RPA and large model-based form question classification reply method and system, and medium

Abstract

The invention discloses a form question classification reply method and system based on RPA and a large model, and a medium, and belongs to the field of question handling, the form question classification reply method based on RPA and the large model comprises the following steps: reading each question in a question collection form by using an RPA client, extracting a question description of each question, obtaining a plurality of question descriptions corresponding to the question collection form; inputting a plurality of problem descriptions into a problem classifier to obtain a plurality of problem classification results, generating a problem classification prompt word of each problem classification result, generating a problem reason and a solution of each problem, obtaining a plurality of problem reasons and solutions of a plurality of problems, generating a problem reason core point and a solution core point of each problem, and generating a problem classification result of each problem; and generating a question answer of each question, obtaining a plurality of question answers, correspondingly sending the plurality of question answers to the acceptor in the question form, and completing question processing feedback. According to the method, the questions in the question form collected by an enterprise can be classified and answered.

Description

Technical Field

[0001] The present invention application belongs to the field of problem processing technology, and specifically relates to a method, system, and medium for classifying and answering form questions based on RPA and a large model. Background Art

[0002] Traditional enterprise problem handling usually relies on manual operations, and the process is as follows: Problem collection: manually fill out paper forms or manually enter problems in electronic forms, and submit them to relevant departments or personnel. Problem classification: manually read and understand the collected problems, and then classify them according to experience or established rules. Problem analysis and solution: for the classified problems, manually search for relevant information, analyze the causes of the problems, and develop solutions. Reply and feedback: manually organize the analysis results and solutions of the problems into text, and manually send them to the person who submitted the problem through email, instant messaging tools, etc. This traditional method has many shortcomings: first, manual operation is inefficient, especially when faced with a large number of problems, the processing speed is difficult to improve; second, manual classification and analysis are prone to deviations, resulting in unstable problem handling quality; in addition, manual replies are prone to problems such as unclear expression and lack of precision, which affects user satisfaction. Summary of the Invention

[0003] This invention provides a method, system, and medium for categorizing and answering form questions based on RPA and a large model, aiming to partially or completely resolve the existing technical issues of email reading, summarizing, and replying. To achieve the above objectives, this invention adopts the following technical solutions:

[0004] The first aspect is the form question classification and response method based on RPA and large models, including:

[0005] Step S100: Use the RPA client to read each question in the question collection form, extract the problem description of each question, and then obtain multiple problem descriptions corresponding to the question collection form;

[0006] Step S200: Multiple question descriptions are input into a question classifier, and the multiple question descriptions are classified to obtain a question classification result for each question description, thereby obtaining multiple question classification results; and based on each question classification result, a question classification prompt word is generated for each question classification result;

[0007] Step S300: Each problem description and each problem classification prompt word are associated to form a problem description classification prompt word set, thereby obtaining multiple problem description classification prompt word sets; the multiple problem description classification prompt word sets are input into a large model to generate a problem cause and solution for each problem, thereby obtaining multiple problem causes and solutions for multiple problems;

[0008] Step S400: Extract the core points of the cause of each problem and the core points of the solution, generate the core points of the cause of each problem and the solution, and obtain multiple core points of the cause of each problem and the solution; generate a question answer for each problem based on the multiple core points of the cause of each problem and the solution, and obtain multiple question answers;

[0009] Step S500: Sending multiple responses to the questions to the recipients in the question form to complete the problem handling feedback.

[0010] Optionally, step S100 includes:

[0011] Step S101: Use the RPA client to read the question form:

[0012] Step S102: Loop through each data in the question form, each data includes: proposer, acceptor, question ID, submission time and question description fields:

[0013] Step S103: extract the question description of each question plain text content, and then obtain multiple question descriptions corresponding to the question collection form.

[0014] Optionally, step S200 includes:

[0015] Step S201: Multiple problem descriptions are input into a problem classifier, and the multiple problem descriptions are classified to obtain a problem classification result for each problem description, thereby obtaining multiple problem classification results;

[0016] Step S202: Generate question classification prompt words for each question classification result based on each question classification result.

[0017] Optionally, step S202 includes:

[0018] Step S2021, defining a classification template: designing a prompt word template for each question classification, including guiding instructions;

[0019] Step S2022, matching question classification: selecting a corresponding template according to the question classification result of step S201;

[0020] Step S2023, generating prompt words: embed the question classification results into the template, and generate question classification prompt words for each question classification result. The prompt words are used to guide the large model to focus on content and direction.

[0021] Optionally, step S400 includes:

[0022] Step S401: extracting the core points of the cause of each problem and the core points of the solution, generating the core points of the cause of each problem and the core points of the solution, and obtaining multiple core points of the cause of the problem and the core points of the solution;

[0023] Step S402: Generate a question answer for each question based on multiple problem cause core points and solution core points, and obtain multiple question answers.

[0024] Optionally, step S401 includes:

[0025] Step S4011: The big model receives the cause and solution of each problem;

[0026] Step S4012: Identify key entities in the text of the cause and solution of each problem, extract keywords, extract core sentences using summary generation technology, and extract the core points of the cause and solution of the problem;

[0027] Step S4013: Output the core point of the problem cause and the core point of the solution for each problem, and generate multiple core points of the problem cause and the solution.

[0028] Optionally, in step S500, the DingTalk robot obtains the Webhook address, extracts the recipient information of each question from the question form, and constructs a text message according to the message format of the DingTalk robot. The content of the text message includes: the answer content of the question, additional information: question ID, submission time, recipient's mobile phone number or user ID, and completes the problem processing feedback.

[0029] In a second aspect, a system for classifying and answering form questions based on RPA and a large model is provided, which is used to implement any of the methods for classifying and answering form questions based on RPA and a large model described in the first aspect, including:

[0030] The problem description acquisition module uses the RPA client to read each problem in the problem collection form, extract the problem description of each problem, and then obtain multiple problem descriptions corresponding to the problem collection form;

[0031] The question classification prompt module inputs multiple question descriptions into the question classifier, classifies the multiple question descriptions, obtains the question classification results of each question description, and then obtains multiple question classification results; based on each question classification result, generates a question classification prompt word for each question classification result;

[0032] The problem cause and solution acquisition module associates each problem description with each problem classification prompt word to form a problem description classification prompt word set, thereby obtaining multiple problem description classification prompt word sets. The multiple problem description classification prompt word sets are input into the large model to generate the problem cause and solution for each problem, thereby obtaining multiple problem causes and solutions for multiple problems.

[0033] The question-answer generation module extracts the core points of the problem cause and the core points of the solution for each problem, generates the core points of the problem cause and the solution for each problem, and obtains multiple core points of the problem cause and the solution; based on the multiple core points of the problem cause and the solution, generates a question answer for each question, and obtains multiple question answers;

[0034] The problem handling feedback module sends multiple responses to the recipients in the problem form to complete the problem handling feedback.

[0035] In a third aspect, a computer-readable storage medium stores instructions, which, when executed on a computer, executes any one of the RPA and large model-based form question classification and answering methods described in the first aspect.

[0036] In a fourth aspect, a device for classifying and answering form questions based on RPA and a large model includes a memory and a processor that are communicatively connected, wherein the memory is used to store a computer program, and the processor is used to read the computer program and execute the method for classifying and answering form questions based on RPA and a large model as described in any one of the first aspects.

[0037] (1) In the present invention, the efficiency of problem handling is greatly improved by combining RPA automation process with big model. RPA can quickly read and traverse form data, and big model can quickly classify, analyze and generate responses to problems. Compared with traditional manual processing methods, the processing speed is significantly improved, and it can respond to internal enterprise problem needs in a timely manner. Big model has strong semantic understanding and analysis capabilities, can accurately classify and deeply analyze problems, and the generated response content is more accurate and professional, effectively avoiding the wrong answers caused by lack of experience or misunderstanding in manual processing, improving the quality of problem handling, and enhancing user satisfaction with the problem solving results. The entire process realizes automated operations from problem collection to response feedback without manual intervention, reducing labor costs. At the same time, the application of big model enables the system to have a certain level of intelligence, which can adapt to problems of different types and complexities, improve the intelligence level of enterprise problem handling, and provide strong support for enterprise management and development. It is built based on common RPA platform and big model technology, which is easy to integrate and deploy in the existing information environment of the enterprise, and feedback is provided through DingTalk robot, which utilizes the enterprise's commonly used instant messaging tool to facilitate users to receive and view responses.

[0038] (2) In the present invention application, firstly, RPA is used to automatically process repetitive tasks such as form data acquisition and text extraction, and combined with the semantic understanding and generation capabilities of the big model, the whole process from problem classification to reply sending is automated. The semantic understanding capabilities of the big model ensure the accuracy of problem classification, cause analysis and solution generation; in addition, the system adopts a modular design, and each module processes data in sequence to form a smooth workflow: the classification module provides guidance for cause analysis, the problem cause and problem solution module provides input for reply generation, the processing feedback module ensures that the reply is accurately delivered, the classification module ensures that the problem is correctly classified, the core point extraction removes redundant information, the reply generation optimizes the language quality, and the processing feedback module finally outputs high-quality and reliable problem replies, allowing each module to be optimized independently. The automated process greatly reduces manual processing time and reduces labor costs. The rapid execution capabilities of RPA and the big model support large-scale problem processing, shortens response time, and improves user satisfaction. The system is suitable for scenarios such as customer service, technical support, and internal feedback management. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1A flowchart of the method for classifying and answering form questions based on RPA and a large model applied for by the present invention;

[0041] Figure 2 This is a schematic diagram of the composition of the form question classification and answering system based on RPA and large models applied for by the present invention;

[0042] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments; based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] In the description of the present invention, it should be understood that the terms "upper," "lower," "front," "back," "left," "right," "top," "bottom," "inside," and "outside" and the like, indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or components referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. To further clarify the objectives, technical solutions, and advantages of the present invention, embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0045] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means one, two, or more than two, unless otherwise specifically defined.

[0046] In order to make the purpose, technical solutions and advantages of the present invention application clearer, the technical solutions in the embodiments of the present invention application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention application. Obviously, the described embodiments are only part of the embodiments of the present invention application, not all the embodiments; based on the embodiments in the present invention application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention application.

[0047] RPA Client

[0048] An RPA (Robotic Process Automation) client is a software tool or application used to implement RPA functionality. It is installed on a user's computer or server and is used to configure, run, and manage RPA software robots (bots). These robots automate repetitive, rule-driven business process tasks by simulating human operations on a graphical user interface (GUI).

[0049] The RPA client is a core component of the RPA platform. It typically serves as the user interface for interacting with the RPA system, allowing users to design automated processes, monitor robot execution, and manage task scheduling. Users install RPA client software (such as UiPath Studio or Automation Anywhere Enterprise Client) on a local computer or server and configure a connection to the RPA server or cloud platform (such as UiPath Orchestrator) to manage robots and tasks.

[0050] RPA software robots perform predefined tasks, simulating human operations within applications (such as logging into a system or filling out forms). They support two modes: Attended RPA, in which robots collaborate with humans, with user triggers or input. Unattended RPA, in which robots operate independently, executing tasks based on triggers or schedules. Task management and monitoring: A dashboard or control center is provided to monitor robot operation status, task progress, and error logs. Task scheduling is supported, allowing the robot to set execution times or trigger conditions. Integration with existing applications (such as ERP and email clients) is possible through user interface (UI) interaction or API connections.

[0051] A method for classifying and answering form questions based on RPA and large models

[0052] like Figure 1 As shown, the first aspect is the form question classification and answering method based on RPA and large models, including:

[0053] Step S100: Use the RPA client to read each question in the question collection form, extract the problem description of each question, and then obtain multiple problem descriptions corresponding to the question collection form;

[0054] Specifically, step S100 includes: using the RPA client to read the problem form storing the problem, looping through each data item in the problem form, extracting the problem description of each problem in plain text, and then obtaining multiple problem descriptions corresponding to the problem collection form;

[0055] In some embodiments, step S100 includes:

[0056] Step S101: Use the RPA client to read the question form:

[0057] In some embodiments, the RPA client (such as UiPath Studio, Automation Anywhere) serves as the entry point for automated operations. The client can be started through preset scheduling (such as daily scheduled operation) or manual triggering to connect to the problem form that stores the problem. The form can be in the form of Excel files, Google Sheets, databases, or web forms.

[0058] Step S102: Loop through each data in the question form, each data includes: proposer, acceptor, question ID, submission time and question description fields:

[0059] In some embodiments, the built-in loop function of the RPA client (such as UiPath's "For Each Row" activity) can be used to traverse each data in the form row by row. Each data typically includes: proposer, recipient, question ID, submission time, and question description fields. The RPA client can recognize and process these structured data.

[0060] Step S103: extract the problem description of each question's plain text content, and then obtain multiple problem descriptions corresponding to the question collection form;

[0061] In some embodiments, for each piece of data, the content of the "Problem Description" field is extracted through the text reading action of the RPA client (such as the "Get RowItem" activity). If the problem description contains HTML tags or rich text format, text processing functions (such as regular expressions) are used to remove non-text content to ensure that the extraction result is in plain text format.

[0062] In some embodiments, each extracted plain text problem description is stored in a variable, a list, or a data table to form a plurality of problem description sets corresponding to the problem collection form for use in subsequent processes.

[0063] In the present invention application, first, the RPA client automatically completes the form reading and data extraction tasks without human intervention, greatly improving processing efficiency. Through predefined rules and activities, it ensures that the data extraction process is consistent and error-free, avoiding errors that may be caused by human operations. The RPA client supports multiple data sources (such as Excel, CSV, databases) and can adapt to different form formats and storage methods; in addition, there is no need to change the existing system or form structure. RPA seamlessly connects to existing applications by simulating user operations, can be easily expanded to process large-scale data, supports multi-robot collaboration or parallel processing, and provides a solid foundation for problem handling processes.

[0064] Step S200: Multiple question descriptions are input into a question classifier, and the multiple question descriptions are classified to obtain a question classification result for each question description, thereby obtaining multiple question classification results; and based on each question classification result, a question classification prompt word is generated for each question classification result;

[0065] Specifically, step S200 includes:

[0066] Step S201: Multiple problem descriptions are input into a problem classifier, and the multiple problem descriptions are classified to obtain a problem classification result for each problem description, thereby obtaining multiple problem classification results;

[0067] In some embodiments, a large model (such as Grok, ChatGPT or LLaMA) is used as a problem classifier, which is implemented through API calls or local deployment. The texts of multiple problem descriptions are obtained from step S100, and prompts for classification tasks are designed for the large model, for example: "Please classify the following problem descriptions as technical problems, business process problems, or product usage problems: {problem description}", each problem description is combined with the prompt, and sent to the large model for classification through the API, and the classification results returned by the large model (such as "technical problem") are received and associated with the corresponding problem description, and the problem classification results of each problem description are saved in a data structure to form a collection of multiple problem classifications.

[0068] In some embodiments, the input is: Problem description: My software cannot start, displaying error code 1234. Prompt: Please categorize the following problem description as a technical issue, business process issue, or product usage issue: {My software cannot start, displaying error code 1234}. Large model processing: The large model analyzes the keywords "software" and "error code," outputs the classification result as "technical issue," and records it in the classification list.

[0069] Step S202, generating a question classification prompt word for each question classification result according to each question classification result;

[0070] Specifically, step S202 includes:

[0071] Step S2021, defining a classification template: designing a prompt word template for each question classification, including guiding instructions;

[0072] Step S2022, matching question classification: selecting a corresponding template according to the question classification result of step S201;

[0073] Step S2023, generating prompt words: embed the question classification results into the template, and generate question classification prompt words for each question classification result. The prompt words are used to guide the large model to focus on content and direction.

[0074] For example, input: problem category "technical problem", template: "Please analyze the following technical problem, focusing on the error code, system environment and possible technical solutions: {problem description}", generate prompt words: "Please analyze the following technical problem, focusing on the error code, system environment and possible technical solutions.

[0075] In the present invention application, firstly, the semantic understanding ability of the large model ensures the accuracy of the classification results, supports batch processing, and quickly completes the classification of a large number of questions. The classification categories can be customized according to needs and adapted to various scenarios. By using pre-trained models, the cost of manual labeling is reduced. The large model can provide a basis for classification and improve the transparency of the problem classification results. In addition, the prompt words are customized according to the classification to improve the focus of the analysis, and are automatically generated to save manual writing time. The template ensures that the prompt words have a unified style, and the template can be adjusted to meet different needs.

[0076] Step S300: Each problem description and each problem classification prompt word are associated to form a problem description classification prompt word set, thereby obtaining multiple problem description classification prompt word sets; the multiple problem description classification prompt word sets are input into a large model to generate a problem cause and solution for each problem, thereby obtaining multiple problem causes and solutions for multiple problems;

[0077] In some embodiments, multiple problem descriptions can be obtained from step S201, and prompt words corresponding to each problem classification can be obtained from step S202. Each problem description is paired with its corresponding classification prompt word to form a "problem description + prompt word" combination (i.e., a problem description classification prompt word set), ensuring that each combination contains a complete problem description and prompt words to guide the large model analysis, and the prompt word set is input into the large model.

[0078] In some embodiments, a pre-trained large language model (such as Grok, ChatGPT, or LLaMA) is used to call the model through an API or local deployment. The prompt word and problem description of each prompt word set are concatenated into a complete input. For example, please analyze the following technical problem, focusing on the error code, system environment, and possible technical solutions: The software fails to start and displays error code 1234. Multiple prompt word sets are sent to the large model in batches or in parallel to optimize processing efficiency.

[0079] In some embodiments, the large model analyzes the input and, guided by prompt words, generates text containing the cause of the problem and a solution. The output format can be structured or natural language paragraphs, depending on the prompt word design. For example, the cause of the problem: Error code 1234 typically indicates that the software is missing a necessary runtime library or that the system environment configuration is incompatible; the solution: It is recommended to check whether the system has the latest version of the runtime library installed, or try running the software in compatibility mode.

[0080] In the present invention application, the big model can deeply understand the semantics of the problem description, and combined with the guidance of the classification prompt words, generate accurate problem causes and solutions. For example, the big model can infer technical reasons such as "missing runtime library" from "the software cannot start", and the customized design of the classification prompt words makes the model output meet the needs of specific scenarios. For example, technical problem prompt words focus on error codes, and business process problem prompt words may focus on process bottlenecks. It can be expanded to support multilingual problem descriptions or more classification types to adapt to international or complex scenarios. Step S300 realizes the automated generation from problem description to cause and solution through the collaborative work of RPA and the big model, providing high-quality input for subsequent response generation.

[0081] Step S400: Extract the core points of the cause of each problem and the core points of the solution, generate the core points of the cause of each problem and the solution, and obtain multiple core points of the cause of each problem and the solution; generate a question answer for each problem based on the multiple core points of the cause of each problem and the solution, and obtain multiple question answers;

[0082] Specifically, step S400 includes:

[0083] Step S401: extracting the core points of the cause of each problem and the core points of the solution, generating the core points of the cause of each problem and the core points of the solution, and obtaining multiple core points of the cause of the problem and the core points of the solution;

[0084] In some embodiments, the parameter extractor is a tool based on natural language processing (NLP) technology that is used to extract key information from the conversation content generated by the large model. Its implementation includes the following steps:

[0085] Step S4011: The big model receives the cause and solution of each problem;

[0086] Step S4012: Identify key entities in the text of the cause and solution of each problem, extract keywords, extract core sentences through summary generation technology, and refine the core points of the cause and solution of the problem.

[0087] In some embodiments, NER recognition technology, namely named entity recognition (Named Entity Recognition), can be used to identify and classify entities with specific meanings from unstructured text. These entities are usually called "named entities". Named entities refer to specific objects or concepts in the real world, such as names of people, places, organizations, etc. Through NER recognition, computers can convert disorganized text into structured information to provide support for subsequent analysis and application, so that NER recognition can be used to identify the key entities in the text of the cause and solution of each problem. Keyword extraction can be achieved through methods such as TF (TF, Term Frequency)-IDF (IDF, Inverse Document Frequency), TextRank or BERT. Summary generation technology is divided into extractive summary and generative summary. Extractive summary directly selects core sentences from the original text, while generative summary generates new text.

[0088] Step S4013: Output the core point of the problem cause and the core point of the solution for each problem, and generate multiple core points of the problem cause and the solution.

[0089] Example: "User feedback indicates the device is running slowly, possibly due to insufficient memory. We recommend closing background programs or increasing memory."

[0090] NER recognition: "out of memory", "background program".

[0091] Keyword extraction: "device running slowly", "insufficient memory", "close background programs", "increase memory".

[0092] Text summary: "Problem cause: Insufficient memory; Solution: Close background programs or increase memory."

[0093] Output: The core cause of the problem is: "Insufficient memory", the core solution is: "Close background programs or increase memory".

[0094] Step S402: Generate a question answer for each question based on multiple problem cause core points and solution core points, and obtain multiple question answers.

[0095] In some embodiments, the core points of the cause of the problem and the core points of the solution extracted in step S401 are received, and prompts are provided to the big model. The prompts are, for example: Generate professional answers based on the following core points: Cause of the problem: {core point of the cause}; Solution: {core point of the solution}. The big model generates answers to the questions based on the prompts, and then generates multiple answers to the questions.

[0096] For example, the core point of the problem cause is "Insufficient memory", and the core point of the solution is "Close background programs or increase memory". The prompt is: "Generate a professional answer based on the following core points: Problem cause: Insufficient memory; Solution: Close background programs or increase memory." The output problem answer is: "Dear user, hello! Regarding your feedback on the slow device running issue, we analyzed that the cause may be insufficient memory. To resolve this issue, we recommend that you close background programs or increase device memory. If you need further assistance, please feel free to contact us."

[0097] In this application, first, through natural language processing (NLP) technologies such as named entity recognition (NER) and keyword extraction, S401 can accurately identify key information in the conversation, ensuring that the extracted core points are highly relevant to the cause of the problem and the solution. By automatically processing large amounts of text, the core points can be quickly extracted, significantly reducing the time for manual reading and sorting. Based on predefined rules or models, step S401 ensures the standardization and consistency of core point extraction, avoiding differences caused by human judgment. Step S401 converts unstructured text into structured data for subsequent processing and analysis. This structured output can be directly used for knowledge base construction or automatic reply systems. In addition, step S402 utilizes the generation capability of a large language model to generate natural and easy-to-understand replies, ensuring that users can easily understand them. It can generate professional and authoritative replies based on the extracted core points, thereby enhancing user trust. The generated replies use terminology and expressions that comply with industry standards to enhance credibility. The reply style can be adjusted to different user groups or scenarios, such as formal, friendly, or concise. From automatically extracting core points from the conversation content to generating professional replies, the entire process is highly automated, reducing manual intervention, reducing the time and cost of manual processing, and improving work efficiency. .

[0098] Step S500: Send the multiple answers to the questions to the recipients in the question form to complete the problem handling feedback.

[0099] In some embodiments, the DingTalk robot obtains the webhook address, extracts the recipient information of each question from the question form, usually including the recipient's DingTalk user ID or mobile phone number, uses an RPA client (such as UiPath, Automation Anywhere) or a script (such as Python) to read the question form data, ensures that each question answer is correctly associated with the corresponding recipient, and constructs a text message (text type) according to the DingTalk robot message format. The text message content includes: question answer content (from step S402), additional information: question ID, submission time, recipient's mobile phone number or user ID, and completes the problem processing feedback.

[0100] In the present invention application, the DingTalk robot automatically sends answers to questions without the need for manual replies one by one, which significantly improves feedback efficiency and supports batch sending, making it suitable for handling large-scale problem feedback scenarios. The recipient information in the form (such as mobile phone number) is used to ensure that the answer is accurately delivered to the target user. The DingTalk robot supports signature verification and custom keywords to ensure that the source of the message is credible and prevent unauthorized access. Webhook communication is based on HTTPS, and data transmission is encrypted to protect sensitive information. It provides an efficient and secure solution for problem handling feedback, improving the problem handling efficiency and user satisfaction of enterprises or teams.

[0101] like Figure 2 As shown, in the second aspect, a system for classifying and answering form questions based on RPA and a large model is used to implement any of the methods for classifying and answering form questions based on RPA and a large model described in the first aspect, including:

[0102] It should be noted that the form question classification and answering system based on RPA and big models applied for in the present invention is used to implement any of the form question classification and answering methods based on RPA and big models in the first aspect, and accordingly also includes: all the recorded technical problems, technical solutions and technical effects of any of the form question classification and answering methods based on RPA and big models in the first aspect, and the present invention application will not be repeated here.

[0103] The problem description acquisition module uses the RPA client to read each problem in the problem collection form, extract the problem description of each problem, and then obtain multiple problem descriptions corresponding to the problem collection form;

[0104] The question classification prompt module inputs multiple question descriptions into the question classifier, classifies the multiple question descriptions, obtains the question classification results of each question description, and then obtains multiple question classification results; based on each question classification result, generates a question classification prompt word for each question classification result;

[0105] The problem cause and solution acquisition module associates each problem description with each problem classification prompt word to form a problem description classification prompt word set, thereby obtaining multiple problem description classification prompt word sets. The multiple problem description classification prompt word sets are input into the large model to generate the problem cause and solution for each problem, thereby obtaining multiple problem causes and solutions for multiple problems.

[0106] The question-answer generation module extracts the core points of the cause of each problem and the core points of the solution, generates the core points of the cause of each problem and the core points of the solution, and obtains multiple core points of the cause of the problem and the core points of the solution; based on the multiple core points of the cause of the problem and the core points of the solution, generates a question answer for each question and obtains multiple question answers.

[0107] The problem handling feedback module sends multiple responses to the recipients in the problem form to complete the problem handling feedback.

[0108] In some embodiments, the problem description acquisition module 100 uses the RPA client to read each question in the problem collection form, extracts the problem description of each question, and then obtains the specific content of multiple problem descriptions corresponding to the problem collection form, which may be the same as all the technical problems, technical solutions, and technical effects recorded in step S100. The present invention application will not go into details about this.

[0109] In some embodiments, in the problem classification prompt module 200, multiple problem descriptions are input into the problem classifier, the multiple problem descriptions are classified, and the problem classification results of each problem description are obtained, and then multiple problem classification results are obtained; according to each problem classification result, the specific content of the problem classification prompt words of each problem classification result is generated, which can be the same as all the technical problems, technical solutions, and technical effects recorded in step S200, and the present invention application will not go into details about this.

[0110] In some embodiments, the problem cause and solution acquisition module 300 associates each problem description and each problem classification prompt word to form a problem description classification prompt word set, and obtains multiple problem description classification prompt word sets; multiple problem description classification prompt word sets are input into a large model to generate the problem cause and solution for each problem, and obtain the specific content of multiple problem causes and solutions for multiple problems, which can be the same as all technical problems, technical solutions, and technical effects recorded in step S300, and the present invention application will not go into details about this.

[0111] In some embodiments, the question-answer generation module 400 extracts the core points of the cause of each problem and the core points of the solution, generates the core points of the cause of each problem and the core points of the solution, and obtains multiple core points of the cause of the problem and the core points of the solution; based on the multiple core points of the cause of the problem and the core points of the solution, generates a question answer for each question, and obtains the specific content of the multiple question answers, which can be the same as all the technical problems, technical solutions, and technical effects recorded in step S400, and the present invention application will not go into details about this.

[0112] In some embodiments, the problem handling feedback module 500 sends multiple responses to the recipients in the question form to complete the specific content of the problem handling feedback, which can be the same as all the technical problems, technical solutions, and technical effects recorded in step S500. The present invention application will not go into details about this.

[0113] In the form question classification and reply system based on RPA and big models applied for by the present invention, first, RPA is used to automatically process repetitive tasks such as form data acquisition and text extraction, and combined with the semantic understanding and generation capabilities of the big model, the full process automation from question classification to reply sending is realized. The semantic understanding capability of the big model ensures the accuracy of problem classification, cause analysis and solution generation; in addition, the system adopts a modular design, and each module processes data in sequence to form a smooth workflow: the classification module provides guidance for cause analysis, the problem cause and problem solution module provides input for reply generation, the processing feedback module ensures that the reply is accurately delivered, the classification module ensures that the problem is correctly classified, the core point extraction removes redundant information, the reply generation optimizes the language quality, and the processing feedback module finally outputs high-quality and reliable question replies, allowing each module to be optimized independently. The automated process greatly reduces manual processing time and reduces labor costs. The rapid execution capabilities of RPA and big models support large-scale problem processing, shorten response time, and improve user satisfaction. The system is suitable for scenarios such as customer service, technical support, and internal feedback management.

[0114] Computer-readable storage medium

[0115] In a third aspect, the present invention application provides a computer-readable storage medium having instructions stored thereon. When the instructions are run on a computer, the form question classification and answering method based on RPA and a large model as described in any one of the first aspects is executed.

[0116] Form question classification and answering device based on RPA and big model

[0117] In fourth aspect, the present invention application provides a device for classifying and answering form questions based on RPA and large models, comprising a memory and a processor that are communicatively connected, wherein the memory is used to store a computer program, and the processor is used to read the computer program and execute the method for classifying and answering form questions based on RPA and large models as described in any one of the first aspects.

[0118] In some embodiments, an apparatus for categorizing and answering form questions based on RPA and a large model includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of any of the methods for categorizing and answering form questions based on RPA and a large model in the first aspect described above.

[0119] Those skilled in the art will understand that the schematic diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. The terminal device may include more or fewer components than shown in the diagram, or a combination of certain components, or different components. For example, the terminal device may also include input and output devices, network access devices, buses, etc.

[0120] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0121] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0122] Wherein, if the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention applies to implement all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0123] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0124] The above is a preferred embodiment of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and modifications without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A form question classification and answering method based on RPA and a large model, characterized by: include: Step S100: Use the RPA client to read each question in the question collection form, extract the problem description of each question, and then obtain multiple problem descriptions corresponding to the question collection form; Step S200: multiple problem descriptions are input into a problem classifier, the multiple problem descriptions are classified, and a problem classification result for each problem description is obtained, thereby obtaining multiple problem classification results; Generate question classification prompt words for each question classification result according to each question classification result; Step S300: Associating each question description with each question classification prompt word to form a question description classification prompt word set, thereby obtaining multiple question description classification prompt word sets; Multiple problem descriptions are classified and prompted by word sets and input into a large model to generate the cause and solution of each problem, thus obtaining multiple cause and solutions for multiple problems. Step S400: extracting the core points of the cause of each problem and the core points of the solution, generating the core points of the cause of each problem and the core points of the solution, and obtaining multiple core points of the cause of the problem and the core points of the solution; Generate answers to each question based on multiple core points of problem causes and solutions, and obtain multiple answers to the questions; Step S500: Send the multiple answers to the questions to the recipients in the question form to complete the problem handling feedback.

2. The form question classification and answering method based on RPA and large models according to claim 1 is characterized in that: Step S100 includes: Step S101: Use the RPA client to read the question form: Step S102: Loop through each data in the question form, each data includes: proposer, acceptor, question ID, submission time and question description fields: Step S103: extract the question description of each question plain text content, and then obtain multiple question descriptions corresponding to the question collection form.

3. The form question classification and answering method based on RPA and large models according to claim 2 is characterized in that: Step S200 includes: Step S201: Multiple problem descriptions are input into a problem classifier, and the multiple problem descriptions are classified to obtain a problem classification result for each problem description, thereby obtaining multiple problem classification results; Step S202: Generate question classification prompt words for each question classification result based on each question classification result.

4. The form question classification and answering method based on RPA and large models according to claim 3 is characterized in that: Step S202 includes: Step S2021, defining a classification template: designing a prompt word template for each question classification, including guiding instructions; Step S2022, matching question classification: selecting a corresponding template according to the question classification result of step S201; Step S2023, generating prompt words: embed the question classification results into the template, and generate question classification prompt words for each question classification result. The prompt words are used to guide the content and direction of the large model's attention.

5. The form question classification and answering method based on RPA and large models according to claim 4 is characterized in that: Step S400 includes: Step S401: extracting the core points of the cause of each problem and the core points of the solution, generating the core points of the cause of each problem and the core points of the solution, and obtaining multiple core points of the cause of the problem and the core points of the solution; Step S402: Generate a question answer for each question based on multiple problem cause core points and solution core points, and obtain multiple question answers.

6. The form question classification and answering method based on RPA and large models according to claim 6 is characterized in that: Step S401 includes: Step S4011: The big model receives the cause and solution of each problem; Step S4012: Identify key entities in the text of the cause and solution of each problem, extract keywords, extract core sentences using summary generation technology, and extract the core points of the cause and solution of the problem; Step S4013: Output the core point of the problem cause and the core point of the solution for each problem, and generate multiple core points of the problem cause and the solution.

7. The form question classification and answering method based on RPA and large models according to claim 1 is characterized in that: In step S500, the DingTalk robot obtains the Webhook address, extracts the recipient information of each question from the question form, and constructs a text message according to the message format of the DingTalk robot. The text message content includes: the answer content of the question, additional information: question ID, submission time, recipient's mobile phone number or user ID, and completes the problem processing feedback.

8. A system for classifying and answering form questions based on RPA and a large model, used to implement the method for classifying and answering form questions based on RPA and a large model as described in any one of claims 1 to 7, characterized in that: include: The problem description acquisition module uses the RPA client to read each problem in the problem collection form, extract the problem description of each problem, and then obtain multiple problem descriptions corresponding to the problem collection form; The problem classification prompt module inputs multiple problem descriptions into the problem classifier, classifies the multiple problem descriptions, obtains the problem classification results of each problem description, and then obtains multiple problem classification results; Generate question classification prompt words for each question classification result according to each question classification result; The problem cause and solution acquisition module associates each problem description with each problem classification prompt word to form a problem description classification prompt word set, thereby obtaining multiple problem description classification prompt word sets. The multiple problem description classification prompt word sets are input into the large model to generate the problem cause and solution for each problem, thereby obtaining multiple problem causes and solutions for multiple problems. The question-answer generation module extracts the core points of the cause of each question and the core points of the solution, generates the core points of the cause of each question and the core points of the solution, and obtains multiple core points of the cause of the problem and the core points of the solution; Generate answers to each question based on multiple core points of problem causes and solutions, and obtain multiple answers to the questions; The problem handling feedback module sends multiple responses to the recipients in the problem form to complete the problem handling feedback.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed on the computer, the form question classification and answering method based on RPA and a large model as described in any one of claims 1 to 7 is executed.

10. A device for classifying and answering form questions based on RPA and a large model, comprising a memory and a processor that are communicatively connected, wherein the memory is used to store a computer program, and the processor is used to read the computer program and execute the method for classifying and answering form questions based on RPA and a large model as described in any one of claims 1 to 7.

Images