Intelligent process approval data processing method and system based on large model and medium

Abstract

The invention discloses a process approval data intelligent processing method based on a large model, and the method comprises the steps: obtaining an approval process initiated by a user, and the approval process comprises a process form and a process attachment; generating an approved to-do transaction according to the approval process; carrying out asynchronous processing on the approved to-do affair to obtain processed data, and uploading the processed data to a large model knowledge base; obtaining a process arriving approval node, and providing a visual question and answer window; obtaining a question proposed by the user for the approval process content in the question and answer window; performing intelligent analysis on the problem by the large model, and generating an abstract required by the examination and approval decision according to the problem; and displaying the corresponding summary information required by the answer and the approval decision in the question and answer window. According to the method, a large model is adopted to intelligently analyze and summarize the content of the process form and the process attachments, summarizing and answering are performed according to questions, and abstracts required by examination and approval decision making are given, so that the difficulty and the period of process decision making are greatly reduced, and the enterprise collaborative office efficiency is integrally improved.

Description

Technical Field

[0001] The present invention relates to the field of process approval technology, and in particular to a method, system, device and medium for intelligent processing of process approval data based on a large model. Background Art

[0002] The current enterprise process approval system generally adopts the following technical architecture:

[0003] 1. Online form system: Based on the OA system, the process is electronically transferred and basic data is collected through structured form fields (such as applicant information, amount, and approval type).

[0004] 2. Rule engine driven: Use a preset rule library (such as amount thresholds and permission matrices) to automate approval decisions, and use RPA technology to trigger simple processes.

[0005] 3. Document attachment management: Manage process attachments (such as contract scans, quotations, and design drawings) in the form of file storage, relying on manual downloading, review, and verification.

[0006] 4. Statistical reporting system: Generates basic statistical indicators such as approval time and rejection rate based on process nodes, but lacks in-depth analysis capabilities.

[0007] Defects and shortcomings of existing technologies:

[0008] 1. Manual processing efficiency bottleneck: Approval personnel must verify form fields item by item. When faced with complex attachments (such as technical agreements exceeding 20 pages), the average processing time for a single process exceeds 30 minutes. This is especially true when cross-departmental collaborative approvals involve repeated verification of information, extending the decision-making cycle by over 40%.

[0009] 2. Weak processing of unstructured data: Existing systems rely on basic OCR recognition to parse PDF / image attachments and are unable to extract semantic-level information (such as special clauses in contracts and technical parameters of drawings). The omission rate of key information is as high as 32%.

[0010] 3. Lack of decision-making support capabilities: Traditional rule engines can only process clearly defined approval rules and lack the ability to dynamically judge fuzzy rules such as "reasonable travel expenses" and "appropriate entertainment standards." As a result, more than 60% of exceptions require manual processing.

[0011] 4. Rigid information interaction: The approval interface is limited to fixed fields and lacks an interactive Q&A mechanism. When an approver needs to verify whether the technical parameters on the third page of the attachment meet procurement standards, they must manually flip through the document, reducing operational efficiency by 70%.

[0012] 5. Low data asset utilization: Historical approval data isn't correlated with business performance, making it impossible to optimize approval strategies through machine learning. Inadequate multimodal processing capabilities: Existing solutions don't enable joint analysis of form structured data and unstructured attachments. For example, they can't automatically verify the consistency of invoice amounts with payment terms in contract attachments.

[0013] 6. Security and scalability deficiencies: Most systems use static permission control and lack a dynamic desensitization mechanism for sensitive information based on large models, posing a risk of data leakage. Furthermore, traditional architectures struggle to support real-time retrieval of hundreds of billions of vector data. Summary of the Invention

[0014] The purpose of the present invention is to provide a method, system, equipment and medium for intelligent processing of process approval data based on a big model. By utilizing the summary and analysis capabilities of the big model, key information in process forms and process attachments can be quickly obtained according to user questions during the approval process to generate summary information required for answers and decisions, greatly reducing the difficulty and cycle of process decision-making and improving enterprise collaboration efficiency.

[0015] The present invention is achieved through the following technical solutions:

[0016] In a first aspect, a first embodiment of the present invention provides a method for intelligently processing process approval data based on a large model, comprising:

[0017] Obtain the approval process initiated by the user, the approval process including the process form and process attachments;

[0018] Generate approval pending tasks according to the approval process;

[0019] Asynchronously process pending approval tasks to obtain processed data, and upload the processed data to the large model knowledge base;

[0020] Acquire the process and reach the approval node, and provide a visual Q&A window;

[0021] Get the questions raised by users in the Q&A window regarding the approval process content;

[0022] The big model intelligently analyzes the problem and generates the summary required for approval decision-making based on the problem;

[0023] The corresponding answers and summary information required for approval decisions are displayed in the Q&A window.

[0024] Furthermore, the specific method of generating process approval pending tasks according to the approval process includes:

[0025] Summarize the process form content into natural language to form text content, obtain the access address of the process attachment, and temporarily store the text content and access address.

[0026] Furthermore, the specific method for asynchronously processing pending approval tasks includes:

[0027] Determine whether the knowledge base has been created;

[0028] If yes, update the knowledge base metadata, where the metadata in the knowledge base corresponds to the to-do task ID;

[0029] If not, create a knowledge base;

[0030] Convert text content into text attachments;

[0031] The contents of both text attachments and process attachments are vectorized.

[0032] Furthermore, the specific method of the large model for intelligent analysis of the problem includes:

[0033] Extract text content from questions;

[0034] Segment the text content to obtain multiple text segments;

[0035] Generate a metadata set for each text segment;

[0036] Vectorize the text fragment to obtain a text vector;

[0037] The text segment, metadata and text vector are stored separately.

[0038] Furthermore, the metadata set includes basic information, semantic information, and structured tags. The basic information includes file name, document content, and page number. The semantic information includes calling a large model to generate summaries, keywords, FAQs, and entities. The structured tags include storing chapters, titles, tables, and pictures.

[0039] Furthermore, the specific method of generating the summary required for the approval decision based on the question includes:

[0040] Performing enhancement processing on the problem to obtain the enhanced problem;

[0041] Performing a dual-channel hybrid search on the enhanced question, wherein the hybrid search includes a keyword search and a vector search, and obtaining a keyword search result and a vector search result respectively;

[0042] The keyword search results and vector search results are input into the rearrangement model for comprehensive analysis, and the search results are sorted and optimized to obtain the top k text segments with the highest similarity, where k>0;

[0043] The top k text segments with the highest similarity and contextual content are input into the large model to generate the summary required for approval decision-making.

[0044] Furthermore, the specific method of enhancing the problem includes:

[0045] Rephrase user questions;

[0046] Generate a set of questions with different representations based on the large model;

[0047] Correct any typos or grammatical errors in the questions.

[0048] In a second aspect, another embodiment of the present invention provides a process approval data intelligent processing system based on a large model, comprising: a process acquisition module, a to-do generation module, an asynchronous processing module, and a question-and-answer module;

[0049] The process acquisition module is used to obtain the approval process initiated by the user, and the approval process includes a process form and a process attachment;

[0050] The to-do generation module is used to generate approval to-do tasks according to the approval process;

[0051] The asynchronous processing module is used to asynchronously process the pending approval tasks to obtain processed data, and upload the processed data to the large model knowledge base;

[0052] The question-and-answer module is used to provide a visual question-and-answer window, obtain questions raised by users on the approval process content in the question window, perform intelligent analysis on the questions, and generate the summary required for the approval decision based on the questions, and display the corresponding answers and summary information required for the approval decision in the question-and-answer window.

[0053] In a third aspect, another embodiment of the present invention provides an electronic device, comprising a processor, an input device, an output device and a memory, wherein the processor is connected to the input device, the output device and the memory respectively, the memory is used to store a computer program, and the computer program includes program instructions, and is characterized in that the processor is configured to call the program instructions to execute the method described in the first embodiment.

[0054] In a fourth aspect, another embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the method described in the first embodiment.

[0055] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0056] The embodiments of the present invention provide a method, system, device and medium for intelligent processing of process approval data based on a big model. After the process creation is initiated, the approval pending tasks are asynchronously processed and then uploaded to the big model. When the process reaches the approval node, a visual question and answer window is provided. The big model intelligently analyzes and summarizes the process form content and process attachments, summarizes and answers the questions, and also provides the summary required for approval decision-making, which greatly reduces the difficulty and cycle of process decision-making and improves the overall efficiency of enterprise collaborative office. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the examples. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be considered as limiting the scope. A person of ordinary skill in the art can also derive other relevant drawings based on these drawings without inventive effort. In the drawings:

[0058] Figure 1 A flowchart of a method for intelligently processing process approval data based on a large model provided by the first embodiment of the present invention;

[0059] Figure 2 This is an interface diagram of the visual question-and-answer window in the first embodiment of the present invention;

[0060] Figure 3 A structural block diagram of a process approval data intelligent processing system based on a large model provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0062] Example 1

[0063] like Figure 1-2 As shown, the first embodiment of the present invention provides a method for intelligently processing process approval data based on a large model, comprising the following steps:

[0064] Get the approval process initiated by the user, where the approval process includes the process form and process attachments;

[0065] Generate approval pending tasks according to the approval process;

[0066] Asynchronously process pending approval tasks to obtain processed data, and upload the processed data to the large model knowledge base;

[0067] Acquire the process and reach the approval node, and provide a visual Q&A window;

[0068] Get the questions raised by users in the Q&A window regarding the approval process content;

[0069] The big model intelligently analyzes the problem and generates the summary required for approval decision-making based on the problem;

[0070] The corresponding answers and summary information required for approval decisions are displayed in the Q&A window.

[0071] Specifically, the specific methods for generating process approval to-do tasks based on the approval process include:

[0072] Summarize the process form content into natural language to form text content, obtain the access address of the process attachment, and temporarily store the text content and access address.

[0073] The method can be executed by a processor in an electronic device. A user fills out a process form, uploads any attachments, and initiates the approval process. The processor receives the user-initiated application process, generates a pending approval task based on the approval process, summarizes the form content into natural language text, obtains the attachment's access address, and temporarily stores both the text and the access address.

[0074] Specifically, the specific methods for asynchronously processing approval pending tasks include:

[0075] Determine whether the knowledge base has been created;

[0076] If yes, update the knowledge base metadata, where the metadata in the knowledge base corresponds to the to-do task ID;

[0077] If not, create a knowledge base;

[0078] Convert text content into text attachments;

[0079] The contents of both text attachments and process attachments are vectorized.

[0080] Encapsulates the calling interface of the large model for file vectorization and dialogue: creating a knowledge base, updating the knowledge base metadata, uploading files to the knowledge base, deleting specified files in the knowledge base according to the metadata ID, and communicating with the knowledge base.

[0081] During asynchronous processing, the processor first determines whether a knowledge base has been created in the large model. If so, it updates the knowledge base metadata. The metadata in the knowledge base corresponds to the pending task ID. Based on the pending task ID, the corresponding metadata can be found, facilitating communication with the knowledge base's specified metadata file. The text content is uniformly converted into text attachments, and the content of the text attachments and process attachments is vectorized. These text attachments and process attachments are then asynchronously uploaded to the large model knowledge base.

[0082] When the process reaches the approval node, the processor provides a visual Q&A window on the pending approval page, allowing the approver to ask questions in natural language. For example, questions like "What is the budget for this application?", "What are the main terms of the contract?", or "What are the conclusions of this report?" Upon receiving the question, the processor converts the ID of the current pending task into metadata and interacts with the metadata file specified in the knowledge base. Based on the approver's question, the large model automatically performs intelligent analysis of the process content and data in the process attachments, and provides concise responses.

[0083] Intelligent analysis mainly uses document vectorization technology to convert unstructured text data into computable and searchable vector representations for subsequent efficient search and intelligent analysis. The specific process is as follows:

[0084] a) Text content extraction

[0085] The goal of text content extraction is to extract pure text information from documents of different formats and unify the data form.

[0086] Text documents (txt, docx, md, etc.): directly parse the file content and extract plain text data.

[0087] Image documents (jpg, png, pdf, etc.): Use OCR technology to perform text recognition and extract text information.

[0088] b) Document segmentation

[0089] The goal of document segmentation is to split long text into semantically coherent short segments, balancing retrieval granularity and context completeness.

[0090] The document is segmented into fixed lengths, leaving a certain overlap area between each paragraph to avoid semantic fragmentation caused by segment boundaries, thereby improving the coherence of subsequent retrieval.

[0091] A hierarchical strategy is adopted, which can be divided into natural paragraphs or sliced by sliding windows with a fixed number of words. Natural paragraphs prioritize maintaining semantic integrity, and fixed word count slicing is suitable for texts without obvious structure.

[0092] c) Generate metadata

[0093] The goal of generating metadata is to add multi-dimensional descriptive information to text fragments to enhance data retrieval and semantic understanding.

[0094] Generates a metadata set for each text fragment, including: Basic information: file name, document content, page number, etc. Semantic information: uses the large model to generate summaries, keywords, FAQs, entities, etc. Structured tags: stores relevant information such as chapters, titles, tables, images, etc. to enhance retrieval accuracy.

[0095] d) Text vectorization

[0096] The goal of text vectorization is to convert natural language into machine-computable vectors to achieve semantic similarity measurement.

[0097] Call the embedding model (bge-large-zh-v1.5, bge-m3, etc.) to convert each text fragment into a high-dimensional vector.

[0098] A dimensionality reduction optimization strategy is adopted to compress vector dimensions through algorithms such as PCA and HNSW to improve storage and retrieval efficiency.

[0099] e) Data storage

[0100] Data storage goal: Categorize and store data in different forms (text, metadata, vectors) to support efficient query and calculation.

[0101] Text fragment storage: The original text data is stored in the database and supports direct reading.

[0102] Metadata storage: Metadata is stored in a structured manner in the database for quick retrieval.

[0103] Vector storage: Text vectors are stored in vector databases (such as pgvector, FAISS, Milvus, Weaviate, etc.), supporting efficient similarity searches.

[0104] The use of multiple data storage methods ensures the structured storage of document data, making subsequent queries more accurate and efficient.

[0105] By incorporating a large model to intelligently analyze and summarize process data and attachments, approvers can quickly access key approval-related information without having to manually review each attachment or itemize the process. This not only simplifies decision-making but also significantly shortens the approval cycle. Approval personnel can focus more on core decisions rather than repeatedly reviewing lengthy documents or process details, thereby improving overall office efficiency.

[0106] When the user enters a query, the processor uses a multi-step strategy to retrieve relevant content from the knowledge base. Finally, the retrieved relevant content is contextually analyzed and summarized through a large model combined with prompt words to ensure efficient and accurate output of key information.

[0107] The specific methods for generating the summary required for approval decision based on the question include:

[0108] a) Problem Enhancement

[0109] Semantic reconstruction: Rewrite user questions to make them more suitable for vector retrieval.

[0110] Extended search: Generate a set of questions with different expressions based on the large model to improve the recall rate.

[0111] Spelling Correction: Intelligent corrections are provided for possible typos or grammatical errors.

[0112] b) Dual-channel hybrid retrieval

[0113] Keyword search path

[0114] Keyword extraction: Use a large model or entity extraction model to extract core keywords from the question.

[0115] Direct matching: Perform keyword fuzzy matching in the vector database to retrieve text fragments containing relevant keywords.

[0116] Vector search path

[0117] Vector conversion: Call the embedding model to convert the augmented question into a vector.

[0118] Vector search: Use vector similarity calculation methods (such as cosine similarity, Euclidean distance, HNSW, etc.) to match the k most similar text fragments in the database, where k>0.

[0119] c) Rearrange the results

[0120] Comprehensive analysis: Combine the results of keyword search and vector search and input them into the rearrangement model.

[0121] Deep reranking: Use the Reranker model (bge-reranker, etc.) to finely rank the search results.

[0122] d) Summary generation output: The top k text segments with the highest similarity and their context are fed into the large model. Combined with carefully designed prompts, the large model is guided to extract the core content, ensuring that the summary covers the key information in the document and generating a concise, comprehensive, and accurate summary.

[0123] When users ask questions, the system not only provides simple answers but also generates concise summaries based on the context, highlighting the key elements needed for approval decisions. For example, when an approver queries a contract appendix, the system summarizes the most important clauses, such as the contract amount, breach of contract clause, and contract term, for quick review and reference. This intelligent summary generation and summarization significantly reduces information overload for approvers during their busy workdays, enabling more efficient and accurate decision-making.

[0124] An embodiment of the present invention provides an intelligent processing method for process approval data based on a big model. After the process creation is initiated, the approval pending tasks are asynchronously processed and then uploaded to the big model. When the process reaches the approval node, a visual question and answer window is provided. The big model intelligently analyzes and summarizes the process form content and process attachments, summarizes and answers the questions, and also provides the summary required for approval decisions, which greatly reduces the difficulty and cycle of process decisions and improves the overall efficiency of enterprise collaborative office.

[0125] Example 2

[0126] like Figure 3 As shown, the second embodiment of the present invention provides a process approval data intelligent processing system based on a big model, which corresponds one-to-one to the process approval data intelligent processing method based on a big model described in the first embodiment. The system includes: a process acquisition module, a to-do generation module, an asynchronous processing module and a question-and-answer module; the process acquisition module is used to obtain the approval process initiated by the user, and the approval process includes a process form and a process attachment; the to-do generation module is used to generate approval to-do transactions according to the approval process; the asynchronous processing module is used to asynchronously process the approval to-do transactions to obtain processed data, and upload the processed data to the big model knowledge base; the question-and-answer module is used to provide a visual question-and-answer window, obtain questions raised by users on the approval process content in the question window, perform intelligent analysis on the questions, and generate the summary required for the approval decision based on the questions, and display the corresponding answers and summary information required for the approval decision in the question-and-answer window.

[0127] Among them, the execution process of each module can be executed according to the process steps of the process approval data intelligent processing method based on a large model in Example 1, and will not be repeated one by one in this embodiment.

[0128] Example 3

[0129] A third embodiment of the present invention provides an electronic device, which includes a processor, an input device, an output device and a memory. The processor is connected to the input device, the output device and the memory respectively. The memory is used to store a computer program. The computer program includes program instructions. The processor is configured to call the program instructions to execute the large model-based process approval data intelligent processing method described in the first embodiment above.

[0130] It should be understood that in the embodiments of the present invention, the processor referred to may be a central processing unit (CPU), and the processor may also be 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.

[0131] Input devices may include a touchpad, a fingerprint collection sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and output devices may include a display (LCD, etc.), a speaker, etc.

[0132] The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store information about the device type.

[0133] In specific implementations, the processor, input device, and output device described in the embodiments of the present invention can execute the implementation methods described in the method embodiments provided in the embodiments of the present invention, and can also execute the implementation methods of the system embodiments described in the embodiments of the present invention, which will not be repeated here.

[0134] Example 4

[0135] The fourth embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program. The computer program includes program instructions. When the program instructions are executed by a processor, the processor executes the large model-based process approval data intelligent processing method described in the first embodiment above.

[0136] The computer-readable storage medium may be the internal storage unit of the terminal described in the aforementioned embodiment, such as the hard disk or memory of the terminal. The computer-readable storage medium may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Furthermore, the computer-readable storage medium may also include both the internal storage unit of the terminal and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the terminal. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0137] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the composition and steps of each example according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0138] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the terminals and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0139] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or can be electrical, mechanical or other forms of connection.

[0140] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for intelligent processing of process approval data based on a large model, characterized in that: include: Obtain the approval process initiated by the user, the approval process including the process form and process attachments; Generate approval pending tasks according to the approval process; Asynchronously process pending approval tasks to obtain processed data, and upload the processed data to the large model knowledge base; Acquire the process and reach the approval node, and provide a visual Q&A window; Get the questions raised by users in the Q&A window regarding the approval process content; The big model intelligently analyzes the problem and generates the summary required for approval decision-making based on the problem; The corresponding answers and summary information required for approval decisions are displayed in the Q&A window.

2. The method for intelligent processing of process approval data based on a large model according to claim 1, characterized in that: The specific method of generating process approval to-do tasks according to the approval process includes: Summarize the process form content into natural language to form text content, obtain the access address of the process attachment, and temporarily store the text content and access address.

3. The method for intelligent processing of process approval data based on a large model according to claim 2, characterized in that: The specific method for asynchronously processing pending approval tasks includes: Determine whether the knowledge base has been created; If yes, update the knowledge base metadata, where the metadata in the knowledge base corresponds to the to-do task ID; If not, create a knowledge base; Convert text content into text attachments; The contents of both text attachments and process attachments are vectorized.

4. The method for intelligently processing process approval data based on a large model according to claim 1, characterized in that: The specific method of intelligent analysis of the problem by the large model includes: Extract text content from questions; Segment the text content to obtain multiple text segments; Generate a metadata set for each text segment; Vectorize the text fragment to obtain a text vector; The text segment, metadata and text vector are stored separately.

5. The method for intelligent processing of process approval data based on a large model according to claim 4, characterized in that: The metadata set includes basic information, semantic information, and structured tags. The basic information includes file name, document content, and page number. The semantic information includes calling a large model to generate summaries, keywords, frequently asked questions, and entities. The structured tags include storing chapters, titles, tables, and pictures.

6. The method for intelligent processing of process approval data based on a large model according to claim 4 or 5, characterized in that: The specific method of generating the summary required for the approval decision based on the question includes: Performing enhancement processing on the problem to obtain the enhanced problem; Performing a dual-channel hybrid search on the enhanced question, wherein the hybrid search includes a keyword search and a vector search, and obtaining a keyword search result and a vector search result respectively; The keyword search results and vector search results are input into the rearrangement model for comprehensive analysis, and the search results are sorted and optimized to obtain the top k text segments with the highest similarity, where k>0; The top k text segments with the highest similarity and contextual content are input into the large model to generate the summary required for approval decision-making.

7. The method for intelligently processing process approval data based on a large model according to claim 6, characterized in that: The specific method for enhancing the problem includes: Rephrase user questions; Generate a set of questions with different representations based on the large model; Correct any typos or grammatical errors in the questions.

8. A process approval data intelligent processing system based on a large model, characterized by: include: Process acquisition module, to-do generation module, asynchronous processing module and question-and-answer module; The process acquisition module is used to obtain the approval process initiated by the user, and the approval process includes a process form and a process attachment; The to-do generation module is used to generate approval to-do tasks according to the approval process; The asynchronous processing module is used to asynchronously process the pending approval tasks to obtain processed data, and upload the processed data to the large model knowledge base; The question-and-answer module is used to provide a visual question-and-answer window, obtain questions raised by users on the approval process content in the question window, perform intelligent analysis on the questions, and generate the summary required for the approval decision based on the questions, and display the corresponding answers and summary information required for the approval decision in the question-and-answer window.

9. An electronic device comprising a processor, an input device, an output device, and a memory, wherein the processor is connected to the input device, the output device, and the memory respectively, and the memory is used to store a computer program, wherein the computer program comprises program instructions, and wherein: The processor is configured to call the program instructions and execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 7.

Images