Intelligent prompting method and system, computer device and storage medium

Abstract

The application discloses an intelligent prompting method and system, a computer device and a storage medium. The method comprises the following steps: receiving a user-defined professional label and a knowledge bias label, and configuring a data collection path; collecting and analyzing multi-dimensional behavior data of the user on a personal computer in real time to generate structured knowledge items; starting a cleaning process during a computer idle period or at a fixed time every day, using a large model to judge the novelty and relevance of the knowledge, and storing the knowledge items determined by the judgment in a knowledge base; in the knowledge base, automatically obtaining vertical field frontier knowledge from the network based on the user label and generating an abstract through a large model, and periodically deleting low-frequency obsolete knowledge; capturing user input content in real time, and combining historical behavior data and the knowledge base to perform RAG retrieval to provide intelligent prompts. The embodiments of the application aim to solve the problems of the existing knowledge base system, such as static, insufficient user adaptability and low interaction efficiency.

Description

Technical Field

[0001] This application relates to an intelligent prompting method, system, computer device, and storage medium, belonging to the fields of artificial intelligence and personal computers. Background Technology

[0002] With the popularization of Artificial Intelligence Personal Computer (AIPC) technology, users' personalized needs for human-computer interaction are becoming increasingly prominent. Current AI assistants or knowledge base systems for personal computers are mainly geared towards general scenarios, lacking the ability to deeply perceive and analyze individual user behaviors (such as search habits, document processing preferences, and information acquisition patterns). Specific problems include:

[0003] Static knowledge base: It relies on a pre-set general knowledge system and cannot absorb personalized knowledge generated by users during computer use (such as high-frequency operation procedures, industry-specific terminology, custom work templates, etc.).

[0004] Insufficient user adaptability: The knowledge filtering and generation process does not take into account users' professional characteristics (such as designers, copywriters, finance professionals, etc.) and device usage habits (such as commonly used software types and file storage structures), resulting in a low degree of matching between recommended content and actual needs.

[0005] Limitations of interactive scenarios: In high-frequency scenarios such as document editing and information retrieval, it is impossible to provide dynamic intelligent prompts based on the user's historical behavior, requiring users to manually repeat operations or search for information across platforms, which is inefficient. Summary of the Invention

[0006] In view of this, this application provides an intelligent prompting method, system, computer device, and storage medium. The embodiments of this application aim to solve the problems of static nature, insufficient user adaptability, and low interaction efficiency of existing knowledge base systems. It proposes a method for dynamic knowledge capture, intelligent generation, and contextualized prompting based on user computer operation behavior to build a highly personalized AIPC-specific knowledge system.

[0007] The first aspect of this application discloses a smart suggestion method, the method comprising:

[0008] Receive user-defined occupational tags and knowledge bias tags, and configure the data collection path;

[0009] Real-time collection and analysis of multi-dimensional user behavior data on personal computers to generate structured knowledge entries;

[0010] The cleaning process is started during computer downtime or at a fixed time each day. The large model is used to determine the novelty and relevance of knowledge, and knowledge items that pass the determination are stored in the knowledge base.

[0011] In the knowledge base, cutting-edge knowledge in vertical fields is automatically obtained from the Internet based on user tags and a large model is used to generate summaries, while low-frequency and outdated knowledge is regularly deleted;

[0012] It captures user input in real time and combines historical behavior data with a knowledge base to perform RAG retrieval to provide intelligent suggestions.

[0013] Furthermore, the data collection path includes search behavior, browsing search results, screenshot operations, document reading, keyboard input, and software usage logs.

[0014] Furthermore, the real-time capture of user input, combined with historical behavior data and a knowledge base for RAG retrieval to provide intelligent suggestions, includes:

[0015] In interactive scenarios, user input is captured in real time, and intelligent prompts related to the current task context are generated based on the knowledge base and historical behavior data.

[0016] The intelligent prompts are presented in the form of interactive fragments and can be directly inserted into the current working interface by triggering events;

[0017] The interactive scenarios include one of them: document editing, information retrieval, and software operation;

[0018] The triggering events include one of the following: keyboard shortcuts and mouse operations.

[0019] A second aspect of this application discloses an intelligent prompting system, the system comprising:

[0020] The user tag and data path configuration module is used to receive user-defined occupation tags and knowledge bias tags, and configure the data collection path;

[0021] The multi-dimensional behavioral data real-time acquisition and parsing module is connected to the user tag and data path configuration module, and is used to collect and parse the user's multi-dimensional behavioral data on the personal computer in real time to generate structured knowledge entries.

[0022] The dynamic knowledge cleaning and intelligent storage module is connected to the multi-dimensional behavioral data real-time acquisition and analysis module. It is used to start the cleaning process during computer idle periods or at fixed times every day, use a large model to judge the novelty and relevance of knowledge, and store the knowledge items that pass the judgment into the knowledge base.

[0023] The knowledge base dynamic expansion and anti-fixation module is connected to the dynamic knowledge cleaning and intelligent storage module. It is used to automatically obtain cutting-edge knowledge in vertical fields from the network based on user tags and generate summaries through a large model, as well as to periodically delete low-frequency outdated knowledge.

[0024] The contextualized intelligent suggestion service module is connected to the knowledge base dynamic expansion and anti-fixation module. It is used to capture user input in real time and perform RAG retrieval in combination with historical behavior data and knowledge base to provide intelligent suggestions.

[0025] Furthermore, the multi-dimensional behavioral data real-time acquisition and analysis module includes:

[0026] The search behavior capture module is used to capture keywords from the browser or local search box and call the vertical domain search engine to obtain result summaries.

[0027] The document reading and parsing module is used to parse the titles, charts, and annotations in a document to extract key information.

[0028] The screenshot OCR parsing module is used to recognize text and tables in screenshots to extract useful information;

[0029] The keyboard input high-frequency phrase statistics module is used to count high-frequency input patterns and generate a user-specific phrase library;

[0030] The software uses a log parsing module to record and parse the operation steps of commonly used software.

[0031] Furthermore, the dynamic knowledge cleaning and intelligent storage module includes:

[0032] The RAG pre-retrieval module is used to match the knowledge to be cleaned with the existing knowledge base to mark duplicate content.

[0033] The large model judgment module is used to filter content that can be retained based on local or remote large models.

[0034] The tag weight dynamic adjustment module is used to automatically adjust the weight of user tags based on newly added knowledge.

[0035] Furthermore, there are multiple occupational tags, including a first occupational tag and a second occupational tag, and the tag weight dynamic adjustment module includes:

[0036] The occupation type label determination module is used to determine occupation type labels based on newly added knowledge.

[0037] The weight dynamic execution module is used to increase the weight of the first occupation label if the occupation type label is the first occupation label, and at the same time, perform non-negative attenuation on the weight of the second occupation label according to the preset attenuation coefficient; if the occupation type label includes the first occupation label and the second occupation label, the weights of the first occupation label and the second occupation label remain unchanged; wherein, the weight change of any occupation label is restricted to a preset weight range.

[0038] Furthermore, the occupational type label determination module includes:

[0039] The dual keyword extraction module is used to automatically extract two types of keywords from the input knowledge content text using natural language processing technology: domain keywords and action keywords;

[0040] The tag comparison module is used to compare the extracted domain keywords and action keywords with a predefined occupational tag dictionary, which contains multiple occupational tags, each of which is associated with a specific set of keywords;

[0041] The tag matching module is used to identify a job tag as a job type tag if a keyword matches a certain job tag; if the keyword does not match any job tag, the job type tag is marked as uncategorized and manual review is triggered.

[0042] A third aspect of this application discloses a computer-readable storage medium comprising a stored program, wherein the program, when running, controls the execution of the intelligent prompting method of the above embodiments in the processor of the device.

[0043] A fourth aspect of this application discloses a computer device, the computer device including a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and executed by the above-described intelligent prompting method.

[0044] Compared with the prior art, the embodiments of this application have the following beneficial effects:

[0045] 1) This embodiment significantly improves the accuracy of knowledge recommendation by fusing tag weights with behavioral data;

[0046] 2) This embodiment realizes the automatic capture of new knowledge every day, which improves the timeliness of the knowledge base content and can dynamically reflect the latest work needs of users;

[0047] 3) This embodiment provides intelligent prompts in office software, reducing the number of steps required for users. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0049] Figure 1 A flowchart of an intelligent prompting method provided in an embodiment of this application.

[0050] Figure 2 This is a structural block diagram of an intelligent prompting system provided in an embodiment of this application.

[0051] Figure 3 This is a structural block diagram of a multi-dimensional behavioral data real-time acquisition and analysis module provided in an embodiment of this application.

[0052] Figure 4 This is a structural block diagram of a dynamic knowledge cleaning and intelligent data entry module provided in an embodiment of this application.

[0053] Figure 5 This is a structural block diagram of a tag weight dynamic adjustment module provided in an embodiment of this application.

[0054] Figure 6 This is a structural block diagram of a job type label determination module provided in an embodiment of this application. Detailed Implementation

[0055] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0056] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0057] Example 1:

[0058] I. Local Deployment Solution:

[0059] Technical Implementation: Deployment on the AIPC side using open-source large models (such as DeepSeek-R1, moonlight-16B-A3B-Instruct) combined with localized tools (Ollama, Flowy, LM Studio). For example, on an Intel Core Ultra platform, models can be run on a local server using Ollam, or offline access can be achieved using Flowy with the help of the Radiant GPU for inference.

[0060] limitation:

[0061] Lagging knowledge updates: It relies on users to manually import or periodically crawl the network to update the knowledge base, and cannot capture dynamic behavioral data of users on the computer in real time (such as keywords searched temporarily, content of newly created documents).

[0062] Personalization is limited to a single dimension: it only adapts to users by adjusting model parameters, lacks integrated analysis of multi-dimensional behavioral data (search, browsing, input, screenshots, etc.), and makes it difficult to build a user-specific knowledge graph.

[0063] Limited scenario coverage: It mainly focuses on dialogue interaction and cannot provide context-aware intelligent services in complex scenarios such as document processing and software control, based on the user's historical operation trajectory.

[0064] II. End-to-End Cloud Deployment Solution:

[0065] Technical implementation: It adopts a collaborative model of "edge-side inference + cloud resources", such as "Xiaoshuo Knows" in cooperation with ASUS. It provides functions such as PC control and local knowledge base by flexibly calling edge-cloud models. It supports inference using edge processor when offline and expands cloud services (such as AI painting and hardware control) when connected to the network.

[0066] limitation:

[0067] Insufficient dynamic adaptability: User behavior data is only used for switching model invocation strategies (such as prioritizing the use of client or cloud models), and is not deeply used for real-time updates and personalized generation of knowledge bases. Knowledge recommendation is still mainly based on general content.

[0068] Lack of interaction depth: In scenarios such as document editing and information retrieval, dynamic prompts (such as high-frequency phrase completion and related document recommendations) are not generated based on user input history and knowledge base content, and users still need to actively trigger the search.

[0069] To address at least one of the aforementioned problems, embodiments of this application propose the following solutions:

[0070] Figure 1 A flowchart illustrating an intelligent suggestion method provided in an embodiment of this application. Figure 1 As shown, the method includes:

[0071] 101 receives user-defined occupational tags and knowledge bias tags, and configures the data collection path.

[0072] In this step, the data collection path includes search behavior, browsing search results, screenshot operations, document reading, keyboard input, and software usage logs.

[0073] Specifically, as a user, you can customize occupational tags (such as designer, finance, and copywriter) and knowledge-biased tags (such as graphic design, financial reporting, and new media writing). The weight of these tags can be dynamically adjusted based on user habits and behavioral data during the knowledge filtering and generation process. Simultaneously, you can configure data collection paths, covering search behavior, browsing search results, screenshot operations, document reading (supporting PDF, Word, or Excel formats), keyboard input (such as high-frequency phrases and paragraphs), and software usage logs (such as operation steps for Adobe tools).

[0074] 102. Real-time collection and analysis of multi-dimensional behavioral data of users on personal computers to generate structured knowledge entries.

[0075] Example 1) Capture keywords from the browser or local search box (such as poster design color scheme), and call the vertical search engine in the background to get the results and generate a summary (such as the key points of design website tutorials).

[0076] Example 2) Parse the document content, extract key information from the title, charts, and annotations (such as the formula logic in the financial statements and the time nodes in the planning proposal), and generate "knowledge point-tag" related records.

[0077] Specific example: Title: Financial Statements for the Second Quarter of 2025;

[0078] chart:

[0079] Revenue trend chart: Shows monthly revenue growth in the second quarter of 2025;

[0080] Cost analysis table: lists the percentage and trend of each cost item;

[0081] annotation:

[0082] Formula logic: Revenue = Sales Revenue - Cost;

[0083] Key metric: Net profit = Revenue - Taxes and fees.

[0084] Example 3) Use OCR technology to recognize text and tables in screenshots and extract valid information (such as parameters in drawings and data charts in web pages).

[0085] Example 4) Analyze high-frequency input patterns (such as meeting minutes templates and expense reimbursement processes in office scenarios) and generate a user-specific phrase library.

[0086] Specific Example: At the monthly sales summary meeting on July 22, 2025, attendees included Zhang San (Sales Manager), Li Si (Marketing Specialist), and Wang Wu (Sales Representative). The meeting was chaired by Zhao Liu, with Qian Qi taking notes. The meeting time was 9:00-11:00 AM in the company's main conference room. The meeting focused on the previous month's sales performance, discussing regional sales data, market feedback, and customer complaint handling. The discussion resulted in the development of specific improvement plans for regions with unsatisfactory sales data, clarification of the next steps in market promotion strategies, and optimization of the customer complaint handling process. The next steps and responsibilities were as follows: Zhang San was responsible for adjusting the overall sales strategy, Li Si for developing the market promotion plan, and Wang Wu for following up on customer complaint handling. In the meeting summary, Zhao Liu emphasized the importance of teamwork and required all departments to actively advance their work according to their assigned responsibilities. In this passage, the user used the following phrases from the phrase library:

[0087] Meeting minutes phrase library:

[0088] Meeting Topic: Monthly Sales Summary Meeting;

[0089] Attendees: Zhang San (Sales Manager), Li Si (Marketing Specialist), Wang Wu (Sales Representative);

[0090] Meeting time: 9:00 - 11:00;

[0091] Meeting Location: Company's First Conference Room;

[0092] Meeting moderator: Zhao Liu;

[0093] Meeting recorder: Qian Qi;

[0094] The meeting focused primarily on last month's sales performance, discussing regional sales data, market feedback, and customer complaint handling.

[0095] Sales performance discussion: Sales data for each region, market feedback, and handling of customer complaints;

[0096] Discussion results: Specific improvement plan, marketing strategy, and optimization of customer complaint handling process;

[0097] Next steps and responsibilities: Zhang San is responsible for adjusting sales strategies, Li Si is responsible for marketing plans, and Wang Wu is responsible for handling customer complaints.

[0098] Meeting summary: Zhao Liu emphasized teamwork and required each department to advance its work according to its assigned tasks.

[0099] 103. During computer downtime or at a fixed time each day, start the cleaning process, use a large model to judge the novelty and relevance of knowledge, and store the knowledge items that pass the judgment into the knowledge base.

[0100] The trigger mechanism for this step is to start the cleaning process during idle computer hours (such as when the screensaver is activated) or at a fixed time each day (such as 22:00).

[0101] The processing steps for this step are as follows:

[0102] Step 1: RAG pre-search: Use the knowledge to be cleaned to match the existing knowledge base and mark duplicate content (such as existing invoice reimbursement process documents).

[0103] Step 2: Large Model Judgment: Analyze the novelty and relevance of knowledge through local or remote large models, and filter out content that can be retained (such as new popular color trends in the design field).

[0104] It's important to note that quantifying the novelty and relevance of knowledge is a key task in knowledge management. Knowledge novelty measures its uniqueness and innovativeness, and can be assessed by calculating the diversity of knowledge units, topic novelty (such as inverse document frequency (IDF) of keywords), or variation and distance within the knowledge network. Knowledge relevance measures its connection to a specific topic or domain, and can be quantified through text similarity (such as cosine similarity, TF-IDF weights), knowledge network embedding features (such as centrality, structural holes), or machine learning models combined with multi-dimensional features. For comprehensive evaluation, these indicators are standardized and linearly summed, and their effectiveness is verified through recall and precision. In short, novelty focuses on the "newness" of knowledge, relevance focuses on the "usefulness" of knowledge, and combining the two allows for a comprehensive assessment of knowledge value.

[0105] In this embodiment, a large model is selected to evaluate the novelty and relevance of knowledge. The large model can be a locally deployed model, such as an industry-specific model based on the BERT architecture within a company, which offers advantages in rapid response and data privacy protection, enabling direct analysis and evaluation of knowledge in a local environment; or it can be a remotely deployed model, such as OpenAI's GPT-4, which has more powerful computing capabilities and a richer knowledge base, capable of handling more complex tasks and providing more accurate evaluation results. By flexibly selecting a local or remote large model, the novelty and relevance of knowledge can be efficiently quantified according to actual needs and resource conditions, providing strong support for knowledge management and innovative research.

[0106] Step 3: Tag Update: Automatically adjust user tag weights based on newly added knowledge (e.g., if you frequently process Excel numerical data, the weight of the financial tag will be increased to 85%).

[0107] 104 In the knowledge base, cutting-edge knowledge in vertical fields is automatically obtained from the network based on user tags and a large model is used to generate summaries, as well as low-frequency outdated knowledge is deleted regularly.

[0108] Based on user tags, cutting-edge knowledge is automatically retrieved from industry websites and academic platforms (such as the 2024 UI design trend report for design users), and summaries are generated using a large model and stored according to tags. Infrequent and outdated knowledge (such as old software operation guides that have not been accessed for more than a year) is regularly deleted to keep the knowledge base active.

[0109] 105 captures user input in real time and combines historical behavior data with a knowledge base to perform RAG retrieval to provide intelligent suggestions.

[0110] In this embodiment, the real-time capture of user input and the RAG retrieval based on historical behavior data and knowledge base to provide intelligent prompts include: capturing user input in real time in interactive scenarios and generating intelligent prompts related to the current task context based on the knowledge base and historical behavior data; the intelligent prompts are presented in the form of interactive fragments and can be directly inserted into the current working interface by triggering events.

[0111] For example, in scenarios such as document editing (e.g., Word), information retrieval (e.g., browsers), and software operation (e.g., Photoshop), by capturing user input in real time and combining it with historical user behavior data and a knowledge base for RAG (Retrieval-Augmented Generation) retrieval, intelligent suggestions can be generated to help users complete tasks more efficiently. For instance, in a document editing scenario, when a user enters "meeting," the system will suggest relevant meeting minutes templates or schedule document links based on historical data. In a design software scenario, when a user creates a canvas using Photoshop, the system will suggest standard sizes (e.g., WeChat official account cover size, poster standard size) and recommended color schemes based on common design needs. Users can quickly insert these suggestions using keyboard shortcuts or mouse operations (e.g., clicking and moving), thereby reducing repetitive input and improving work efficiency.

[0112] In summary, unlike static knowledge bases or models that rely on manual updates by users, this embodiment achieves automated, personalized, and timely expansion of the knowledge system.

[0113] Figure 2 This is a structural block diagram of an intelligent prompting system provided in an embodiment of this application. Figure 2 As shown, the system includes:

[0114] The user tag and data path configuration module 201 is used to receive user-defined occupation tags and knowledge bias tags, and configure the data collection path.

[0115] In this embodiment, the user tags are set as the main tag "Designer" and the sub-tags "Graphic Design" and "E-commerce Poster". Regarding the data acquisition path, the user selected "Screenshot Operation" to capture competitor poster designs, "Document Reading" to parse comments in PSD source files, and "Software Usage Log" to record layer operations in Photoshop.

[0116] The multi-dimensional behavioral data real-time acquisition and parsing module 202 is connected to the user tag and data path configuration module, and is used to collect and parse the user's multi-dimensional behavioral data on the personal computer in real time to generate structured knowledge entries.

[0117] In this embodiment, when the user was designing, they captured a promotional poster from an e-commerce platform using a screenshot. OCR recognized the text "618 Big Sale" and "Discount Rules," as well as the color codes (#FF6B6B, #4ECDC4), and generated a knowledge entry titled "Promotional Design - Color Scheme." Simultaneously, regarding PSD document reading, the user's opened "Double 11 Main Visual.psd" file was parsed, extracting the font style (Source Han Sans Bold) and element layout (left image, right text structure) from the layers, and associating them with the "E-commerce Poster - Design Specifications" tag. Furthermore, software operation log monitoring revealed frequent use of the "Pen Tool + Path Fill" combination, generating an operation record titled "Image Cutout Techniques - Advanced Pen Tool."

[0118] The dynamic knowledge cleaning and intelligent storage module 203 is connected to the multi-dimensional behavioral data real-time acquisition and analysis module. It is used to start the cleaning process during computer idle periods or at fixed times every day, use a large model to judge the novelty and relevance of knowledge, and store the knowledge items that pass the judgment into the knowledge base.

[0119] The knowledge base dynamic expansion and anti-fixation module 204 is connected to the dynamic knowledge cleaning and intelligent storage module. It is used to automatically obtain cutting-edge knowledge in vertical fields from the network based on user tags and generate summaries through a large model, as well as to periodically delete low-frequency outdated knowledge.

[0120] In this embodiment, the large model judges the collected content: it considers "618 promotional poster color scheme" to be retained in the database due to its timeliness; while "pen tool basic tutorial" is marked as to be eliminated because it is low-frequency content and is already covered in the knowledge base. At the same time, the system automatically retrieves articles on "2024 flat icon design trends" from design platforms such as Dribbble and Behance every week, generates summaries, and associates them with the "graphic design - design trends" tag.

[0121] The contextualized intelligent prompting service module 205 is connected to the knowledge base dynamic expansion and anti-fixation module. It is used to capture user input content in real time and perform RAG retrieval in combination with historical behavior data and knowledge base to provide intelligent prompts.

[0122] In this embodiment, when creating a new canvas in Photoshop, the system recommends a commonly used e-commerce poster size of 1920×600 pixels based on the user's tag "Designer" and sub-tags "Graphic Design" and "E-commerce Poster," and suggests the most recently used color scheme (#FF6B6B as the main color). When the user enters the keyword "coupon," the system will suggest the path to the coupon template file in historical design examples, as well as a link to the document "Promotional Element Layout Techniques."

[0123] Furthermore, such as Figure 3 As shown, the multi-dimensional behavioral data real-time acquisition and analysis module 202 includes:

[0124] The Search Behavior Capture Module 2021 is used to capture keywords from the browser or local search box and call the vertical search engine to obtain result summaries.

[0125] The Document Reading and Parsing Module 2022 is used to parse the titles, charts, and annotations in a document to extract key information.

[0126] The screenshot OCR parsing module 2023 is used to recognize text and tables in screenshots to extract valid information.

[0127] The Keyboard Input High-Frequency Phrases Statistics Module 2024 is used to count high-frequency input patterns and generate a user-specific phrase library.

[0128] The software uses the log parsing module 2025 to record and parse the operation steps of commonly used software.

[0129] Furthermore, such as Figure 4 As shown, the dynamic knowledge cleaning and intelligent data entry module 203 includes:

[0130] RAG pre-retrieval module 2031 is used to match the knowledge to be cleaned with the existing knowledge base to mark duplicate content.

[0131] The large model judgment module 2032 is used to filter content that can be retained based on local or remote large models.

[0132] The tag weight dynamic adjustment module 2033 is used to automatically adjust the weight of user tags based on newly added knowledge.

[0133] As a possible implementation, there can be multiple occupational tags, including a first occupational tag and a second occupational tag, such as... Figure 5 As shown, the label weight dynamic adjustment module 2033 includes:

[0134] The Occupation Type Label Determination Module 20331 is used to determine occupation type labels based on newly added knowledge.

[0135] The weight dynamic execution module 20332 is used to increase the weight of the first occupation label if the occupation type label is the first occupation label, and at the same time, perform non-negative attenuation on the weight of the second occupation label according to the preset attenuation coefficient; if the occupation type label includes the first occupation label and the second occupation label, the weights of the first occupation label and the second occupation label remain unchanged; wherein, the weight change of any occupation label is restricted within the preset weight range.

[0136] It should be noted that users can choose from a variety of career tags, such as graphic designer and programmer.

[0137] For example, the user selects the occupational tags "graphic designer" and "programmer". When the user uploads a document involving "image processing using Python", the occupational type tag determination module 20331 extracts keywords such as "Python" and "image processing" and compares them with a predefined occupational tag dictionary. Since "Python" is highly related to "programmer" and "image processing" is also related to "graphic designer", the occupational type tags for the document are determined to be "programmer" and "graphic designer". The weight dynamic execution module 20332, with an initial weight of 4 for "graphic designer" and 3 for "programmer", maintains the weights of both tags unchanged since the document involves both occupational tags; that is, the weight of "graphic designer" remains 4 and the weight of "programmer" remains 3. If a user subsequently uploads a document that only involves "Python programming," the occupation type label determination module 20331 determines the occupation type label as "programmer," and the dynamic weight execution module 20332 increases the weight of "programmer" to 4, while simultaneously reducing the weight of "graphic designer" to 3.2 by a preset attenuation coefficient of 0.8. All adjusted weights remain within the preset weight range [1, 10]. In this way, the system can dynamically adjust the weights of occupation labels based on newly added knowledge, better reflecting the importance and relevance of each occupation label within the knowledge system.

[0138] As a possible implementation, such as Figure 6 As shown, the occupational type label determination module 20331 includes:

[0139] The dual keyword extraction module 203311 is used to automatically extract two types of keywords from the input knowledge content text using natural language processing technology: domain keywords and action keywords.

[0140] The tag comparison module 203312 is used to compare the extracted domain keywords and action keywords with a predefined occupational tag dictionary, which contains multiple occupational tags, each of which is associated with a set of specific keywords.

[0141] The tag matching module 203313 is used to identify a job tag as a job type tag if a keyword matches a certain job tag; if the keyword does not match any job tag, the job type tag is marked as uncategorized and manual review is triggered.

[0142] For example, when processing the user input text "I am developing an AI-based recommender system. I need to write an algorithm using Python and optimize the model's performance.", the dual-keyword extraction module 203311 extracts the domain keywords "artificial intelligence," "recommender system," "Python," "algorithm," and "model," as well as the action keywords "develop," "use," "write," and "optimize." The tag comparison module 203312 compares these keywords with a predefined occupational tag dictionary, finding that "Python," "write," "algorithm," and "optimize" are related to the occupational tag "programmer," while "artificial intelligence," "recommender system," "model," and "optimize" are related to the occupational tag "AI researcher." Finally, the tag matching module 203313 determines the occupational type tags as "programmer" and "AI researcher."

[0143] Example 2:

[0144] Embodiments of this application also provide a computer device, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of the present invention during runtime.

[0145] The aforementioned memory can refer to devices inside a computer used to store data and programs, including RAM, hard disks, etc. RAM can be used to temporarily store running programs and data, while hard disks can be used to store programs and data long-term. Memory enables the computer to read and write data and execute programs. The aforementioned processor is responsible for executing instructions in computer programs and performing data processing. It can also be responsible for controlling and executing various operations, including arithmetic operations, logical operations, and data transmission.

[0146] Example 3:

[0147] Embodiments of this application also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of the present invention.

[0148] The aforementioned computer storage media can refer to the media used in computer memory to store certain discontinuous physical quantities. Computer storage media mainly include semiconductors, magnetic cores, magnetic drums, magnetic tapes, laser discs, etc. Computer-readable storage media include stored programs, which can be a set of instructions that a computer can recognize and execute, running on an electronic computer to meet certain information needs.

[0149] Example 4:

[0150] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of the present invention.

[0151] The aforementioned computer program products can refer to software programs that have been written, tested, and released, and can run on computers or other devices. Computer program products can include application programs, operating systems, utility software, etc., used to achieve specific functions or solve specific problems.

[0152] Example 5:

[0153] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program that, when executed by a processor, implements the methods in various embodiments of the present invention.

[0154] The aforementioned non-volatile computer-readable storage medium can refer to a medium for storing data. Non-volatile computer-readable storage media can retain data without loss when power is off and can be used to store long-term data, such as operating systems, applications, and user files. Non-volatile storage media can include hard disk drives, solid-state drives, optical disks, and flash memory storage devices, etc.

[0155] Example 6:

[0156] Embodiments of this application also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of the present invention.

[0157] The aforementioned computer program can refer to a set of instructions used to tell the computer to perform specific tasks or operations. Computer programs can be written by programmers using specific programming languages ​​and can include algorithms, data structures, logic, and control flow. Computer programs can be used for a variety of purposes, including application software, operating systems, etc.

[0158] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0159] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0160] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0161] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0162] If the integrated unit is implemented as 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 technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0163] In summary, this embodiment utilizes large-scale model technology to monitor users' multi-dimensional actions on their personal computers in real time, such as search behavior, browsing results, screenshot operations, document reading, and keyboard input, enabling in-depth analysis and dynamic knowledge extraction. It combines a user-defined tag system to achieve personalized knowledge filtering and leverages RAG technology to deeply integrate a dynamically updated knowledge base with user input scenarios, constructing an end-to-end personalized knowledge service chain. Specifically, this includes: first, a multi-dimensional behavioral data collection mechanism that globally intercepts user-configured personalized data paths, enabling structured parsing of unstructured data and constructing a dynamic user knowledge demand graph through full-scenario behavioral trajectory capture; second, a dynamic knowledge base expansion and anti-stagnation strategy that proactively acquires cutting-edge knowledge in vertical fields based on user tags, generates summaries through a large-scale model, and categorizes and stores them, while periodically eliminating low-frequency and outdated content to maintain knowledge base activity and achieve automated, personalized, and timely expansion of the knowledge system.

[0164] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An intelligent prompting system, characterized in that, include: The user tag and data path configuration module is used to receive user-defined occupation tags and knowledge bias tags, and configure the data collection path; The multi-dimensional behavioral data real-time acquisition and parsing module is connected to the user tag and data path configuration module, and is used to collect and parse the user's multi-dimensional behavioral data on the personal computer in real time to generate structured knowledge entries. The dynamic knowledge cleaning and intelligent storage module is connected to the multi-dimensional behavioral data real-time acquisition and analysis module. It is used to start the cleaning process during computer idle periods or at fixed times every day, use a large model to judge the novelty and relevance of knowledge, and store the knowledge items that pass the judgment into the knowledge base. The knowledge base dynamic expansion and anti-fixation module is connected to the dynamic knowledge cleaning and intelligent storage module. It is used to automatically obtain cutting-edge knowledge in vertical fields from the network based on user tags and generate summaries through a large model, as well as to periodically delete low-frequency outdated knowledge. The scenario-based intelligent prompting service module is connected to the knowledge base dynamic expansion and anti-fixation module. It is used to capture user input content in real time and perform RAG retrieval in combination with historical behavior data and knowledge base to provide intelligent prompts. The multi-dimensional behavioral data real-time acquisition and analysis module includes: The search behavior capture module is used to capture keywords from the browser or local search box and call the vertical domain search engine to obtain result summaries. The document reading and parsing module is used to parse the titles, charts, and annotations in a document to extract key information. The screenshot OCR parsing module is used to recognize text and tables in screenshots to extract useful information; The keyboard input high-frequency phrase statistics module is used to count high-frequency input patterns and generate a user-specific phrase library; The software uses a log parsing module to record and parse the operation steps of commonly used software by users; The dynamic knowledge cleaning and intelligent storage module includes: The RAG pre-retrieval module is used to match the knowledge to be cleaned with the existing knowledge base to mark duplicate content. The large model judgment module is used to filter content that can be retained based on local or remote large models. The tag weight dynamic adjustment module is used to automatically adjust the weight of user tags based on newly added knowledge. There are multiple occupational tags, including a first occupational tag and a second occupational tag. The tag weight dynamic adjustment module includes: The occupation type label determination module is used to determine occupation type labels based on newly added knowledge. The weight dynamic execution module is used to increase the weight of the first occupation label if the occupation type label is the first occupation label, and at the same time, to non-negatively decrease the weight of the second occupation label according to a preset attenuation coefficient; if the occupation type label includes the first occupation label and the second occupation label, the weights of the first occupation label and the second occupation label remain unchanged; wherein, the weight change of any occupation label is limited to a preset weight range. The occupation type label determination module includes: The dual keyword extraction module is used to automatically extract two types of keywords from the input knowledge content text using natural language processing technology: domain keywords and action keywords; The tag comparison module is used to compare the extracted domain keywords and action keywords with a predefined occupational tag dictionary, which contains multiple occupational tags, each of which is associated with a specific set of keywords; The tag matching module is used to identify a job tag as a job type tag if a keyword matches a certain job tag; if the keyword does not match any job tag, the job type tag is marked as uncategorized and manual review is triggered.

2. A method for intelligent prompting, characterized in that, The system as described in claim 1 is employed, and the method comprises: Receive user-defined occupational tags and knowledge bias tags, and configure the data collection path; Real-time collection and analysis of multi-dimensional user behavior data on personal computers to generate structured knowledge entries; The cleaning process is started during computer downtime or at a fixed time each day. The large model is used to determine the novelty and relevance of knowledge, and knowledge items that pass the determination are stored in the knowledge base. In the knowledge base, cutting-edge knowledge in vertical fields is automatically obtained from the Internet based on user tags and a large model is used to generate summaries, while low-frequency and outdated knowledge is regularly deleted; It captures user input in real time and combines historical behavior data with a knowledge base to perform RAG retrieval to provide intelligent suggestions.

3. The intelligent prompting method according to claim 2, characterized in that, The data collection path includes search behavior, browsing search results, screenshot operations, document reading, keyboard input, and software usage logs.

4. The intelligent prompting method according to claim 2, characterized in that, The real-time capture of user input, combined with historical behavior data and a knowledge base for RAG retrieval to provide intelligent suggestions, includes: In interactive scenarios, user input is captured in real time, and intelligent prompts related to the current task context are generated based on the knowledge base and historical behavior data. The intelligent prompts are presented in the form of interactive fragments and can be directly inserted into the current working interface by triggering events; The interactive scenarios include one of them: document editing, information retrieval, and software operation; The triggering events include one of the following: keyboard shortcuts and mouse operations.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent prompting method as described in any one of claims 2-4.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the intelligent prompting method as described in any one of claims 2-4.

Images