complaint report and the complaint report requirement; If the semantic similarity is greater than a first threshold, determining that the verification result is verification passed; If the semantic similarity is less than or equal to the first threshold, it is determined that the verification result is verification failure.

5. The method for generating a complaint report according to claim 4, wherein: When numerical data is included in said complaint report; Calling the professional model to perform the verification task based on the complaint report required by the complaint report, and obtaining the verification results including: Invoking a professional model to calculate a data error rate based on the numerical data in the complaint report and the numerical data of the task target; If the data error rate is less than a second threshold, determining that the verification result is a verification pass; If the data error rate is greater than or equal to the second threshold, it is determined that the verification result is verification failure.

6. The method for generating a complaint report according to claim 4 or 5, characterized in that: The determining whether to output the complaint report required according to the verification result includes: When the verification result is failure, calling the large model to adjust the generative task, and returning to the step of calling the large model and / or the professional model to execute the generative task to obtain the complaint report corresponding to the complaint report requirement; Until the verification result is verification passed, the complaint report required by the complaint report is output.

7. The method for generating a complaint report according to claim 6, wherein: The complaint report outputting the complaint report requirement includes: Use the template engine to generate and output complaint report documents that meet the set requirements.

8. A complaint report generating device, characterized in that: include: Demand acquisition module, used to obtain complaint report requirements; A task target determination module is used to call the large model to analyze the complaint reporting requirements and obtain the task target of the complaint reporting requirements; A task determination module, configured to determine a generation task and a verification task based on the task objective; A generative task execution module, configured to call the large model and / or the specialized model to execute the generative task and obtain the complaint report required by the complaint report; A verification task execution module, configured to call the large model and / or the specialized model to execute the verification task based on the complaint report required by the complaint report, and obtain a verification result; The complaint report output module is used to output the complaint report required according to the verification result.

9. An electronic device, characterized in that: include: processor; as well as a memory for storing executable instructions of the processor; Wherein, the processor is configured to execute the complaint report generating method described in any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the complaint report generating method according to any one of claims 1 to 7 is implemented.

11. A computer program product comprising: A computer program or instruction, characterized in that when the computer program or instruction is executed by a processor, it implements the complaint report generation method described in any one of claims 1 to 7.

Images