An online document intelligent operating system and operating method based on large language model

By utilizing an intelligent online document operating system based on a large language model and employing intelligent interaction and tool modules for instruction processing, the problem of cumbersome online document operations has been solved. This system enables the decomposition and sorting of natural language instructions, thereby improving the operational efficiency and user experience of online documents.

CN118504533BActive Publication Date: 2024-11-08STARRAY (BEIJING) INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410969846.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-08
Estimated Expiration
2044-07-19

AI Technical Summary

Technical Problem

Online document operations are cumbersome, requiring users to adapt to web-based operation methods, resulting in low efficiency and failing to achieve the same intuitiveness and convenience as desktop applications.

Method used

The online document intelligent operating system based on a large language model includes an intelligent interaction module, an instruction processing module, and a tool module. Through instruction processing in the initial layer, deep layer, adjustment layer, integration layer, and memory layer, it realizes the decomposition and sorting of natural language instructions. Combined with the toolset of online Excel and online Word, it provides rich editing functions.

Benefits of technology

It simplifies online document operations, improves user experience, achieves intuitive operation similar to desktop applications, enhances editing efficiency and system stability, and establishes an experience database to optimize subsequent instruction analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118504533B_ABST
    Figure CN118504533B_ABST
Patent Text Reader

Abstract

The present invention discloses an online document intelligent operating system and operation method based on a large language model, including an intelligent interaction module, an instruction processing module, and a tool module. The intelligent interaction module provides an instruction input platform; the instruction processing module identifies and processes the instructions collected by the intelligent interaction module, decomposes the complex instructions, and determines the execution order of the decomposed multiple instructions. The instruction processing module performs instruction identification processing based on the large language model; the tool module is used to call the tool based on the instruction obtained by the instruction processing module to edit the document online, and the tool module includes an online Excel work set and an online word work set. The present invention is simple to operate, and only requires to make requirements in a natural language manner to realize online editing of documents; ensure the stability and reliability of system operation; by setting a memory layer and establishing an experience database, a reference is provided for the subsequent instruction analysis process, thereby improving the system operation efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of online document processing, and in particular to an online document intelligent operating system and an operating method based on a large language model. Background Art

[0002] In today's digital, fast-paced era, the use of online documents has become an indispensable part of improving work efficiency and promoting team collaboration. It breaks the limitations of time and space, allowing team members in multiple locations and time zones to share information and edit files in real time, greatly promoting the fluency of project management and knowledge exchange.

[0003] However, when using online documents, many users will experience an operating experience that is different from that of desktop applications, the most notable of which is the cumbersome and inconvenient operation. Compared with the intuitive interface layout and familiar operating logic of desktop applications, online documents often need to be operated on web pages, which not only means that users need to frequently switch between different pages and tabs, but also requires users to spend more time and energy to search and select. In addition, since online documents usually do not support the same shortcut keys and mouse operations as desktop applications, users need to re-adapt to the new operating methods, which is undoubtedly a challenge for users who are accustomed to desktop applications. In short, compared with desktop applications, the cumbersome and inconvenient operation of online documents has become a problem that many users cannot ignore during use. Summary of the invention

[0004] In order to overcome the above problems existing in the prior art, the present invention proposes an online document intelligent operating system and operating method based on a large language model.

[0005] The technical solution adopted by the present invention to solve its technical problems is: an online document intelligent operating system based on a large language model, comprising an intelligent interaction module, an instruction processing module, and a tool module, wherein the intelligent interaction module provides an instruction input platform; the instruction processing module identifies and processes the instructions collected by the intelligent interaction module, decomposes complex instructions, and determines the execution order of the decomposed multiple instructions, and the instruction processing module performs instruction identification processing based on the large language model; the tool module is used to call a tool based on the instructions obtained by the instruction processing module to perform online editing of the document, and the tool module includes an online Excel workset and an online word workset.

[0006] The above-mentioned online document intelligent operating system based on a large language model, the instruction processing module includes an initial layer, a depth layer, an adjustment layer, an integration layer, and a memory layer. The initial layer is used to identify the instructions collected by the intelligent interaction module and pass the executable instructions to the depth layer. The depth layer is used to analyze, evaluate and split the instructions transmitted by the initial layer, and pass the split instructions to the adjustment layer; the adjustment layer determines the execution order for multiple instructions transmitted by the depth layer, feeds the results back to the depth layer, further analyzes the rationality of the instruction execution order through the depth layer, and passes the execution order to the integration layer; the integration layer is used to integrate the execution order determined by the depth layer and the adjustment layer to form the final instruction execution order; all data of the initial layer, depth layer, adjustment layer, and integration layer are passed to the memory layer, and the memory layer stores the data to form an experience database to provide a reference for the subsequent instruction analysis process.

[0007] In the above-mentioned online document intelligent operating system based on a large language model, the online Excel work set includes:

[0008] Selection operation: supports single selection, multiple selection, all selection of cells, and selection of specific rows or columns;

[0009] Data editing: directly enter text, numbers or formulas in cells and provide intelligent recommendations; delete and modify entered cell contents, and adjust the cell format and the format of the cell contents according to the instructions;

[0010] Copy and Paste: quickly copy, cut and paste row and column data, including the option to preserve formats and formulas;

[0011] Insertion and deletion: support inserting or deleting rows and columns, as well as batch operations;

[0012] Move operation: provide drag and drop function for cells, rows and columns;

[0013] Sorting operation: Arrange data in ascending or descending order;

[0014] Data validation function: set rules for input data;

[0015] Find and replace function: search the table for content that matches the information provided by the instruction, and replace the found content;

[0016] Basic counting: automatically calculate the sum, average, maximum and minimum values ​​in the selected area;

[0017] Conditional statistics: conditional judgment and statistics are performed through conditional statistical functions;

[0018] Data visualization: Display data in different forms according to instruction requirements.

[0019] In the above-mentioned online document intelligent operating system based on a large language model, the online word work set includes:

[0020] Intelligent writing assistant: Use language models to automatically generate article paragraphs, summaries or suggestions;

[0021] Real-time collaborative editing: allows multiple users to edit documents online at the same time and view editing traces and comments in real time;

[0022] Style template: preset multiple document style templates, and apply unified format with one click;

[0023] Page layout: supports page margin, paper size, orientation adjustment, as well as header, footer, and page number insertion;

[0024] Automatic table of contents generation: Automatically generate a table of contents based on the heading levels.

[0025] An operating method of an online document intelligent operating system based on a large language model, based on the above-mentioned online document intelligent operating system based on a large language model, specifically includes:

[0026] Step 1: The user inputs instructions through the intelligent interaction module;

[0027] Step 2, the instruction processing module identifies and processes the instruction obtained in step 1. When the instruction obtained in step 1 is a compound instruction, the compound instruction is decomposed into at least two basic instructions, and the execution steps of the at least two basic instructions are determined;

[0028] Step 3: Based on the basic instructions obtained in step 2 and the execution steps of the at least two basic instructions, a tool module is called to edit the online document.

[0029] In the above-mentioned method for operating the online document intelligent operating system based on a large language model, step 2 specifically includes:

[0030] Step 2.1, the initial layer recognizes the instructions in step 1 and passes the executable instructions to the deep layer;

[0031] Step 2.2, the depth layer analyzes and evaluates the instruction transmitted by the initial layer to determine whether it is a compound instruction. If it is a compound instruction, the compound instruction is split and at least two basic instructions after the split are passed to the adjustment layer, and step 2.3 is executed; if it is a basic instruction, the basic instruction is directly passed to the adjustment layer, and step 2.4 is executed;

[0032] Step 2.3, the adjustment layer analyzes at least two basic instructions transmitted by the depth layer, determines the initial execution order of at least two basic instructions, and feeds back the result to the depth layer. The depth layer analyzes the initial execution order to determine whether the compound instruction can be implemented. If it cannot be implemented, the result is fed back to the adjustment layer. The adjustment layer readjusts the initial execution order and feeds back the adjustment result to the depth layer. The above process is repeated until the target execution order that can implement the compound instruction is obtained, and the target execution order is passed to the integration layer. The integration layer outputs the target execution order;

[0033] Step 2.4, the adjustment layer analyzes the basic instructions passed by the depth layer, feeds the results back to the depth layer, and passes the basic instructions to the integration layer, which outputs the target execution order;

[0034] Step 2.5, pass steps 2.1-2.4 to the memory layer, the memory layer stores the instruction processing process, forms an experience database, and provides a reference for the subsequent instruction analysis process.

[0035] The beneficial effects of the present invention are that the present invention is simple to operate and only requires making requirements in natural language to achieve online document editing; by setting adjustment layers and depth layers, instructions are automatically disassembled and sorted, and the execution order is continuously adjusted to ensure the stability and reliability of system operation; by setting a memory layer, an experience database is established to provide a reference for subsequent instruction analysis processes, thereby improving system operation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0037] Figure 1 This is a system architecture diagram of the present invention;

[0038] Figure 2 This is a framework diagram of the instruction processing module of the present invention;

[0039] Figure 3 This is a relationship diagram of the main modules of the system of the present invention;

[0040] Figure 4 This is a schematic diagram of the first step thinking process of the AI ​​Agent in Example 1 of the present invention;

[0041] Figure 5 This is a schematic diagram of the last step of the thinking process of the AI ​​Agent in Example 1 of the present invention. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0043] This embodiment discloses an online document intelligent operating system based on a large language model. Figure 1 As shown, it includes an intelligent interaction module, an instruction processing module, and a tool module. The intelligent interaction module provides an instruction input platform; the instruction processing module recognizes and processes the instructions collected by the intelligent interaction module, decomposes complex instructions, and determines the execution order of multiple decomposed instructions. The instruction processing module performs instruction recognition processing based on a large language model; the tool module is used to call the tool based on the instructions obtained by the instruction processing module to edit the document online, and the tool module includes an online Excel workset and an online word workset.

[0044] ①Online Excel toolset

[0045] (a) Basic operations

[0046] Selection operations: Supports single selection, multiple selection, all selection of cells, and selection of specific rows or columns.

[0047] Data Editing: Allows users to enter text, numbers or formulas directly into cells and provides intelligent recommendations. Allows users to delete and modify entered cell contents. Also adjusts the cell format and the format of the cell contents according to user requirements.

[0048] Copy and Paste: Quickly copy, cut, and paste row and column data, including the option to preserve formatting and formulas.

[0049] Insertion and deletion: Supports inserting or deleting rows and columns, as well as batch operations, to maintain the integrity of the data structure.

[0050] Move operation: Provides drag-and-drop movement of cells, rows, and columns to facilitate data reorganization.

[0051] Sorting operation: Arrange data in ascending or descending order, and can be sorted based on a single column or multiple columns.

[0052] Data validation function: Set rules for data input, such as data type, value range, text length, etc. to ensure the accuracy of data input.

[0053] Find and replace function: You can search the table for content that matches the information provided by the user, and replace the found content.

[0054] (b) Statistical functions

[0055] Basic counting: Automatically calculate the sum, average, maximum, minimum, etc. of the values ​​in the selected area.

[0056] Conditional statistics: implements conditional judgment and statistics such as IF functions, SUMIF, COUNTIF, and supports complex conditional data filtering.

[0057] Data visualization: Display data in different forms such as histograms, pivot tables, etc. according to user requirements.

[0058] ②Online Word Toolset

[0059] (a) Text generation and editing

[0060] Intelligent Writing Assistant: Uses language models to automatically generate article paragraphs, summaries or suggestions to help users quickly complete the first draft.

[0061] Real-time collaborative editing: allows multiple users to edit documents online at the same time, view editing traces and comments in real time, and improve teamwork efficiency.

[0062] (b) Formatting and layout

[0063] Style templates: Preset a variety of document style templates, such as title, body, citation, etc., and apply a unified format with one click.

[0064] Page layout: supports page margin, paper size, orientation adjustment, as well as header, footer, and page number insertion.

[0065] Automatic table of contents generation: Automatically generate a table of contents based on the heading levels to facilitate online document navigation and structure management.

[0066] The above are atomic tools. Composite tools can call several atomic tools and composite tools. AI Agent calls these tools to implement corresponding online document operation functions.

[0067] like Figure 2 As shown, the instruction processing module includes an initial layer, a depth layer, an adjustment layer, an integration layer, and a memory layer. The initial layer is used to identify the instructions collected by the intelligent interaction module and pass the executable instructions to the depth layer. The depth layer is used to analyze, evaluate and split the instructions passed by the initial layer, and pass the split instructions to the adjustment layer; the adjustment layer determines the execution order for multiple instructions passed by the depth layer, feeds the results back to the depth layer, further analyzes the rationality of the instruction execution order through the depth layer, and passes the execution order to the integration layer; the integration layer is used to integrate the execution order determined by the depth layer and the adjustment layer to form the final instruction execution order; all data from the initial layer, depth layer, adjustment layer, and integration layer are passed to the memory layer, and the memory layer stores the data to form an experience database to provide a reference for the subsequent instruction analysis process.

[0068] The intelligent operating system of this embodiment is based on a large language model to build a command processing module. In this embodiment, the command processing module is represented by multiple AI agents, and each layer of the command processing module is an AI agent. Users interact with multiple AI agents through communication tools (such as QQ, WeChat), and multiple AI agents will operate the corresponding online documents to complete user requirements.

[0069] Users complete task publishing by sending messages in QQ groups. The system will monitor messages in QQ groups through go-cqhttp and collect and send messages to multiple AI agents. Multiple AI agents will analyze and solve problems according to the division of labor between modules. Finally, the results will be notified to users by @ in the QQ group.

[0070] Figure 3 This is a diagram showing the relationship between the main modules of the online document intelligent operating system. The user is the user of the system and provides the command information required by the system. The QQ robot is a tool for communicating with users, which is equivalent to the external interface of the system. The AI ​​Agent is the operating logic of the system and is the subject of the control tool call. Querying the information in the online document is mainly responsible for viewing the information content in the online document. Editing the information in the online document is mainly responsible for adding, deleting, and changing the information content in the online document. Joint table editing of the information in the online document is mainly responsible for editing the information in the other table by querying the information in one of the two online tables. Generating the information in the online document is mainly to generate the corresponding information based on the user description and fill it in the online document. The whole process will be interactive in the form of QQ or WeChat.

[0071] This embodiment also discloses an operation method of an online document intelligent operating system based on a large language model, which specifically includes:

[0072] Step 1: The user inputs instructions through the intelligent interaction module;

[0073] Step 2: The instruction processing module identifies and processes the instruction obtained in step 1. When the instruction obtained in step 1 is a compound instruction, the compound instruction is decomposed into at least two basic instructions, and the execution steps of the at least two basic instructions are determined.

[0074] Step 2 specifically includes:

[0075] Step 2.1, the initial layer recognizes the instructions in step 1 and passes the executable instructions to the deep layer;

[0076] Step 2.2, the depth layer analyzes and evaluates the instruction transmitted by the initial layer to determine whether it is a compound instruction. If it is a compound instruction, the compound instruction is split and at least two basic instructions after the split are passed to the adjustment layer, and step 2.3 is executed; if it is a basic instruction, the basic instruction is directly passed to the adjustment layer, and step 2.4 is executed;

[0077] Step 2.3, the adjustment layer analyzes at least two basic instructions transmitted by the depth layer, determines the initial execution order of at least two basic instructions, and feeds back the result to the depth layer. The depth layer analyzes the initial execution order to determine whether the compound instruction can be implemented. If it cannot be implemented, the result is fed back to the adjustment layer. The adjustment layer readjusts the initial execution order and feeds back the adjustment result to the depth layer. The above process is repeated until the target execution order that can implement the compound instruction is obtained, and the target execution order is passed to the integration layer. The integration layer outputs the target execution order;

[0078] Step 2.4, the adjustment layer analyzes the basic instructions passed by the depth layer, feeds the results back to the depth layer, and passes the basic instructions to the integration layer, which outputs the target execution order;

[0079] Step 2.5, pass steps 2.1-2.4 to the memory layer, the memory layer stores the instruction processing process, forms an experience database, and provides a reference for the subsequent instruction analysis process.

[0080] Step 3: Based on the basic instructions obtained in step 2 and the execution steps of the at least two basic instructions, a tool module is called to edit the online document. Example 1

[0081] The language model of this example is GPT-4, and the social application is QQ.

[0082] Follow the prompts below to build an AI Agent

[0083] You are a powerful AI assistant that can automate problem solving using tools and commands.

[0084] You must follow the following constraints to complete the task. {Constraints}

[0085] Your task is: {task description}

[0086] If this task displays "None", "Gone", "Completed" or similar expressions, you can directly input FINISH in the following tool.

[0087] All the files you need are in the following directory: {working directory}

[0088] You can use the following tools or commands, {tools}

[0089] First, output your thought process according to the following format instructions:

[0090] {Thought Process}

[0091] The constraints in the AI ​​Agent are described as follows:

[0092] ①. Use only one tool for each decision, and you can use it any number of times.

[0093] ②. Make sure the command called or the tool used is in the given tool list.

[0094] ③. If all tasks have been completed, make sure to end with the "FINISH" command.

[0095] ④. If the execution of a command or tool fails, try to change the parameters or parameter format and call it again.

[0096] ⑤. The generated response must follow the factual information given above and cannot fabricate information.

[0097] ⑥ Ensure that the generated actions can be accurately executed. The actions can include specific methods and target outputs.

[0098] The thinking process in AI Agent can be described as follows:

[0099] Key concept: The combined concepts or entities involved in the task, i.e., compound instructions.

[0100] Concept decomposition: Decompose the key concepts in the task into a series of sub-elements to be queried, that is, split the compound instructions into multiple basic instructions.

[0101] (1) Analyze the dependencies between elements, for example:

[0102] i. Do we need to obtain the value / definition of A before we can obtain B through A?

[0103] ii. More thoughts can be expanded like this.

[0104] (2) Based on the above analysis, arrange the query priorities among sub-elements.

[0105] (3) Find the sub-element that currently needs to obtain a value.

[0106] Different AI Agents will have different descriptions of tasks in AI Agents to ensure that each AI Agent does only one thing and improve the task completion rate. For example, for the initial layer AI Agent, the description is as follows: Your task is to identify instructions based on the user's input and pass executable instructions to the deep layer. For the deep layer AI Agent, the description is as follows: Your task is to analyze and evaluate the instructions passed by the initial layer to determine whether they are compound instructions. If they are compound instructions, split the compound instructions and pass at least two basic instructions after splitting to the adjustment layer. At the same time, analyze the execution order passed by the adjustment layer to determine whether the compound instructions can be implemented. If not, feed the results back to the adjustment layer. For the adjustment layer AI Agent, the description is as follows: Your task is to analyze at least two basic instructions passed by the deep layer, determine the initial execution order of at least two basic instructions, and feed the results back to the deep layer. For the integration layer AI Agent: The description is as follows, your task is to integrate all analysis results and variants to form a final solution.

[0107] The user starts the batch file in go-cqhttp to realize data transmission with the QQ robot. By setting the group information to be monitored, the URL and header are set through the same address as go-cqhttp to obtain the information in the QQ group. The latest message and the current message are obtained through the res_get_error method for judgment. If they are the same, rest for 3 seconds and then judge again; if they are different, the first 10 lines of content are obtained through the getdoct_URL method to obtain the URL of the Tencent document to be modified, and the res_separate_mes method is used to determine whether the message is @test robot operation. If not, skip this step. If so, the following content and the URL are transmitted to the AI_get_answer method to call the AI ​​Agent. After obtaining the answer, the answer is sent to the QQ group through the robot through the answer_post method. When the user wakes up the AI ​​Agent in the @robot way (such as "@robot, fill in the phone number of each row of the student's grade correctly in the joint table"), the thinking process (part) of the AI ​​Agent is as follows Figure 4 When the AI ​​Agent finishes thinking, the action it takes is FINISH, which means that it has successfully completed the user's task, as shown in Figure 5 As shown, the result is sent to the user in the form of @user.

[0108] The above embodiments are only exemplary embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art may make various modifications or equivalent substitutions to the present invention within the essence and protection scope of the present invention, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present invention.

Claims

1. An online document intelligent operating system based on a large language model, characterized in that: It includes an intelligent interaction module, an instruction processing module, and a tool module. The intelligent interaction module provides an instruction input platform; the instruction processing module recognizes and processes the instructions collected by the intelligent interaction module, decomposes complex instructions, and determines the execution order of multiple instructions decomposed, and the instruction processing module performs instruction recognition processing based on a large language model; the tool module is used to call a tool based on the instructions obtained by the instruction processing module to edit the document online, and the tool module includes an online Excel workset and an online word workset; The instruction processing module includes an initial layer, a depth layer, an adjustment layer, an integration layer, and a memory layer. The initial layer is used to identify the instructions collected by the intelligent interaction module and pass the executable instructions to the depth layer. The depth layer is used to analyze, evaluate and split the instructions passed by the initial layer, and pass the split instructions to the adjustment layer; the adjustment layer determines the execution order for multiple instructions passed by the depth layer, feeds the results back to the depth layer, further analyzes the rationality of the instruction execution order through the depth layer, and passes the execution order to the integration layer; the integration layer is used to integrate the execution order determined by the depth layer and the adjustment layer to form the final instruction execution order; all data of the initial layer, the depth layer, the adjustment layer, and the integration layer are passed to the memory layer, and the memory layer stores the data to form an experience database to provide a reference for the subsequent instruction analysis process; The operation method of using the online document intelligent operating system based on the large language model is as follows: Step 1: The user inputs instructions through the intelligent interaction module; Step 2, the instruction processing module identifies and processes the instruction obtained in step 1. When the instruction obtained in step 1 is a compound instruction, the compound instruction is decomposed into at least two basic instructions, and the execution steps of the at least two basic instructions are determined; Step 3, based on the basic instructions obtained in step 2 and the execution steps of the at least two basic instructions, calling a tool module to edit the online document; The step 2 specifically includes: Step 2.1, the initial layer recognizes the instructions in step 1 and passes the executable instructions to the deep layer; Step 2.2, the depth layer analyzes and evaluates the instruction transmitted by the initial layer to determine whether it is a compound instruction. If it is a compound instruction, the compound instruction is split and at least two basic instructions after the split are passed to the adjustment layer, and step 2.3 is executed; if it is a basic instruction, the basic instruction is directly passed to the adjustment layer, and step 2.4 is executed; Step 2.3, the adjustment layer analyzes at least two basic instructions transmitted by the depth layer, determines the initial execution order of at least two basic instructions, and feeds back the result to the depth layer. The depth layer analyzes the initial execution order to determine whether the compound instruction can be implemented. If it cannot be implemented, the result is fed back to the adjustment layer. The adjustment layer readjusts the initial execution order and feeds back the adjustment result to the depth layer. The above process is repeated until the target execution order that can implement the compound instruction is obtained, and the target execution order is passed to the integration layer. The integration layer outputs the target execution order; Step 2.4, the adjustment layer analyzes the basic instructions passed by the depth layer, feeds the results back to the depth layer, and passes the basic instructions to the integration layer, which outputs the target execution order; Step 2.5, pass steps 2.1-2.4 to the memory layer, the memory layer stores the instruction processing process, forms an experience database, and provides a reference for the subsequent instruction analysis process.

2. The online document intelligent operating system based on a large language model according to claim 1, characterized in that: The online Excel workbook includes: Selection operation: supports single selection, multiple selection, all selection of cells, and selection of specific rows or columns; Data editing: directly enter text, numbers or formulas in cells and provide intelligent recommendations; delete and modify entered cell contents, and adjust the cell format and the format of the cell contents according to the instructions; Copy and Paste: quickly copy, cut and paste row and column data, including the option to preserve formats and formulas; Insertion and deletion: support inserting or deleting rows and columns, as well as batch operations; Move operation: provide drag and drop function for cells, rows and columns; Sorting operation: Arrange data in ascending or descending order; Data validation function: set rules for input data; Find and replace function: search the table for content that matches the information provided by the instruction, and replace the found content; Basic counting: automatically calculate the sum, average, maximum and minimum values ​​in the selected area; Conditional statistics: conditional judgment and statistics are performed through conditional statistical functions; Data visualization: Display data in different forms according to instruction requirements.

3. The online document intelligent operating system based on a large language model according to claim 1, characterized in that: The online word work set includes: Intelligent writing assistant: Use language models to automatically generate article paragraphs, summaries or suggestions; Real-time collaborative editing: allows multiple users to edit documents online at the same time and view editing traces and comments in real time; Style template: preset multiple document style templates, and apply unified format with one click; Page layout: supports page margin, paper size, orientation adjustment, as well as header, footer, and page number insertion; Automatic table of contents generation: Automatically generate a table of contents based on the heading levels.