Customer service knowledge base construction method and device and computer readable storage medium

Abstract

The invention provides a customer service knowledge base construction method and device and a computer readable storage medium. In the scheme, first information including function requirements, performance indexes and business rules of a target product is extracted from a requirement document of the target product, and second information including interface elements, interface element attributes and interface layout information of the target product is extracted from a prototype graph of the target product; inputting the first information and the second information into a pre-trained GPT large model for analysis to obtain a target knowledge point of the target product; generating customer service knowledge base entries according to the target knowledge points, and obtaining a customer service knowledge base formed by the customer service knowledge base entries; and if it is monitored that the version of the demand document and / or the prototype graph is updated, updating the customer service knowledge base according to the updated demand document and / or the updated prototype graph. According to the scheme, the problem of limitation of construction of the customer service knowledge base in the prior art is solved, wherein the limitation comprises low accuracy, low automation degree and update lag.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a customer service knowledge base construction method, a customer service knowledge base construction device, and a computer-readable storage medium. Background Art

[0002] A customer service knowledge base is a database used to store, manage, and provide information about a company's products and services. In today's rapidly evolving technological landscape, building and maintaining an efficient customer service knowledge base has become a critical factor in ensuring service quality and customer satisfaction. However, existing customer service knowledge base construction methods face significant efficiency bottlenecks and quality challenges. Product managers create requirements specifications, prototypes, and business process diagrams that contain vast amounts of information, but translating and integrating this information often requires significant manpower and time.

[0003] In existing technologies, the construction and updating of customer service knowledge bases are highly dependent on manual operations, which leads to limitations in the construction of customer service knowledge bases, including low accuracy, low degree of automation and delayed updates, making it difficult to adapt to the rapid iterative product development cycle. Summary of the Invention

[0004] The main purpose of this application is to provide a customer service knowledge base construction method, a customer service knowledge base construction device and a computer-readable storage medium, so as to at least solve the problems of limitations in the construction of customer service knowledge base in the prior art, including low accuracy, low degree of automation and delayed updates.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for constructing a customer service knowledge base is provided, comprising: extracting first information from a requirement document of a target product, and extracting second information from a prototype diagram of the target product, wherein the first information includes functional requirements, performance indicators and business rules of the target product, and the second information includes interface elements, interface element attributes and interface layout information of the target product; inputting the first information and the second information into a pre-trained GPT large model for parsing to obtain target knowledge points of the target product, wherein the target knowledge points include functional descriptions, operating guides and frequently asked questions FAQs; utilizing template engine technology to generate customer service knowledge base entries based on the target knowledge points to obtain a customer service knowledge base composed of the customer service knowledge base entries, wherein the customer service knowledge base entries include questions and answers to the questions; monitoring the requirement document and the prototype diagram, and if the version of the requirement document and / or the prototype diagram is updated, updating the customer service knowledge base according to the updated requirement document and / or the updated prototype diagram.

[0006] According to another aspect of the present application, a customer service knowledge base construction device is provided, including: a first acquisition unit, used to extract first information from a requirement document of a target product, and extract second information from a prototype diagram of the target product, wherein the first information includes functional requirements, performance indicators and business rules of the target product, and the second information includes interface elements, interface element attributes and interface layout information of the target product; an input unit, used to input the first information and the second information into a pre-trained GPT large model for parsing to obtain target knowledge points of the target product, wherein the target knowledge points include functional descriptions, operating guidelines and frequently asked questions FAQs; a generation unit, used to generate customer service knowledge base entries based on the target knowledge points using template engine technology, to obtain a customer service knowledge base composed of the customer service knowledge base entries, wherein the customer service knowledge base entries include questions and answers to the questions; an update unit, used to monitor the requirement document and the prototype diagram, and if the version of the requirement document and / or the prototype diagram is updated, update the customer service knowledge base according to the updated requirement document and / or the updated prototype diagram.

[0007] According to another aspect of the present application, a computer-readable storage medium is provided, which includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute any one of the customer service knowledge base construction methods.

[0008] Applying the technical solution of the present application, first information is extracted from the requirement document of the target product, and second information is extracted from the prototype diagram of the target product, wherein the first information includes the functional requirements, performance indicators and business rules of the target product, and the second information includes the interface elements, interface element attributes and interface layout information of the target product; the first information and the second information are input into the pre-trained GPT large model for parsing to obtain the target knowledge points of the target product, and the target knowledge points include functional descriptions, operating guidelines and FAQs; using template engine technology, customer service knowledge base entries are generated according to the target knowledge points to obtain a customer service knowledge base composed of customer service knowledge base entries, and the customer service knowledge base entries include questions and answers to the questions; monitoring the requirement document and prototype diagram, if the version of the requirement document and / or prototype diagram is updated, the customer service knowledge base is updated according to the updated requirement document and / or updated prototype diagram. In this solution, by acquiring and parsing the target product's requirement documents and prototype images, combined with the deep analysis capabilities of the GPT large model, accurate target knowledge points are efficiently generated. Then, template engine technology is used to quickly build a standardized customer service knowledge base, monitor version updates of documents and prototype images, and trigger automatic updates of the knowledge base in real time to ensure the timeliness and accuracy of the content, thereby solving the limitations of the construction of customer service knowledge bases in existing technologies, including low accuracy, low degree of automation, and delayed updates. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The drawings that constitute part of this application are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0010] Figure 1 A hardware structure block diagram of a mobile terminal for executing a customer service knowledge base construction method provided in an embodiment of the present application is shown;

[0011] Figure 2 A flow chart of a method for building a customer service knowledge base according to an embodiment of the present application is shown;

[0012] Figure 3 A flowchart of a specific customer service knowledge base construction method provided according to an embodiment of the present application is shown;

[0013] Figure 4 The figure shows a structural block diagram of a customer service knowledge base construction device provided according to an embodiment of the present application.

[0014] The above drawings include the following reference numerals:

[0015] 102. Processor; 104. Memory; 106. Transmission device; 108. Input / output device. DETAILED DESCRIPTION

[0016] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0017] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0018] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0019] As introduced in the background technology, the construction of the customer service knowledge base in the prior art has many limitations due to its high reliance on manual operations. In order to solve the problems of limitations in the construction of the customer service knowledge base in the prior art, including low accuracy, low degree of automation and delayed updates, the embodiments of the present application provide a customer service knowledge base construction method, a customer service knowledge base construction device and a computer-readable storage medium.

[0020] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0021] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure diagram of a mobile terminal of a customer service knowledge base construction method according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown) a processor 102 (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA and other processing devices) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0022] The memory 104 can be used to store computer programs, such as software programs and modules of application software, such as the computer program corresponding to the device information display method in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implementing the above-mentioned method. The memory 104 may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned networks include but are not limited to the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include a wireless network provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0023] In this embodiment, a method for constructing a customer service knowledge base running on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0024] Figure 2 Schematic diagram of the customer service knowledge base construction method according to the embodiment of the present application. Figure 2 As shown, the method includes the following steps:

[0025] Step S201: extracting first information from a target product's requirements document and second information from a prototype diagram of the target product, wherein the first information includes functional requirements, performance indicators, and business rules of the target product, and the second information includes interface elements, interface element attributes, and interface layout information of the target product;

[0026] Specifically, the requirements document includes a requirements specification and a business process diagram. The requirements specification is a document that describes in detail the functions, performance, interface, interface and other non-functional requirements of the software system or product; the business process diagram is a graphical representation used to describe business activities and processes in the software system. The requirements document usually contains detailed requirements for the design and development of the target product, including functional requirements, performance indicators, business rules, etc. Functional requirements describe the functions that the product must provide, such as user login, search, payment, etc.; performance indicators specify the performance of the product under specific conditions, such as response time, number of concurrent users, data processing speed, etc.; business rules define the constraints of the business logic in the product, such as order processing flow, user authority control, etc. The functional requirements, performance indicators and business rules of the target product are extracted from the requirements document of the target product for subsequent parsing using the pre-trained GPT large model.

[0027] The prototype provides a preliminary design of the target product's user interface, including interface elements (such as buttons and input boxes), interface element attributes (such as position, size, and color), and interface layout information (such as page flow and element grouping). The target product's interface elements, interface element attributes, and interface layout information are extracted from the prototype for subsequent analysis using the pre-trained GPT large model.

[0028] Through step S201, the key information required to build a customer service knowledge base can be extracted from different perspectives and data sources (requirements documents and prototype diagrams). The first information provides a description of the product functions and business logic, while the second information provides an intuitive user interface design. Combining the first information and the second information, a more comprehensive and accurate knowledge base entry can be generated. For example, from the requirements document, it can be extracted that "the user needs to enter a username and password on the login page", and at the same time, the specific design and element attributes of the login page can be identified from the prototype diagram, such as "the username input box is located in the upper left corner of the screen" and "the password input box is immediately below the username input box and has a hidden character function". By integrating these two parts of information, an accurate "how to log in" operation guide can be generated, including the location of the input box, input prompts, and any specific user interface prompts, such as "the password field contains an eye icon, which can be clicked to display the password".

[0029] This method of extracting information from multiple sources ensures the comprehensiveness and accuracy of the customer service knowledge base content, and also provides a data foundation for subsequent automated update processes.

[0030] Step S202: Input the first information and the second information into a pre-trained GPT large model for parsing to obtain target knowledge points of the target product, including function descriptions, operation guides, and FAQs.

[0031] Specifically, the first information comes from the requirements document, including functional requirements, performance indicators and business rules; the second information comes from the prototype diagram, covering interface elements, interface element attributes and interface layout information. The first information and the second information are input into the pre-trained GPT large model, so that the pre-trained GPT large model can fully understand the design intent and user interface of the product. The pre-trained GPT large model is trained with massive text data for powerful language understanding and generation capabilities. In the process of building a customer service knowledge base, the pre-trained GPT large model can understand functional requirements and business rules and convert them into concise and clear functional descriptions. For example, for the "user login function", the model can generate: "Log in with username and password, and support remembering login status."

[0032] Combining the first and second pieces of information, the pre-trained GPT model can create detailed instructions for users on how to use specific features. For example, in the login process, it can generate: "Click the 'Login' button on the login page, enter your registered email address and password, and then click 'Submit'."

[0033] Based on its understanding of product functionality and predictions of potential user questions, the pre-trained GPT large model can generate FAQs covering common user questions and their solutions. For example, "Question: What should I do if I forgot my password? Solution: Click the 'Forgot Password' link on the login page and follow the instructions to reset your password."

[0034] After the first information and the second information are input into the pre-trained GPT large model, the pre-trained GPT large model will perform in-depth analysis and comprehensive understanding, and can convert these structured information into target knowledge points in the form of natural language. Specifically, first, the pre-trained GPT large model understands the meaning of the input first information and the second information, which includes functional requirements, performance indicators and business rules extracted from the requirements document, as well as interface elements, interface element attributes and interface layout information extracted from the prototype diagram. The pre-trained GPT large model combines the abstract concepts in the requirements document with the specific interface elements in the prototype diagram to generate a guide containing specific operation steps to ensure the practicality of the guide. After understanding all the input information, the pre-trained GPT large model generates easy-to-understand natural language text (ie, target knowledge points), such as functional descriptions, operation guides and FAQs. For example, based on the position and associated text of the "Login" button in the prototype diagram, an operation guide related to it can be generated.

[0035] In summary, by inputting the structured first information and second information into the pre-trained GPT large model for parsing, the target knowledge points of the customer service knowledge base can be generated, including function descriptions, operation guides and FAQs. This process makes full use of the natural language understanding and generation capabilities of the GPT large model, providing an efficient and accurate method for the construction of the customer service knowledge base.

[0036] Step S203: Generate customer service knowledge base entries based on the target knowledge points using template engine technology to obtain a customer service knowledge base consisting of the customer service knowledge base entries, wherein the customer service knowledge base entries include questions and answers to the questions;

[0037] Specifically, a template engine is a programming tool used to fill data into a predefined template to generate dynamic output. In this embodiment, template engine technology is used to generate customer service knowledge base entries. Typically, a template contains a fixed structure and variable placeholders, while the data is the actual content extracted from the target knowledge point. The target knowledge point is generated by analyzing the requirements document and prototype diagram by a pre-trained GPT large model, including functional descriptions, operating instructions, FAQs, etc. Among them, the functional description is a brief description of a specific function of the product; the operating instructions are detailed steps or instructions on how to use the product or its features; the FAQs are a preset collection of questions and answers that answer common questions that users may encounter.

[0038] First, select an appropriate template based on the target knowledge point type (such as functional description, operation guide, FAQ, etc.). For example, an operation guide can use the structure "Step 1: [action description]; Step 2: [action description];...", while an FAQ can adopt the format "Q: [problem description]; A: [solution description]." Next, fill the corresponding template placeholders with the target knowledge point data obtained from the pre-trained GPT large model. For example, in an operation guide, "[action description]" will be replaced with the specific user operation steps, such as "Click the login button; enter the username and password; click Submit." The template engine automatically combines the template and data to generate a complete customer service knowledge base entry. For example, a simple FAQ template might be: "Q: How to perform operation X? A: {step description}," where "{step description}" is dynamically generated and filled in based on the operation guide in the target knowledge point. The generated customer service knowledge base entries need to be integrated into a structured knowledge base. This usually involves classifying (such as by function, by problem type), sorting and indexing entries to facilitate users to find and manage information within the system, resulting in a customer service knowledge base.

[0039] Here's an example:

[0040] Let's assume the target knowledge point is "How to change your password in Product Y?" The instructions for this part are described as "Visit the Account Settings page; find the Change Password option in Security Settings; enter your current password; set and confirm your new password." Using template engine technology, you can select a question-and-answer FAQ template, such as "Q: How do I change my password in Product Y? A: Visit the Account Settings page; find the Change Password option in Security Settings; enter your current password; set and confirm your new password." This creates a complete customer service knowledge base entry.

[0041] Template engine technology transforms information parsed by a pre-trained GPT large model into structured, user-friendly knowledge base content. This ensures consistent formatting across customer service knowledge base entries, making the knowledge base professional, understandable, and easy to use. This process requires virtually no human intervention, significantly improving the efficiency of building the customer service knowledge base and reducing workload and human error. Furthermore, the template engine allows for flexible data insertion and format adjustments, making it easy to add new knowledge points or optimize existing templates to adapt to product development and changes.

[0042] Step S204 , monitoring the requirement document and the prototype diagram. If the version of the requirement document and / or the prototype diagram is updated, updating the customer service knowledge base according to the updated requirement document and / or the updated prototype diagram.

[0043] Specifically, the version changes of the requirements document and prototype drawings are monitored in real time. Once a new submission or change is detected in the requirements document and / or prototype drawings, the subsequent update process is automatically triggered. When a version update is detected, the updated requirements document and / or updated prototype drawings are extracted to update the customer service knowledge base to reflect the latest status of the product documentation. This includes updating the text descriptions in the FAQ, operation guide or any related documents, and updating the screenshots or operation procedures related to the prototype drawings changes. Changes to the requirements document may not affect the prototype drawings, or they may affect the prototype drawings. Changes to the prototype drawings may not affect the requirements document, or they may affect the requirements document. For example, if the "Registration" page in the prototype drawing is redesigned, the system will automatically update the interface screenshots in the "How to Register" entry and modify the operation steps to match the new design.

[0044] Through continuous monitoring and automatic updates, the customer service knowledge base is always synchronized with the latest status of the product's requirements documents and prototypes. Furthermore, the automated update process reduces the need for manual monitoring and updates, saving time and resources. Change points can be precisely identified and all relevant entries in the customer service knowledge base are updated, avoiding issues such as missing or inconsistent information. In summary, through step S204, the automated update of the customer service knowledge base achieves real-time response to changes in the product's requirements documents and prototypes, ensuring the timeliness and accuracy of the customer service knowledge base content, and improving customer service quality and user experience.

[0045] Through this embodiment, the target product's requirement documents and prototype drawings are acquired and parsed, and combined with the deep analysis capabilities of the GPT large model, accurate target knowledge points are efficiently generated. Then, a standardized customer service knowledge base is quickly constructed using template engine technology, and version updates of documents and prototype drawings are monitored. This can trigger automatic updates of the knowledge base in real time, ensuring the timeliness and accuracy of the content, thereby solving the limitations of the construction of customer service knowledge bases in the prior art, including low accuracy, low degree of automation, and delayed updates.

[0046] During the specific implementation process, the above-mentioned requirement document and the above-mentioned prototype diagram are monitored. If the version of the above-mentioned requirement document and / or the above-mentioned prototype diagram is updated, the above-mentioned customer service knowledge base is updated according to the updated requirement document and / or the updated prototype diagram, including: monitoring the above-mentioned requirement document and the above-mentioned prototype diagram through the version control system interface; extracting the above-mentioned updated requirement document and / or the above-mentioned updated prototype diagram when it is detected that the version of the above-mentioned requirement document and / or the above-mentioned prototype diagram is updated; using a difference analysis algorithm to determine the changed content of the above-mentioned requirement document and / or the above-mentioned prototype diagram, the above-mentioned changed content includes the requirement document version change information and / or the prototype diagram version change information; according to the above-mentioned changed content, using the above-mentioned pre-trained GPT large model to parse the above-mentioned updated requirement document and / or the above-mentioned updated prototype diagram to generate updated target knowledge points; updating the above-mentioned customer service knowledge base according to the above-mentioned updated target knowledge points.

[0047] Specifically, Git is an open source distributed version control system used to track file changes during the software development process and record modifications made by multiple users to multiple branches. It is mainly designed to efficiently and reliably manage projects from very small to very large. It is particularly suitable for tracking code changes, but it is also widely used for version management of documents, configuration files, media files, etc. By integrating a version control system interface such as Git, it is ensured that any changes to the requirements document and / or prototype diagram can be captured in real time. By continuously monitoring the version status of the requirements document and prototype diagram, changes to the requirements document and prototype diagram can be monitored in real time, thereby triggering the automatic update process of the knowledge base to ensure that the knowledge base is always consistent with the product's requirements document and prototype diagram. This greatly simplifies the work of knowledge base maintenance and improves its accuracy and timeliness.

[0048] After extracting the updated requirements document and / or updated prototype, use the Diff Match Patch library to identify version changes in the requirements document. The Diff Match Patch library is an open source library designed for efficiently calculating and applying text differences. It consists of three main components: diff (difference calculation), match (matching algorithm), and patch (patch application). The diff component calculates the differences between two texts, the match component finds and marks identical sections in the text, and the patch component applies these differences to modify the text.

[0049] Using the diff algorithm in the Diff Match Patch library, the requirements document and the updated requirements document are compared to calculate the text differences between them. This includes identifying all additions and deletions to the text, as well as formatting changes (such as changes in font, size, color, or paragraph structure). The calculated text differences are analyzed and classified into specific change information (i.e., requirements document version change information), such as modifications to functional descriptions, additions or deletions to operational procedures, or updates to performance indicators. All identified changes are recorded and serve as the basis for updating the customer service knowledge base.

[0050] After extracting the updated requirements document and / or updated prototype, use the OpenCV library to identify prototype version changes. OpenCV is an open-source computer vision and machine learning software library that includes numerous image and video processing modules for object detection, image recognition, and feature extraction. Use OpenCV to compare the prototype with the updated prototype to identify any prototype version changes.

[0051] Through the above process, we can accurately identify changes in requirements documents and prototypes. Whether it's subtle text additions or deletions or formatting adjustments in requirements documents, or significant changes to interface elements or layout updates in prototypes, all can be captured efficiently and accurately. This change information is then used in the parsing process of the pre-trained GPT large model to generate updated target knowledge points. Ultimately, this ensures the dynamic updating of the customer service knowledge base and the timeliness of its content, thereby improving customer service quality and user satisfaction.

[0052] After determining the specific changes in the requirements document and / or prototype diagram, the changes are input into the pre-trained GPT large model. The pre-trained GPT large model parses the changes and generates updated target knowledge points. Based on the generated updated target knowledge points, the customer service knowledge base is updated in real time to reflect the latest changes in the product documentation.

[0053] By integrating the version control system interface, this embodiment can monitor the version updates of requirement documents and prototype drawings in real time. Once an update is detected, it automatically extracts the latest requirement documents and prototype drawings, and uses a differential analysis algorithm to accurately identify the changed content. Based on this change information, the pre-trained GPT large model performs in-depth analysis to generate updated target knowledge points, including new function descriptions, operation guides, and FAQs. Ultimately, based on these updated knowledge points, the customer service knowledge base is updated in a timely and accurate manner to ensure that it is always synchronized with the product documents it serves, thereby significantly improving customer service efficiency and user experience, while reducing maintenance costs and error rates.

[0054] In this embodiment, user behavior analysis and user profiling technology are introduced, leveraging big data to analyze user habits, preferences, and common problems encountered in using the customer service knowledge base, thereby establishing a personalized user model. User profiling technology is a data analysis method that aims to construct a digital representation of user characteristics. By collecting and analyzing data such as user behavior, preferences, and demographic information, it helps to gain a deeper understanding of users and, in turn, provide more personalized and precise service. Based on user profiling, it is possible to predict the types of problems users may encounter and proactively recommend relevant customer service knowledge base entries. For example, for users who frequently access product feature guides, the latest feature update guides or advanced feature tutorials can be recommended. Users can select knowledge points of interest based on their needs, generating a customized customer service knowledge base that only contains content that interests them. This customization not only improves the user experience but also reduces the overall maintenance burden of the knowledge base. Through personalized recommendations and a customized knowledge base, this embodiment can significantly improve the practicality and user satisfaction of the customer service knowledge base. User profiling technology ensures the accuracy and relevance of recommendations, while customization meets the specific needs of different user groups, improving query efficiency and service quality.

[0055] In some embodiments of the present application, first information is extracted from a requirement document of a target product, and second information is extracted from a prototype image of the target product, including: using a natural language processing algorithm to parse the requirement document of the target product to obtain the first information, the natural language processing algorithm including a BERT model or a GPT model; using an image processing algorithm to parse the prototype image of the target product to obtain the second information, the image processing algorithm including a SIFT algorithm and a YOLO algorithm.

[0056] Specifically, the requirements document includes key information about product functions, performance requirements, user interface descriptions, etc. In order to extract useful first information from the requirements document, a natural language processing algorithm is used, in particular a BERT model or a GPT model based on deep learning. Among them, the BERT (Bidirectional Encoder Representations from Transformers) model is a bidirectional Transformer model that understands text through context and is particularly suitable for performing tasks such as text classification, sentiment analysis, and entity recognition. It can accurately extract key entities and relationships from the requirements document. The GPT (Generative Pretrained Transformer) model is an autoregressive Transformer model that is good at generating coherent text. In an embodiment of the present application, the GPT model is used to parse the requirements document, understand its structure and semantics, and thus identify text paragraphs that describe product functions, operating procedures, or business rules.

[0057] The requirements document is fed into a BERT or GPT model. The model uses deep learning technology to perform semantic parsing on the document, identifying key entities, operational instructions, error messages, user interface descriptions, and other information within the document. The first information extracted includes statements describing product features, operational steps, performance indicators, and textual descriptions of interface elements.

[0058] The prototype image is a visual expression of the product design, which contains information such as the user interface layout, element position and style. In order to accurately extract the second information from the prototype image, the embodiment of the present application utilizes image processing algorithms, in particular the SIFT algorithm and the YOLO algorithm. Among them, the SIFT (Scale-Invariant Feature Transform) algorithm is used to identify key points and descriptors in the image, and can maintain robustness even when the image is scaled, rotated or the lighting changes. In this embodiment, the SIFT algorithm helps to identify interface elements in the prototype image, such as buttons, input boxes, links, etc. The YOLO (You Only LookOnce) algorithm is a real-time target detection algorithm that can locate and identify multiple objects directly from an image. Using the YOLO algorithm, all interface elements in the prototype image can be quickly and accurately identified, and their positions and categories can be determined.

[0059] The prototype image is first preprocessed by OpenCV to obtain a preprocessed prototype image. The preprocessing operations include scaling, cropping and enhancing the image to ensure the accuracy of the image processing algorithm. Among them, image enhancement is to enhance the key details of the image by adjusting the brightness, contrast, sharpness and other properties of the image, so that the image processing algorithm can more easily recognize features; cropping is to remove parts of the prototype image that are not related to the target, focus on the user interface area, reduce the complexity of processing, and improve the accuracy and recognition speed of the algorithm; scaling is to uniformly scale the image to a suitable size, usually to match the image size used when the algorithm is trained, to ensure that the algorithm can correctly recognize the elements in the image. Feature extraction is used to identify unique patterns or features in the image, such as edges, textures, shapes, etc. In this embodiment, the feature extraction algorithms used include SIFT (Scale-Invariant Feature Transform) algorithm, SURF (Speeded-Up Robust Features) algorithm or ORB (Oriented FAST and Rotated BRIEF) algorithm. The SIFT algorithm is used to detect key points in the image and calculate the descriptors of these points. Even if the image is scaled, rotated or the lighting changes, the SIFT algorithm can accurately identify interface elements. Similar to the SIFT algorithm, the SURF or ORB algorithm can also identify feature points in an image and is suitable for processing large-scale image data. The SIFT, SURF, or ORB algorithm is applied to the preprocessed prototype image to detect and extract the feature points of interface elements. Based on these feature points, the algorithm calculates descriptors, which contain attribute information such as the shape and texture of the interface element. The extracted interface element attributes include the element's size, shape, color, and position coordinates.

[0060] Pre-trained object detection models, such as the YOLO algorithm or other models based on convolutional neural networks, can identify objects of specific categories in images. The YOLO algorithm is a real-time object detection algorithm that treats object detection as a regression problem, predicting bounding boxes and categories directly from images. It is fast and highly accurate, making it very suitable for processing interface element recognition in prototype images. The convolutional neural network algorithm is a deep learning algorithm specifically designed to process image data. By learning the spatial hierarchy in the image, it can identify and locate elements in the image. Use the pre-trained YOLO or convolutional neural network to scan the pre-processed prototype image to identify various elements in the interface, such as buttons, input boxes, labels, etc. The algorithm locates each element and assigns a category label, resulting in a detailed list of interface elements, including their location and type.

[0061] Through graph theory and network analysis techniques, a spatial relationship diagram between elements can be constructed, including the hierarchical structure, relative positions, and potential interactions between elements. The identified interface elements are regarded as nodes in the graph, and the spatial relationships between nodes (such as up and down, left and right, and included) are regarded as edges. Graph theory algorithms, such as breadth-first search and depth-first search, are applied to analyze the hierarchical structure and layout relationships of interface elements. Network analysis methods are used to understand the interaction logic between elements, such as the association between buttons and input boxes, and the functional relationships between element groups.

[0062] The combination of preprocessing, feature extraction, element recognition, and layout analysis enables a comprehensive understanding of the interface design in the prototype image, including the attributes and location of each interface element, as well as the layout and interaction relationships between them. This information forms an important basis for building a customer service knowledge base, ensuring the precise matching and description of the customer service knowledge base content with the product interface. Through steps such as image preprocessing, feature extraction, object detection, and layout analysis, an in-depth analysis of the prototype image is achieved, extracting the detailed attributes, location, and layout information of the interface elements. This is one of the keys to building a customer service knowledge base. Through this series of processing, an accurate and comprehensive interface description can be formed, providing users with more detailed operation guides and product information.

[0063] To obtain the first piece of information, structured information extraction is performed first. Structured information extraction converts unstructured text data into a format that is easier to process and analyze. For the requirements document, this means splitting and extracting the document according to the hierarchical structure of titles and paragraphs. Using a natural language processing algorithm, the requirements document is segmented according to titles and paragraphs. Each title or paragraph is considered an information segment, and key information is extracted from each information segment, such as functional descriptions, operating procedures, performance indicators, etc. This information will serve as the basis for subsequent analysis. Text preprocessing aims to standardize text data and remove noise so that subsequent analysis can be performed more accurately. The extracted structured information is preprocessed by formatting the text to unify the format, such as removing excess spaces and standardizing punctuation, to ensure the consistency of the text data. De-noising is used to remove irrelevant information in the text, such as comments, repeated paragraphs, or text unrelated to the requirements, so that the algorithm can better focus on key information.

[0064] Perform text analysis on the preprocessed structured information to obtain text information. Text analysis aims to extract deeper insights from the preprocessed structured information, including identifying specific patterns and matching keywords. Regular expressions are used to search for specific patterns in the text, such as phone numbers, dates, specific commands, or functional descriptions, to quickly locate and extract structured data. Keywords related to product features, performance, or interface, such as "register," "login," and "search," are identified and further refined through keyword matching.

[0065] Information extraction is performed on text information to obtain extracted information. Information extraction is the conversion of the results of text analysis (text information) into structured data, including the identification of key information, entities, and relationships. Information extraction includes the extraction of key information, entity recognition, and relationship extraction. Among them, the extraction of key information is to determine the core information points in the document, such as the description of the user interface, the details of the functional process, etc., based on the results of regular expression extraction and keyword matching. Entity recognition is the use of natural language processing technology to identify entities in the document, such as function names, performance parameters, user roles, etc. These entities are the basis for building a customer service knowledge base. Relationship extraction is the analysis of the relationship between entities, such as the interaction between functions and user roles, and the association between performance indicators and operations. These relationships are the key to understanding the content of the document.

[0066] Natural language processing algorithms are used to deeply analyze the text and extracted information to gain a deeper understanding and obtain the initial first information. Specifically, deep learning models such as BERT or GPT are used to parse the text and extracted information, understanding the contextual meaning of the text and the relationships between entities. The algorithms can identify the underlying meaning of the text and parse complex statements to ensure a comprehensive interpretation of the requirements document. The parsed initial first information is formatted into a standardized structure and converted into a unified format such as JSON or XML to obtain the first information.

[0067] The natural language processing steps described above enable in-depth and accurate extraction of key information from requirement documents, including functional descriptions, performance specifications, and user interface guidelines. The combination of preprocessing, structured information extraction, text analysis, information extraction, deep parsing, and formatting ensures the comprehensiveness and accuracy of the extracted information, providing a solid data foundation for the automated construction of the customer service knowledge base. This series of natural language processing operations significantly improves the efficiency of customer service knowledge base construction, reduces manual errors, and ensures the clarity and relevance of the customer service knowledge base content, better meeting user query needs.

[0068] In summary, by using natural language processing algorithms (BERT or GPT models) to parse requirements documents and image processing algorithms (SIFT and YOLO algorithms) to parse prototype images, we can automatically, efficiently, and accurately extract key information from requirements documents and prototype images, providing a solid data foundation for the automated construction of the customer service knowledge base. This information includes not only product functional descriptions and performance requirements, but also the detailed location and style of user interface elements, forming a comprehensive material for building the customer service knowledge base. This improves the efficiency of customer service knowledge base construction, information accuracy, and ease of maintenance, while also significantly enhancing user experience and interaction satisfaction.

[0069] In other embodiments of the present application, template engine technology is used to generate customer service knowledge base entries based on the above-mentioned target knowledge points to obtain a customer service knowledge base composed of the above-mentioned customer service knowledge base entries, and the above-mentioned customer service knowledge base entries include questions and answers to the above-mentioned questions, including: using the above-mentioned template engine technology to select corresponding templates according to different above-mentioned target knowledge points, the above-mentioned templates include titles and content, and the above-mentioned template engine technology is Jinja2 or Thymeleaf; filling the information of the above-mentioned target knowledge points into the corresponding templates of the above-mentioned target knowledge points to generate the above-mentioned customer service knowledge base entries; integrating all the above-mentioned customer service knowledge base entries to obtain the above-mentioned customer service knowledge base composed of the above-mentioned customer service knowledge base entries.

[0070] Specifically, after extracting target knowledge points from the requirements document and prototype diagram, the next step is to convert these knowledge points into customer service knowledge base entries—namely, questions and their corresponding answers. This process is accomplished using template engine technology to ensure entry consistency and readability. This embodiment utilizes template engine technologies such as Jinja2 or Thymeleaf, which allow for the generation of dynamic content using predefined templates. Template engines can handle variable substitution, conditional logic, and loops, making them ideal for generating structured customer service knowledge base entries. Each template consists of two parts: a title and content. The title is the main theme of the entry, while the content details the question and its answer. Template design takes into account the characteristics of different knowledge points, such as product feature descriptions, user operation guides, and FAQs. The most appropriate template is selected based on the type and content of the extracted target knowledge point. For example, if the target knowledge point is instructions for registering a new account, a template describing the registration process is selected. The extracted target knowledge point information is then populated into the corresponding template. Variables in the template, such as product names, feature descriptions, and operation steps, can be recognized and replaced to generate specific entry content. Through template engine technology, the filled template is converted into a complete customer service knowledge base entry. The customer service knowledge base entry is similar in form to a common FAQ or operation guide, containing questions and detailed answers.

[0071] All generated customer service knowledge base entries are integrated into a complete customer service knowledge base. This process includes categorizing, sorting, and reviewing the entries to ensure a clear structure, comprehensive content, and easy searchability. Ultimately, all customer service knowledge base entries are integrated into the customer service knowledge base, which is accessible to customer service personnel and users for self-service query, improving service efficiency and user experience.

[0072] By leveraging template engine technology and extracting target knowledge points, we can efficiently and accurately construct structured customer service knowledge base entries. This process automates the generation of customer service knowledge base content, reducing the need for manual editing while ensuring the standardization and consistency of customer service knowledge base entries. By integrating multiple customer service knowledge base entries into a complete customer service knowledge base, we provide standardized guidance for customer service and user inquiries, effectively improving customer service efficiency and user satisfaction.

[0073] In order to ensure that the content of the customer service knowledge base can be dynamically adjusted according to the user's actual experience and needs, after using template engine technology to generate customer service knowledge base entries based on the above-mentioned target knowledge points and obtain a customer service knowledge base composed of the above-mentioned customer service knowledge base entries, the above-mentioned method also includes: obtaining feedback information from users and maintenance personnel of the above-mentioned target products; using a machine learning model to parse the above-mentioned feedback information to obtain parsing results about the above-mentioned feedback information, wherein the above-mentioned parsing results include the problem type and improvement requirements of the above-mentioned target products; and adjusting the parameters in the model for parsing the above-mentioned requirement document and the above-mentioned prototype diagram of the above-mentioned target products based on the above-mentioned parsing results.

[0074] Specifically, after the customer service knowledge base is built, feedback from users and maintenance personnel is continuously collected. This feedback can include evaluations of the accuracy, clarity, and completeness of customer service knowledge base entries, as well as user-generated problems and suggestions, and deficiencies identified by maintenance personnel when managing the knowledge base. Methods for collecting feedback include, but are not limited to, online questionnaires, analysis of user interaction records, and customer service logs. The purpose of obtaining feedback is to continuously optimize the content and structure of the customer service knowledge base, ensuring that it meets the needs of users and maintenance personnel, and improving overall user satisfaction and system efficiency.

[0075] Machine learning models (such as deep learning-based text classification models or sentiment analysis models) are used to parse this feedback, aiming to automatically identify problem types and improvement needs. The model can understand the semantics of the text, identify the emotional tendencies (positive, negative, or neutral) in user feedback, and categorize the feedback content, such as user interface issues, unclear function descriptions, and missing operation steps. First, the feedback information is input into a pre-trained machine learning model. The model deeply analyzes the text to identify keywords and phrases in the user feedback and understand the essence of the feedback. Then, based on the identified content, the model classifies the feedback information into pre-set problem types and uses sentiment analysis to determine the user's attitude towards specific issues (satisfied, dissatisfied, neutral). The results of feedback analysis reveal potential problems in the knowledge base and user improvement needs. This information is used to optimize the parameters of the model that parses the requirements documents and prototype diagrams, thereby improving the quality and efficiency of the customer service knowledge base generation.

[0076] Associate the problem types in the feedback with key steps or parameters in the parsing process of requirements documents and prototype images. For example, if user feedback repeatedly mentions inaccurate descriptions of the location of interface elements, it means that the location recognition parameters in the image processing algorithm need to be adjusted. Based on the analysis of problem types and improvement needs, model parameters can be adjusted automatically or semi-automatically to improve performance in specific aspects. For example, increase the feature point matching threshold in image processing to improve the accuracy of interface element attribute recognition; or adjust the weights of the natural language processing model to make the model more accurate in identifying specific entities or relationships. After the parameters are adjusted, retrain the model with the updated parameters, and improve the model's parsing ability and accuracy through iterative optimization. Training data includes historical requirements documents, prototype images, and feedback information to ensure that the model can adapt to changing user needs and product features.

[0077] By continuously collecting and analyzing feedback from users and maintenance personnel, model parameters can be optimized in a targeted manner, thereby continuously improving the parsing quality of requirements documents and prototype diagrams. This feedback-based learning mechanism can promptly identify and resolve problems, improving the accuracy, completeness, and user satisfaction of the customer service knowledge base. Over time, through continuous learning and adjustment, the model can more accurately understand the information in requirements documents and prototype diagrams, generating knowledge base entries that better meet user needs, forming a continuously self-improving and optimizing intelligent customer service knowledge base automation construction system. This ensures that the content of the customer service knowledge base can be dynamically adjusted based on the actual user experience and needs, thereby improving the level of customer service and the overall user experience.

[0078] In other embodiments of the present application, a context-aware and context-responsive mechanism is introduced. By integrating context-aware technology, it is possible to identify user contextual information based on the user's current usage scenario and status (such as usage time, location, device type, operation history, etc.). For example, if a user uses an app late at night, it can be inferred that the user may have encountered an urgent problem or is looking for instructions on how to use a specific function. Based on the context-aware results, context-responsive rules are formulated, including which knowledge base items are prioritized in specific situations, adjusting the weights of recommendation algorithms, and changing the information presentation method (such as optimizing the layout on mobile devices). The formulation of context-responsive rules aims to maximize the effectiveness and user experience of the customer service knowledge base in different situations. The content and presentation of the customer service knowledge base can be dynamically adjusted according to changes in the user's context, such as highlighting the registration guide and function introduction when the user first logs in, and providing more in-depth operation tips and troubleshooting information when the user frequently uses a certain function. By introducing context-aware and context-responsive mechanisms, this embodiment can significantly improve the practicality and interactivity of the customer service knowledge base. Unlike traditional static knowledge bases, context-aware and context-responsive mechanisms can provide the most relevant and timely information and support based on the user's specific context, greatly improving the efficiency and satisfaction of user problem solving. The application cases of situational awareness and situational response mechanisms are as follows:

[0079] Imagine a user attempting to access the company's enterprise software while on an overseas business trip. Late at night, the user encounters a problem logging into the system. The context-aware module recognizes that the user is using a mobile device and is not working during business hours. This immediately activates appropriate response rules, prioritizing recommendations for overseas login guides, mobile-specific instructions, and emergency contact information. Furthermore, taking into account the non-business hours, detailed self-help options are provided, including answers to frequently asked questions, troubleshooting steps, and links to online documentation.

[0080] Contextual awareness and response mechanisms not only solve current user problems but also prevent similar issues in the future. By identifying user context in real time and dynamically adjusting knowledge base content, we can provide more personalized and contextualized customer support. This mechanism further enhances the intelligent automation of the customer service knowledge base, enabling it to better adapt to user needs in different scenarios and improve overall service quality.

[0081] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the customer service knowledge base construction method of the present application will be described in detail below with reference to specific embodiments.

[0082] This embodiment relates to a specific method for building a customer service knowledge base, such as Figure 3As shown, through a series of refined processes, we can achieve automatic parsing, knowledge extraction, structured construction of requirement specifications, prototype diagrams and business process diagrams, and dynamic updating of the knowledge base.

[0083] 1. Document automatic parsing process:

[0084] Natural language processing (NLP) technology, especially deep learning models such as BERT or GPT, is used to parse requirement specifications (SRS) and business process diagrams (BPDs). This process involves steps such as text preprocessing, entity recognition, and dependency analysis to extract key information from the documents, such as functional requirements, performance indicators, and business rules.

[0085] 2. Image recognition and interface element extraction process:

[0086] Computer vision techniques, such as convolutional neural networks (CNNs), analyze prototype images to automatically identify interface elements (such as buttons and input boxes) and their attributes. This process also includes layout analysis to understand the spatial relationships between interface elements, providing an intuitive interface reference for subsequent knowledge base construction.

[0087] 3. Knowledge extraction and integration process:

[0088] Combining the results of document parsing and image recognition, the GPT large model conducts in-depth analysis of the extracted data to identify and integrate key knowledge points. This process involves technologies such as semantic understanding, knowledge graph construction, and information fusion to form coherent and structured knowledge base entries.

[0089] 4. Structured knowledge base construction process:

[0090] Based on the extracted and integrated knowledge content, we apply template engines such as Jinja2 or Thymeleaf to automatically generate structured knowledge base entries. These templates support a variety of document types, such as FAQs, operation guides, and troubleshooting guides, ensuring the standardization and systematization of knowledge base content.

[0091] 5. Knowledge base dynamic update process:

[0092] The version control system monitors changes in requirements documents and prototypes, automatically triggering the update process. Using differential analysis algorithms, we accurately identify change points and update the knowledge base accordingly, ensuring the timeliness and accuracy of the knowledge base.

[0093] Version Control Integration: By integrating with version control system APIs such as Git, we can monitor version updates of requirements documents and prototype files in real time. When a new version is detected, the changes are automatically extracted.

[0094] Difference Analysis: Use advanced text comparison algorithms and image processing techniques, such as Google's Diff MatchPatch library or OpenCV, to identify changes in requirements documents and prototypes. This includes text additions and deletions, formatting changes, and rearrangement of interface elements.

[0095] Smart content updates: Map these changes to the corresponding entries in the knowledge base and automatically update the interface screenshots and instructions to reflect the new user interface prompts. For example, if the "Register" button in the prototype is moved or redesigned, the corresponding screenshots and instructions in the knowledge base will be automatically updated.

[0096] Dynamic update execution: Automatically update the knowledge base content, including FAQs, operation guides, etc., to ensure that all relevant documents reflect the new functional descriptions. This process ensures that the knowledge base content is kept in sync with the latest status of the product.

[0097] 6. User feedback and continuous learning process:

[0098] Feedback from users and maintainers is collected through a user interface, and machine learning techniques are used to continuously optimize the knowledge extraction and generation algorithms. This feedback loop mechanism enables the system to self-learn and improve, thereby improving the quality and relevance of the knowledge base.

[0099] 7. The following is a specific example showing how to convert the requirements document and prototype diagram of APP login and registration into detailed content of the knowledge base:

[0100] Requirements document: describes the login and registration functionality of the app, including form field requirements, password policies, error messages, etc.

[0101] Prototype: shows the layout of the login and registration page, including input boxes, buttons, icons, etc.

[0102] 7.1 Document parsing: Functional example: Extract login and registration form fields from the requirement document, such as "user name", "password", "confirm password", and "mobile phone number".

[0103] 7.2 Image Recognition: Functional Example: Identify login and registration page elements in the prototype image, such as the "Login" button, the "Forgot Password?" link, and the "Create Account" button on the registration page.

[0104] 7.3 Knowledge Extraction and Integration: Functional Example: Combine document and prototype information to extract key steps of the registration process, such as "Enter mobile phone number", "Set password", and "Agree to terms of service".

[0105] 7.4 Structured Construction: Functional Example: Generate structured registration process entries, question and answer form combinations, such as:

[0106] How to register a new account

[0107] 1. Open the app and click the "Create Account" button.

[0108] 2. Enter your mobile phone number and click "Next".

[0109] 3. Set a password that contains uppercase and lowercase letters and numbers, and is at least 8 characters long.

[0110] 4. Enter your password again to confirm, then tap "Register."

[0111] 5. Read and agree to the Terms of Service and Privacy Policy.

[0112] 7.5 Knowledge Base Update: Function Example: When the requirements document is updated and a "Register with a Foreign Mobile Number" option is added, the knowledge base is automatically updated with the following content: Alternatively, you can enter your foreign mobile number and verify it to complete the registration.

[0113] Original situation:

[0114] The requirements document states that the login and registration module only supports domestic mobile phone numbers in the format of XXX-XXXXXXX.

[0115] The prototype image shows a mobile phone number input box with the prompt text "Please enter your domestic mobile phone number."

[0116] Changes:

[0117] The requirements document has been updated to add support for foreign mobile phone numbers. The format description has been changed to +[country code][mobile phone number].

[0118] The prompt text of the mobile phone number input box in the prototype image is updated to "Please enter your mobile phone number" to adapt to the international format.

[0119] For example, an updated FAQ entry might include: "Q: What format of mobile phone number can I use to register? A: You can register using the domestic format (XXX-XXXXXXX) or the international format (+[country code][mobile phone number)."

[0120] 7.6 Quality Control Module: Functional Example: The auditor checks the accuracy of the "How to Register a New Account" entry, including the correctness of the steps, the consistency between the interface screenshots and prototype images, and the clarity of the language.

[0121] 7.7 Feedback Learning Module: Functional Example: User feedback indicates that password complexity requirements are unclear during registration. The module records the feedback and adjusts the description of password settings in the knowledge base. For example, the module sets a strong password with at least 8 characters, including at least one uppercase letter, one lowercase letter, and one number.

[0122] Through this embodiment, it is possible to automatically extract information from requirement specifications, prototype diagrams, and business process diagrams to generate a structured customer service knowledge base. The GPT large model is used for natural language processing and image recognition to achieve in-depth analysis and knowledge point extraction of document and image data. It is capable of processing and integrating data from different modalities (text and images) to form a coherent and comprehensive knowledge representation. It has the ability to monitor document changes and automatically update the knowledge base to ensure the timeliness and accuracy of the knowledge base content. By collecting feedback from users and maintenance personnel, it is possible to continuously learn and optimize to improve the performance of knowledge extraction and generation algorithms. The automated customer service knowledge base generation process of this embodiment, especially the dynamic update process of the customer service knowledge base, ensures the timeliness and accuracy of the customer service knowledge base content through real-time monitoring of document changes, fine-grained difference analysis, intelligent content mapping and updating, and dynamic update execution.

[0123] This application significantly improves the efficiency of customer service knowledge base construction through automation technology. Automated document parsing and image recognition modules reduce reliance on manual operations, thereby accelerating the generation of the customer service knowledge base and reducing errors caused by human error. Furthermore, the dynamic knowledge base update mechanism promptly reflects the latest product changes, resolving the issue of delayed knowledge base updates in the prior art. Another significant advantage of this application is reduced maintenance costs. In the prior art, customer service knowledge bases require constant manual review and updates with product iterations, which not only increases operating costs but also makes it difficult to ensure timely updates. However, this application, through automated update processes and quality control modules, reduces the need for manual maintenance, thereby reducing long-term costs. Furthermore, this application enhances the user experience through user interaction optimization and feedback learning mechanisms. Using natural language processing technology, this application provides a more intuitive and user-friendly query interface, enabling users to quickly obtain accurate information and enhancing user satisfaction. Finally, this application's cross-modal data processing capabilities and modular design not only enhance the richness and accuracy of the customer service knowledge base content but also facilitate system expansion and maintenance.

[0124] The embodiments of the present application also provide a customer service knowledge base construction device. It should be noted that the customer service knowledge base construction device of the embodiments of the present application can be used to execute the customer service knowledge base construction method provided by the embodiments of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the details that have been described will not be repeated here. As used below, the term "module" can implement a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0125] The following is an introduction to the customer service knowledge base construction device provided in the embodiment of the present application.

[0126] Figure 4 This is a structural block diagram of a customer service knowledge base construction device according to an embodiment of the present application. Figure 4 As shown, the device includes a first acquisition unit 10, an input unit 20, a generation unit 30 and an update unit 40. The first acquisition unit is used to extract first information from the requirement document of the target product and extract second information from the prototype diagram of the target product, wherein the first information includes the functional requirements, performance indicators and business rules of the target product, and the second information includes the interface elements, interface element attributes and interface layout information of the target product; the input unit is used to input the first information and the second information into the pre-trained GPT large model for parsing to obtain the target knowledge points of the target product, wherein the target knowledge points include functional descriptions, operation guides and FAQs; the generation unit is used to use template engine technology to generate customer service knowledge base entries based on the target knowledge points, and obtain a customer service knowledge base composed of the customer service knowledge base entries, wherein the customer service knowledge base entries include questions and answers to the questions; the update unit is used to monitor the requirement document and the prototype diagram, and if the version of the requirement document and / or the prototype diagram is updated, the customer service knowledge base is updated according to the updated requirement document and / or the updated prototype diagram.

[0127] Extract the key information needed to build the customer service knowledge base from different perspectives and data sources (requirements documents and prototypes). The first information provides a description of product functionality and business logic, while the second information provides an intuitive user interface design. Combining the first and second information can generate more comprehensive and accurate knowledge base entries. This method of extracting information from multiple sources ensures the comprehensiveness and accuracy of the customer service knowledge base content and also provides the data foundation for subsequent automated update processes.

[0128] The first and second information are fed into a pre-trained GPT model, enabling it to fully understand the product's design intent and user interface. The pre-trained GPT model, trained on massive amounts of text data, possesses powerful language comprehension and generation capabilities. During the construction of the customer service knowledge base, the pre-trained GPT model can understand functional requirements and business rules, translating them into concise and clear functional descriptions.

[0129] By inputting the structured first information and second information into the pre-trained GPT large model for parsing, the target knowledge points of the customer service knowledge base can be generated, including function descriptions, operation guides and FAQs. This process makes full use of the natural language understanding and generation capabilities of the GPT large model, providing an efficient and accurate method for the construction of the customer service knowledge base.

[0130] In this embodiment, template engine technology is used to generate customer service knowledge base entries. Typically, templates contain fixed structures and variable placeholders, while data is the actual content extracted from the target knowledge point. The target knowledge point is generated by a pre-trained GPT large model analyzing the requirements document and prototype diagrams. It includes functional descriptions, operating instructions, and FAQs. A functional description is a brief description of a specific product feature; an operating instruction is a detailed step-by-step guide explaining how to use the product or its features; and an FAQ is a pre-set collection of questions and answers that answer common questions users may encounter.

[0131] Through continuous monitoring and automatic updates, the customer service knowledge base is always synchronized with the latest status of product requirements documents and prototypes. Furthermore, the automated update process reduces the need for manual monitoring and updates, saving time and resources. Change points can be precisely identified and all relevant entries in the customer service knowledge base are updated, avoiding information omissions or inconsistencies. Automated updates to the customer service knowledge base enable real-time response to changes in product requirements documents and prototypes, ensuring the timeliness and accuracy of the customer service knowledge base content, and improving customer service quality and user experience.

[0132] Through this embodiment, the target product's requirement documents and prototype drawings are acquired and parsed, and combined with the deep analysis capabilities of the GPT large model, accurate target knowledge points are efficiently generated. Then, a standardized customer service knowledge base is quickly constructed using template engine technology, and version updates of documents and prototype drawings are monitored. This can trigger automatic updates of the knowledge base in real time, ensuring the timeliness and accuracy of the content, thereby solving the limitations of the construction of customer service knowledge bases in the prior art, including low accuracy, low degree of automation, and delayed updates.

[0133] In the specific implementation process, the update unit includes a monitoring module, an extraction module, a determination module, a first parsing module and an update module. Among them, the monitoring module is used to monitor the above-mentioned requirement document and the above-mentioned prototype diagram through the version control system interface; the extraction module is used to extract the above-mentioned updated requirement document and / or the above-mentioned updated prototype diagram when it is detected that the version of the above-mentioned requirement document and / or the above-mentioned prototype diagram has been updated; the determination module is used to use a difference analysis algorithm to determine the changed content of the above-mentioned requirement document and / or the above-mentioned prototype diagram, and the above-mentioned changed content includes the requirement document version change information and / or the prototype diagram version change information; the first parsing module is used to parse the above-mentioned updated requirement document and / or the above-mentioned updated prototype diagram according to the above-mentioned changed content using the above-mentioned pre-trained GPT large model to generate updated target knowledge points; the update module is used to update the above-mentioned customer service knowledge base according to the above-mentioned updated target knowledge points.

[0134] Through the above process, we can accurately identify changes in requirements documents and prototypes. Whether it's subtle text additions or deletions or formatting adjustments in requirements documents, or significant changes to interface elements or layout updates in prototypes, all can be captured efficiently and accurately. This change information is then used in the parsing process of the pre-trained GPT large model to generate updated target knowledge points. Ultimately, this ensures the dynamic updating of the customer service knowledge base and the timeliness of its content, thereby improving customer service quality and user satisfaction.

[0135] After determining the specific changes in the requirements document and / or prototype diagram, the changes are input into the pre-trained GPT large model. The pre-trained GPT large model parses the changes and generates updated target knowledge points. Based on the generated updated target knowledge points, the customer service knowledge base is updated in real time to reflect the latest changes in the product documentation.

[0136] By integrating the version control system interface, this embodiment can monitor the version updates of requirement documents and prototype drawings in real time. Once an update is detected, it automatically extracts the latest requirement documents and prototype drawings, and uses a differential analysis algorithm to accurately identify the changed content. Based on this change information, the pre-trained GPT large model performs in-depth analysis to generate updated target knowledge points, including new function descriptions, operation guides, and FAQs. Ultimately, based on these updated knowledge points, the customer service knowledge base is updated in a timely and accurate manner to ensure that it is always synchronized with the product documents it serves, thereby significantly improving customer service efficiency and user experience, while reducing maintenance costs and error rates.

[0137] In some embodiments of the present application, the first acquisition unit includes a second parsing module and a third parsing module. The second parsing module is configured to parse the requirements document of the target product using a natural language processing algorithm, such as a BERT model or a GPT model, to obtain the first information; and the third parsing module is configured to parse the prototype image of the target product using an image processing algorithm, such as a SIFT algorithm or a YOLO algorithm, to obtain the second information.

[0138] Specifically, the requirements document is input into the BERT model or the GPT model. The model uses deep learning technology to perform semantic parsing on the requirements document and identify key entities, operating instructions, error messages, user interface descriptions and other information in the requirements document. The first information extracted includes statements describing product functions, operating steps, performance indicators, text descriptions of interface elements, etc. The prototype image is preprocessed using OpenCV to obtain a preprocessed prototype image. The SIFT algorithm, SURF algorithm or ORB algorithm is applied to the preprocessed prototype image to detect and extract the feature points of the interface elements. Based on the feature points, the algorithm calculates descriptors. These descriptors contain attribute information such as the shape and texture of the interface elements. The extracted interface element attributes include the size, shape, color, position coordinates, etc. of the elements.

[0139] Pre-trained object detection models, such as the YOLO algorithm or other models based on convolutional neural networks, can identify objects of specific categories in images. Using a pre-trained YOLO or convolutional neural network to scan the pre-processed prototype image, the algorithm identifies various elements in the interface, such as buttons, input boxes, and labels. The algorithm locates each element and assigns a category label, resulting in a detailed list of interface elements, including their location and type.

[0140] Through graph theory and network analysis techniques, a spatial relationship diagram between elements can be constructed, including the hierarchical structure, relative positions, and potential interactions between elements. The identified interface elements are regarded as nodes in the graph, and the spatial relationships between nodes (such as up and down, left and right, and included) are regarded as edges. Graph theory algorithms, such as breadth-first search and depth-first search, are applied to analyze the hierarchical structure and layout relationships of interface elements. Network analysis methods are used to understand the interaction logic between elements, such as the association between buttons and input boxes, and the functional relationships between element groups.

[0141] The combination of preprocessing, feature extraction, element recognition, and layout analysis enables a comprehensive understanding of the interface design in the prototype image, including the attributes and location of each interface element, as well as the layout and interaction relationships between them. This information forms an important basis for building a customer service knowledge base, ensuring the precise matching and description of the customer service knowledge base content with the product interface. Through steps such as image preprocessing, feature extraction, object detection, and layout analysis, an in-depth analysis of the prototype image is achieved, extracting the detailed attributes, location, and layout information of the interface elements. This is one of the keys to building a customer service knowledge base. Through this series of processing, an accurate and comprehensive interface description can be formed, providing users with more detailed operation guides and product information.

[0142] To obtain the first piece of information, structured information extraction is performed first. Structured information extraction converts unstructured text data into a format that is easier to process and analyze. For the requirements document, this means splitting and extracting the document according to the hierarchical structure of titles and paragraphs. Using a natural language processing algorithm, the requirements document is segmented according to titles and paragraphs. Each title or paragraph is considered an information segment, and key information is extracted from each information segment, such as functional descriptions, operating procedures, performance indicators, etc. This information will serve as the basis for subsequent analysis. Text preprocessing aims to standardize text data and remove noise so that subsequent analysis can be performed more accurately. The extracted structured information is preprocessed by formatting the text to unify the format, such as removing excess spaces and standardizing punctuation, to ensure the consistency of the text data. De-noising is used to remove irrelevant information in the text, such as comments, repeated paragraphs, or text unrelated to the requirements, so that the algorithm can better focus on key information.

[0143] Perform text analysis on the preprocessed structured information to obtain text information. Text analysis aims to extract deeper insights from the preprocessed structured information, including identifying specific patterns and matching keywords. Regular expressions are used to search for specific patterns in the text, such as phone numbers, dates, specific commands, or functional descriptions, to quickly locate and extract structured data. Keywords related to product features, performance, or interface, such as "register," "login," and "search," are identified and further refined through keyword matching.

[0144] Information extraction is performed on text information to obtain extracted information. Information extraction is the conversion of the results of text analysis (text information) into structured data, including the identification of key information, entities, and relationships. Information extraction includes the extraction of key information, entity recognition, and relationship extraction. Among them, the extraction of key information is to determine the core information points in the document, such as the description of the user interface, the details of the functional process, etc., based on the results of regular expression extraction and keyword matching. Entity recognition is the use of natural language processing technology to identify entities in the document, such as function names, performance parameters, user roles, etc. These entities are the basis for building a customer service knowledge base. Relationship extraction is the analysis of the relationship between entities, such as the interaction between functions and user roles, and the association between performance indicators and operations. These relationships are the key to understanding the content of the document.

[0145] Natural language processing algorithms are used to deeply analyze the text and extracted information to gain a deeper understanding and obtain the initial first information. Specifically, deep learning models such as BERT or GPT are used to parse the text and extracted information, understanding the contextual meaning of the text and the relationships between entities. The algorithms can identify the underlying meaning of the text and parse complex statements to ensure a comprehensive interpretation of the requirements document. The parsed initial first information is formatted into a standardized structure and converted into a unified format such as JSON or XML to obtain the first information.

[0146] By using natural language processing algorithms (BERT or GPT models) to parse requirements documents and image processing algorithms (SIFT and YOLO algorithms) to parse prototype images, key information can be automatically, efficiently, and accurately extracted from requirements documents and prototype images, providing a solid data foundation for the automated construction of the customer service knowledge base. This information includes not only product functional descriptions and performance requirements, but also the detailed location and style of user interface elements, forming a comprehensive material for building the customer service knowledge base. This improves the efficiency of building the customer service knowledge base, information accuracy, and ease of maintenance, while also significantly enhancing user experience and interaction satisfaction.

[0147] In some other embodiments of the present application, the generation unit includes a selection module, a filling module, and an integration module. The selection module is used to utilize the template engine technology to select corresponding templates according to different target knowledge points, wherein the templates include titles and contents, and the template engine technology is Jinja2 or Thymeleaf; the filling module is used to fill the information of the target knowledge point into the corresponding template of the target knowledge point to generate the customer service knowledge base entry; and the integration module is used to integrate all the customer service knowledge base entries to obtain the customer service knowledge base composed of the customer service knowledge base entries.

[0148] By leveraging template engine technology and extracting target knowledge points, we can efficiently and accurately construct structured customer service knowledge base entries. This process automates the generation of customer service knowledge base content, reducing the need for manual editing while ensuring the standardization and consistency of customer service knowledge base entries. By integrating multiple customer service knowledge base entries into a complete customer service knowledge base, we provide standardized guidance for customer service and user inquiries, effectively improving customer service efficiency and user satisfaction.

[0149] To ensure that the customer service knowledge base content can be dynamically adjusted based on users' real-world experiences and needs, the device also includes a second acquisition unit, a parsing unit, and an adjustment unit. The second acquisition unit is used to obtain feedback from users and maintenance personnel of the target product; the parsing unit is used to parse the feedback using a machine learning model to obtain parsed results regarding the feedback, including the problem types and improvement requirements for the target product; and the adjustment unit is used to adjust parameters in the model used to parse the requirements document and prototype diagram of the target product based on the parsing results.

[0150] By continuously collecting and analyzing feedback from users and maintenance personnel, model parameters can be optimized in a targeted manner, thereby continuously improving the parsing quality of requirements documents and prototype diagrams. This feedback-based learning mechanism can promptly identify and resolve problems, improving the accuracy, completeness, and user satisfaction of the customer service knowledge base. Over time, through continuous learning and adjustment, the model can more accurately understand the information in requirements documents and prototype diagrams, generating knowledge base entries that better meet user needs, forming a continuously self-improving and optimizing intelligent customer service knowledge base automation construction system. This ensures that the content of the customer service knowledge base can be dynamically adjusted based on the actual user experience and needs, thereby improving the level of customer service and the overall user experience.

[0151] The customer service knowledge base construction device includes a processor and memory. The first acquisition unit, input unit, generation unit, and update unit are all stored as program units in the memory. The processor executes the program units stored in the memory to implement the corresponding functions. The above modules are all located in the same processor; alternatively, the above modules can be located in different processors in any combination.

[0152] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0153] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program. When the program is run, the device where the computer-readable storage medium is located is controlled to execute the customer service knowledge base construction method.

[0154] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the above-mentioned customer service knowledge base construction method are implemented.

[0155] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the steps of the above-mentioned customer service knowledge base construction method.

[0156] Obviously, those skilled in the art will appreciate that the various modules or steps of the present invention described above can be implemented using a general-purpose computing device, can be centralized on a single computing device, or can be distributed across a network of multiple computing devices. They can be implemented using program code executable by the computing device, and thus, can be stored in a storage device and executed by the computing device. In some cases, the steps shown or described herein can be performed in a different order than that shown, or can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0157] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0158] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0159] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0160] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0161] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0162] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0163] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0164] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0165] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for constructing a customer service knowledge base, characterized in that: include: Extracting first information from a target product's requirements document and extracting second information from a prototype diagram of the target product, wherein the first information includes functional requirements, performance indicators, and business rules of the target product, and the second information includes interface elements, interface element attributes, and interface layout information of the target product; Input the first information and the second information into the pre-trained GPT large model for parsing to obtain target knowledge points of the target product, wherein the target knowledge points include function description, operation guide and FAQ; Generate customer service knowledge base entries based on the target knowledge points using template engine technology to obtain a customer service knowledge base composed of the customer service knowledge base entries, wherein the customer service knowledge base entries include questions and answers to the questions; The requirement document and the prototype diagram are monitored, and if the version of the requirement document and / or the prototype diagram is updated, the customer service knowledge base is updated according to the updated requirement document and / or the updated prototype diagram.

2. The method according to claim 1, characterized in that Monitoring the requirements document and the prototype diagram, and if the versions of the requirements document and / or the prototype diagram are updated, updating the customer service knowledge base according to the updated requirements document and / or the updated prototype diagram, including: Monitoring the requirement document and the prototype diagram through a version control system interface; When it is detected that the version of the requirement document and / or the prototype diagram is updated, extracting the updated requirement document and / or the updated prototype diagram; Determine the changed content of the requirement document and / or the prototype diagram using a difference analysis algorithm, wherein the changed content includes version change information of the requirement document and / or version change information of the prototype diagram; According to the changed content, use the pre-trained GPT large model to parse the updated requirement document and / or the updated prototype diagram to generate updated target knowledge points; The customer service knowledge base is updated according to the update target knowledge point.

3. The method according to claim 2, characterized in that Determine the changed content of the requirement document and / or the prototype diagram using a difference analysis algorithm, wherein the changed content includes the requirement document version change information and / or the prototype diagram version change information, including: Using a Diff Match Patch library to compare the requirement document with the updated requirement document, and identifying version change information of the requirement document, wherein the version change information of the requirement document includes text addition and deletion information and format change information; and / or, The prototype image and the updated prototype image are compared using the OpenCV library to identify the prototype image version change information, wherein the prototype image version change information includes interface element change information and interface layout change information.

4. The method according to claim 1, wherein Extracting first information from a target product requirement document and extracting second information from a prototype diagram of the target product include: Parsing the requirement document of the target product using a natural language processing algorithm to obtain the first information, the natural language processing algorithm including a BERT model or a GPT model; The prototype image of the target product is parsed using an image processing algorithm to obtain the second information, where the image processing algorithm includes a SIFT algorithm and a YOLO algorithm.

5. The method according to claim 4, characterized in that The prototype image of the target product is parsed using an image processing algorithm to obtain the second information, wherein the image processing algorithm includes a SIFT algorithm and a YOLO algorithm, including: Performing an image preprocessing operation on the prototype image to obtain a preprocessed prototype image, wherein the image preprocessing operation includes image enhancement, cropping and scaling; Performing feature extraction on the preprocessed prototype image using a feature extraction algorithm to obtain the interface element attributes in the second information, the feature extraction algorithm including the SIFT algorithm, SURF algorithm or ORB algorithm; Performing element recognition on the preprocessed prototype image using a pretrained target detection model to obtain the interface element in the second information, wherein the pretrained target detection model includes the YOLO algorithm or the convolutional neural network algorithm; Layout analysis of the interface elements is performed using graph theory and network analysis methods to obtain the interface layout information in the second information.

6. The method according to claim 4, characterized in that The requirement document of the target product is parsed using a natural language processing algorithm to obtain the first information, including: Parsing the requirements document using the natural language processing algorithm to extract structured information, wherein the structured information includes information extracted by splitting the requirements document into titles and paragraphs; Performing a text preprocessing operation on the structured information to obtain preprocessed structured information, wherein the text preprocessing operation includes formatting and denoising; Performing text analysis on the preprocessed structured information to obtain text information, wherein the text analysis includes regular expression extraction and keyword matching; Performing information extraction on the text information to obtain extracted information, wherein the information extraction includes extracting key information, entity recognition, and relationship extraction; Parsing the text information and the extracted information using the natural language processing algorithm to obtain initial first information; The initial first information is formatted to obtain the first information.

7. The method according to claim 1, characterized in that Using template engine technology, customer service knowledge base entries are generated according to the target knowledge points to obtain a customer service knowledge base composed of the customer service knowledge base entries, wherein the customer service knowledge base entries include questions and answers to the questions, including: Utilizing the template engine technology, selecting corresponding templates according to different target knowledge points, wherein the templates include titles and contents, and the template engine technology is Jinja2 or Thymeleaf; Filling the target knowledge point information into the corresponding template of the target knowledge point to generate the customer service knowledge base entry; All of the customer service knowledge base entries are integrated to obtain the customer service knowledge base composed of the customer service knowledge base entries.

8. The method according to claim 1, characterized in that After generating customer service knowledge base entries based on the target knowledge points using template engine technology to obtain a customer service knowledge base composed of the customer service knowledge base entries, the method further includes: Obtaining feedback from users and maintenance personnel of the target product; Parsing the feedback information using a machine learning model to obtain parsing results regarding the feedback information, wherein the parsing results include problem types and improvement requirements for the target product; According to the parsing result, parameters in the model for parsing the requirement document and the prototype diagram of the target product are adjusted.

9. A customer service knowledge base construction device, characterized in that: include: a first acquisition unit, configured to extract first information from a target product requirement document and second information from a prototype diagram of the target product, wherein the first information includes functional requirements, performance indicators, and business rules of the target product, and the second information includes interface elements, interface element attributes, and interface layout information of the target product; An input unit, configured to input the first information and the second information into a pre-trained GPT large model for parsing to obtain target knowledge points of the target product, wherein the target knowledge points include function descriptions, operation guides, and FAQs; a generating unit, configured to generate customer service knowledge base entries based on the target knowledge points using a template engine technology, and obtain a customer service knowledge base composed of the customer service knowledge base entries, wherein the customer service knowledge base entries include questions and answers to the questions; An updating unit is configured to monitor the requirement document and the prototype diagram, and if the version of the requirement document and / or the prototype diagram is updated, update the customer service knowledge base according to the updated requirement document and / or the updated prototype diagram.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the customer service knowledge base construction method according to any one of claims 1 to 8.

Images