Test question data generation method, electronic equipment, storage medium and electronic equipment

Abstract

The invention discloses a test question data generation method, electronic equipment, a storage medium and electronic equipment, and relates to the field of large model technology and text data processing. The method comprises the steps that in response to a first input instruction acting on an operation interface, a target document corresponding to the input instruction is displayed on the operation interface, and the target document comprises multi-modal data; in response to a conversion instruction acting on the operation interface, displaying target text data corresponding to the conversion instruction on the operation interface; in response to a second input instruction acting on the operation interface, a generation condition corresponding to the second input instruction is displayed on the operation interface, and the generation condition comprises at least one of a target type, a target difficulty level and a target dimension; and in response to a generation instruction acting on the operation interface, displaying the target test question data on the operation interface. The technical problem of low efficiency and diversity of test question data generation in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the fields of large model technology and text data processing, and specifically, to a method for generating test data, an electronic device, a storage medium and an electronic device. Background Art

[0002] Use computer technology and algorithms to generate various types of questions and question-related data, that is, to generate test question data. The generated test question data can be used in various scenarios such as education and training, test evaluation, and intelligent question-answering systems. Educational institutions can use test question generation technology to generate a large number of practice questions to help students conduct independent learning and review; various examination institutions can use test question generation technology to generate various types of test questions for examinations and evaluations; some intelligent question-answering systems can use test question generation technology to generate question banks to help users quickly obtain information and solve problems.

[0003] At present, in related technologies, test data are usually generated manually, which requires a lot of human resource costs, and the generated test data has a low coverage rate of the content to be tested, and the diversity of the test data is low, resulting in low efficiency and diversity of generated test data.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide a method for generating test data, an electronic device, a storage medium and an electronic device, so as to at least solve the technical problems of low efficiency and low diversity in generating test data in related technologies.

[0006] According to one aspect of an embodiment of the present application, a method for generating test question data is provided, the method comprising: in response to a first input instruction acting on an operation interface, displaying a target document corresponding to the input instruction on the operation interface, wherein the target document comprises multimodal data; in response to a conversion instruction acting on the operation interface, displaying target text data corresponding to the conversion instruction on the operation interface, wherein the target text data is text data obtained by parsing multimodal data using a document parsing large model; in response to a second input instruction acting on the operation interface, displaying a generation condition corresponding to the second input instruction on the operation interface, wherein the generation condition comprises at least one of the following: target type, target difficulty level, target dimension; in response to a generation instruction acting on the operation interface, displaying target test question data on the operation interface, wherein the target test question data is data generated by inputting the target text data into a test question generation large model and guiding the test question generation large model using the generation condition.

[0007] According to another aspect of an embodiment of the present application, a method for generating test question data is also provided, the method comprising: obtaining a target document by calling a first interface, wherein the first interface includes a first parameter, the parameter value of the first parameter includes the target document, and the target document includes multimodal data; parsing the multimodal data using a document parsing large model to obtain target text data; inputting the target text data into a test question generation large model, and generating target test question data using the test question generation large model; outputting the target test question data by calling a second interface, wherein the second interface includes a second parameter, the parameter value of the second parameter includes the target test question data.

[0008] According to another aspect of the embodiments of the present application, an electronic device is further provided, including: a memory storing an executable program; and a processor for running the program, wherein the method of each embodiment of the present application is executed when the program is running.

[0009] According to another aspect of an embodiment of the present application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in each embodiment of the present application.

[0010] According to another aspect of the embodiments of the present application, a computer program product is also provided, including a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present application is implemented.

[0011] According to another aspect of an embodiment of the present application, a computer program product is also provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present application is implemented.

[0012] According to another aspect of the embodiments of the present application, a computer program is also provided, and when the computer program is executed by a processor, the methods in the various embodiments of the present application are implemented.

[0013] In an embodiment of the present application, a method for generating test question data is provided, the method comprising: in response to a first input instruction acting on an operation interface, displaying a target document corresponding to the input instruction on the operation interface, wherein the target document includes multimodal data; in response to a conversion instruction acting on the operation interface, displaying target text data corresponding to the conversion instruction on the operation interface, wherein the target text data is text data obtained by parsing multimodal data using a document parsing large model; in response to a generation instruction acting on the operation interface, displaying target test question data on the operation interface, wherein the target test question data is data generated by inputting the target text data into a test question generation large model using the test question generation large model. It is easy to notice that the target document in the present application uses multimodal data to enrich the content of the document and ensure the diversity of the generated test data; in the process of generating test data, the text data obtained by parsing the multimodal data in the target document using the document parsing model can help realize the conversion from multimodal data to text data. The application of the document parsing model can improve the system's processing ability for multimodal data, so that the system can better parse and utilize multiple data sources, and provide more comprehensive and accurate information for subsequent test generation; the user specifies the generation conditions by entering the second instruction, and the system can display the generation conditions so that the user can clearly understand the generation conditions of the test. The requirements of the test data can be determined by the second input instruction input by the user, and the specific test question generation conditions can be determined, so that the subsequent system can generate test question data that meets the user's requirements based on the generation conditions specified by the user; the target test question data is generated by inputting the target text data into the test question generation large model. The test question data generated by the test question generation large model can ensure the quality and diversity of the test questions and improve the efficiency of generating test question data. Through the above steps, the user can quickly generate diverse and high-quality test question data through simple input instructions, improve the efficiency and diversity of generating test question data, and thus solve the technical problems of low efficiency and diversity in generating test question data in related technologies.

[0014] It is easy to notice that the above general description and the following detailed description are only for the purpose of exemplifying and explaining the present application, and do not constitute a limitation of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0016] Figure 1 It is a schematic diagram of an application scenario of a method for generating test question data according to an embodiment of the present application;

[0017] Figure 2is a flow chart of a method for generating test question data according to an embodiment of the present application;

[0018] Figure 3 is a flowchart of an optional method for generating test question data according to an embodiment of the present application;

[0019] Figure 4 is a schematic diagram of a configuration window in another optional operation interface according to an embodiment of the present application;

[0020] Figure 5 is a flowchart of an optional method for generating test question data according to an embodiment of the present application;

[0021] Figure 6 is a schematic diagram of a system timing of an optional method for generating test question data according to an embodiment of the present application;

[0022] Figure 7 It is a schematic diagram of the interaction of functional modules of an optional method for generating test question data according to an embodiment of the present application;

[0023] Figure 8 is a flow chart of another method for generating test question data according to an embodiment of the present application;

[0024] Fig. 9 is a schematic diagram of a device for generating test question data according to an embodiment of the present application;

[0025] Fig.10 is a schematic diagram of another device for generating test question data according to an embodiment of the present application;

[0026] Fig.11 It is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0028] 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 sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. 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 comprising 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.

[0029] The technical solution provided in this application is mainly implemented by large-scale model technology. The large model here refers to a deep learning model with large-scale model parameters, which can usually contain hundreds of millions, tens of billions, hundreds of billions, trillions or even more than ten trillion model parameters. The large model can also be called a cornerstone model / foundation model (Foundation Model). The large model is pre-trained through large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks, and the model has good generalization ability, such as large-scale language model (Large Language Model, LLM), multi-modal pre-training model (multi-modal pre-training model), etc.

[0030] It should be noted that when the large model is actually applied, the pre-trained model can be fine-tuned through a small number of samples so that the large model can be applied to different tasks. For example, the large model can be widely used in natural language processing (NLP), computer vision, speech processing and other fields, specifically in computer vision tasks such as visual question answering (VQA), image caption (IC), image generation, etc. It can also be widely used in natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. Therefore, the main application scenarios of the large model include but are not limited to digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc.

[0031] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following explanations:

[0032] A large-scale language model can be an artificial intelligence model based on deep learning technology. By training massive text data, it can understand and generate human-like natural language. It usually has billions or even hundreds of billions of parameters and is widely used in automated customer service, content creation, language translation and other fields. It has strong language processing capabilities and adaptability. For example, it can include OpenAI's GPT series and Google's BERT.

[0033] Prompt engineering is a technology for optimizing and designing input prompts. It guides the model to generate more accurate, relevant and high-quality outputs by cleverly constructing and adjusting the input text of a given language model. This technology plays an important role in improving the performance of the model in specific tasks such as text generation, translation, question and answer, etc. It is widely used in automated content generation, customer service, intelligent question and answer, and other fields, helping to achieve more efficient and intelligent language processing systems.

[0034] Multimodal models are an artificial intelligence technology that uses a variety of data types such as text, images, audio, and video for learning and reasoning. They aim to provide more comprehensive and accurate solutions by integrating information from different modalities. These models are widely used in medical diagnosis, autonomous driving, content creation and other fields, greatly improving task performance and application value.

[0035] An agent is an autonomous software entity that has the ability to perceive the environment, make decisions, and execute actions. It can sense external information through sensors and take actions autonomously according to goals and environmental changes, such as adjusting device settings and sending information. It is widely used in intelligent virtual assistants, autonomous driving systems, financial transactions, game AI, and robots. Agents play a key role in improving system intelligence and automation.

[0036] Artificial Intelligence Generated Content (AIGC) has broad application prospects in various industries, especially in the fields of content creation, education and training, film and television entertainment, etc. By reducing the workload and time cost of human beings in content creation, AIGC not only improves production efficiency, but also creates many new possibilities.

[0037] Sys Prompt refers to the internal prompts or guidance information used by the system or model when working. It usually sets the behavior, style or specific context of the model. For example, in a pre-trained model, the system prompt may be used to tell the model its role or expected response method. This prompt is usually not directly seen by the user, but it has an important impact on the effect of the model output.

[0038] User prompts can refer to requests or questions that users directly input to the model. User prompts usually contain specific information that users want to obtain or problems that need to be solved. The model will generate a response based on this prompt. User prompts are the main way for people to interact with machines, and directly affect the specificity and accuracy of the output content.

[0039] The above-mentioned system prompts refer to the background or rules set by the system to guide the generation of responses, while user prompts are the specific content of user input. The combination of the two can determine the final output of the artificial intelligence model.

[0040] Prompt slots refer to specific information locations that need to be filled in the dialogue system, such as time, place, and budget. By identifying these slots, the system can extract user needs, maintain the dialogue context, and generate personalized responses. Slots can be divided into required and optional, which affects the system's understanding of user intent and the accuracy of the response.

[0041] According to an embodiment of the present application, a method for generating test data 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.

[0042] Considering the huge number of model parameters of the large model and the limited computing resources of the mobile terminal, the method for generating the test data provided in the embodiment of the present application can be applied to the following examples: Figure 1 The application scenarios shown are not limited to these. Figure 1 In the application scenario shown, the large model is deployed in the server 10, and the server 10 can be connected to one or more client devices 20 through a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. The client device 20 here may include but is not limited to: a smart phone, a tablet computer, a laptop computer, a PDA, a personal computer, a smart home device, a vehicle-mounted device, etc. The client device 20 can interact with the user through a graphical user interface to implement the call of the large model, thereby implementing the method provided in the embodiment of the present application.

[0043] In an embodiment of the present application, the system composed of a client device and a server can perform the following steps: the client device can generate a first input instruction, a conversion instruction, a second input instruction, and a generation instruction through a graphical user interface, and display a target document, target text data, generation conditions, and target test data in the operation interface; the server can obtain a target document containing multimodal data, convert the target document into target text data using a document parsing model, determine the generation condition corresponding to the second input instruction, and use the generation condition to guide the test question generation model, and generate the target test question data from the target text data. It should be noted that, in the case where the operating resources of the client device can meet the deployment and operating conditions of the large model, the embodiment of the present application can be performed in the client device.

[0044] Under the above operating environment, this application provides Figure 2 The method of generating test data shown in FIG. Figure 2 FIG. 1 is a flow chart of a method for generating test data according to an embodiment of the present application. Figure 2 As shown, the following steps may be specifically included:

[0045] Step S202, in response to a first input instruction applied to the operation interface, a target document corresponding to the input instruction is displayed on the operation interface.

[0046] Among them, the target document contains multimodal data.

[0047] The above-mentioned operation interface may refer to a graphical interface or command line interface used when a user interacts with a computer system. The operation interface may include various menus, buttons, text boxes, icons and other elements that a user can see and operate. The user performs various operations and tasks through the operation interface to communicate and interact with the computer system. The operation interface can be displayed in various mobile terminals, desktop computers, server-side operation / management systems and other devices. The operation interface can also be set according to actual needs, which is not limited here.

[0048] The above-mentioned first input instruction may refer to an operation instruction sent by a user to a computer system through an input device such as a keyboard, mouse or touch screen, which is used to start the task of executing the target document corresponding to the display input instruction. The first input instruction may be a specific command, keyword or action, which is used to indicate the operation that the computer system user wishes to perform. The form of the first input instruction can be determined according to actual needs and is not limited here.

[0049] The above-mentioned multimodal data may refer to a data set containing multiple different types of data information, such as text, image, audio, video, etc. The specific form of the multimodal data may also be determined according to actual needs and is not limited here.

[0050] The above-mentioned target document may refer to a document to be analyzed and processed when generating test data, which is presented in the form of multimodal data such as text, image, audio, video, etc. The target document may adopt a knowledge base system (KBS) document. The KBS system may be a system for managing and organizing knowledge. The KBS document contains a large amount of knowledge and information and can be used to generate education-related content such as test questions and answers. The KBS document is written by experts or domain knowledge personnel and contains knowledge in various fields, such as history, geography, mathematics, science, etc. In the process of generating test data, the knowledge points and information in the KBS document can be analyzed to extract relevant content and generate test content that meets educational requirements. By using the KBS document, a large amount of high-quality test data can be effectively generated to help educational training institutions improve teaching efficiency and quality. The target document can also be determined according to actual needs, which is not limited here.

[0051] In an optional embodiment, the user can input a first input instruction into the operation interface by simulating an operation on the operation interface. The system can determine the corresponding target document based on the first input instruction and display the target document. The displayed target document can be a text document, a picture, a video or other multimodal data form. The determination is made here according to actual needs. Responding to the first input instruction on the operation interface and displaying the corresponding target document can help the user find the required information more quickly and improve user experience and efficiency. When the user enters the first instruction on the operation interface, the system can immediately respond and display the corresponding target document. This can be achieved by real-time monitoring of user input and automatically matching corresponding documents. The system can determine the target document corresponding to the instruction input by the user through keyword matching, semantic analysis and other technologies. The target document of multimodal data can also enrich the content of the document, so that the user can have a more comprehensive understanding of the information.

[0052] Step S204, responding to the conversion instruction applied to the operation interface, and displaying the target text data corresponding to the conversion instruction on the operation interface.

[0053] The target text data is the text data obtained by parsing the multimodal data using a large document parsing model.

[0054] The above-mentioned conversion instructions may refer to instructions for controlling the conversion of the target document into target text data generated by the user interacting with the operation interface through clicking, selecting or speaking. The form of the conversion instructions can be determined according to the specific interaction method between the user and the operation interface, and is not limited here.

[0055] The above-mentioned document parsing large model may refer to a deep learning model that can parse and convert multimodal data with complex formats and difficult to process into easy-to-understand and subsequent processing. The document parsing large model may adopt a multimodal model. The multimodal model may refer to a model that combines a variety of different input data sources, such as text, images, audio, etc. for learning and prediction. The multimodal model can process different types of data and learn more comprehensive and accurate information from it, thereby improving the performance and predictive ability of the model. The multimodal model has a wide range of applications in natural language processing, computer vision, speech recognition and other fields. The document parsing large model can also be determined according to actual needs, which is not limited here.

[0056] In an optional embodiment, in response to a conversion instruction from a user on an operation interface, target text data corresponding to the conversion instruction is displayed on the operation interface. The target text data may be text data obtained by parsing multimodal data using a document parsing model. The document parsing model may be a deep learning model that can process multimodal data, such as images, videos, audio, etc., and convert them into text data. The document parsing model may include multiple layers, such as convolutional neural networks, recurrent neural networks, attention mechanisms, etc., for processing different types of data and extracting features. By training the document parsing model, it can learn the correlation between different types of data, thereby converting multimodal data into text data. The specific process may include inputting multimodal data into the document parsing model, the model will process the data and extract features, and then convert the data into text data through the neural network. The final text data output is calculated. In the process of generating test data, the text data obtained by parsing the multimodal data using the document parsing large model can help realize the conversion from multimodal data to text data, and provide basic data for subsequent test generation. The user enters or selects the conversion instruction on the operation interface, and the system receives and responds to the conversion instruction, and displays the corresponding target text data on the operation interface for the user to view. Through the above steps, the user can obtain the target text data generated by the system through the conversion instruction on the operation interface, realizing the interaction and information transmission between the user and the system. The application of the document parsing large model also improves the system's processing capabilities for multimodal data, allowing the system to better parse and utilize multiple data sources, provide users with more comprehensive and accurate information, enhance user experience, and improve the functionality and practicality of the system.

[0057] Step S206, in response to the second input instruction acting on the operation interface, displaying the generation condition corresponding to the second input instruction on the operation interface.

[0058] The generation conditions include at least one of the following: target type, target difficulty level, and target dimension.

[0059] The above-mentioned second input instruction may refer to an operation instruction sent by a user to the computer system through an input device such as a keyboard, mouse or touch screen, which is used to start the task of executing the display generation conditions. The second input instruction may be a specific command, keyword or action used to indicate the operation that the computer system user wishes to perform. The form of the second input instruction can be determined according to actual needs and is not limited here.

[0060] The above-mentioned target type may refer to the expected test question type of the test question data that needs to be generated. The target type may include but is not limited to single-choice, multiple-choice, fill-in-the-blank, question-and-answer and other types. The target type may be determined according to actual needs and is not limited here.

[0061] The above-mentioned target difficulty level may refer to the expected difficulty level of the test questions data to be generated. The target difficulty level may include but is not limited to simple, ordinary, complex and other levels. The target difficulty level may be determined according to actual needs and is not limited here.

[0062] The above-mentioned target dimension may refer to the expected test content / dimension of the test data to be generated. The target dimension may include but is not limited to the test dimension selected by the user, or the test dimension customized by the user, etc. The target dimension may be determined according to actual needs and is not limited here.

[0063] In an optional embodiment, the user can specify the generation conditions with a second input instruction on the operation interface. The generation conditions may include information such as target type, target difficulty level, target dimension, etc. The second instruction input by the user can be parsed by the system and the corresponding generation conditions can be displayed on the operation interface. The system can determine the specific test question generation conditions through the second input instruction input by the user, so that the user can clearly understand the requirements for generating test question data, so that the subsequent system can generate test question data that meets the user's requirements based on the user-specified generation conditions. In the above process, the user specifies the generation conditions by inputting the second instruction, and the system can display the generation conditions. The user can easily generate test question data that meets the requirements, which can improve the diversity of the generated test question data and improve the efficiency and accuracy of test question generation.

[0064] Step S208, in response to the generation instruction on the operation interface, the target test question data is displayed on the operation interface.

[0065] Among them, the target test question data is the target text data input into the test question generation model, and the data generated by the test question generation model is guided by the generation conditions.

[0066] The above-mentioned generation instructions may refer to instructions for controlling the display of target test data generated by the user interacting with the operation interface through clicks, selections or language. The form of the generation instructions may be determined according to the specific interaction method between the user and the operation interface, and is not limited here.

[0067] The above-mentioned large model for test question generation may refer to a deep learning model that can generate target test question data based on target text data. The large model for test question generation may adopt a large model. The large model may refer to a neural network model with a relatively large number of parameters in deep learning. The large model may have millions to billions of parameters and requires a large amount of data and computing resources to train and tune. The large model performs well in processing complex tasks and large-scale data sets, such as natural language processing, computer vision, speech recognition and other fields. The large model for test question generation may also be determined according to actual needs and is not limited here.

[0068] The above-mentioned target test data may refer to test data for assessing students' knowledge, abilities and skills generated in accordance with certain formats and specifications based on specific teaching, training requirements and assessment standards. The target test data may include different types of questions such as multiple-choice questions, fill-in-the-blank questions, and essay questions, and may cover all aspects of the course content.

[0069] The process of generating target test data needs to take into account many factors such as teaching objectives, student levels, and examination formats to ensure that the test data can comprehensively, accurately, and appropriately reflect students' learning situation and ability level. The generation of target test data also needs to follow certain norms and standards to ensure the validity and fairness of the test questions. The generation of target test data can provide an effective tool for teaching and training evaluation, help understand students' learning situation, promote students' learning improvement, and provide an important reference for students' development.

[0070] In an optional embodiment, the user can trigger a response through a generation instruction on the operation interface and input the target text data into the test question generation model. The test question generation model can generate corresponding test question data based on the input text data. After training, the test question generation model can automatically learn and generate test question data that meets the requirements based on the input text data. The target text data is input into the test question generation model, and the model can start running to generate corresponding test question data based on the learned knowledge and rules. The generated test question data may include question content, options, answers and other information. Finally, the generated test question data will be displayed on the operation interface and the user can view and use the data. In the above process, the user interacts with the test question generation model through the generation instruction on the operation interface to realize the generation of test question data.

[0071] In the above process, the user can interact with the system through the generation instructions on the operation interface to guide the system to generate target test data. Specifically, the user can input the target text data on the operation interface and trigger the generation instruction, so that the system can use the test generation model to generate test data that matches the input text data, and can display the generated target test data on the operation interface, so that the user can intuitively view, review and edit the generated test data. The user can adjust and modify the generated test data as needed to ensure that the generated test data meets the expected requirements. Through the above interactive setting process, the user can easily guide the system to generate test data that meets the requirements and improve the efficiency and accuracy of test generation.

[0072] The present application sets up a first input instruction, a conversion instruction, a second input instruction and a generation instruction that respond to the operation interface. When the user inputs the first input instruction, the operation interface can display the target document containing multimodal data corresponding to the input instruction. The user can quickly view and understand the content of the target document, providing basic information for subsequent operations; when the user enters the conversion instruction, the operation interface can display the target text data corresponding to the conversion instruction. The text data is obtained by parsing the multimodal data using a large document parsing model, and the user can read and understand the text data more conveniently; the user specifies the generation conditions by entering the second instruction, and the system can display the generation conditions so that the user can clearly understand the requirements for generating test question data, and can determine the specific test question generation conditions through the second input instruction entered by the user, so that the subsequent system can generate test question data that meets the user's requirements based on the generation conditions specified by the user; when the user enters the generation instruction, the operation interface can also display the target test question data.

[0073] In an embodiment of the present application, a method for generating test question data is provided, the method comprising: in response to a first input instruction acting on an operation interface, displaying a target document corresponding to the input instruction on the operation interface, wherein the target document includes multimodal data; in response to a conversion instruction acting on the operation interface, displaying target text data corresponding to the conversion instruction on the operation interface, wherein the target text data is text data obtained by parsing multimodal data using a document parsing large model; in response to a generation instruction acting on the operation interface, displaying target test question data on the operation interface, wherein the target test question data is data generated by inputting the target text data into a test question generation large model using the test question generation large model. It is easy to notice that the target document in the present application uses multimodal data to enrich the content of the document and ensure the diversity of the generated test data; in the process of generating test data, the text data obtained by parsing the multimodal data in the target document using the document parsing model can help realize the conversion from multimodal data to text data. The application of the document parsing model can improve the system's processing ability for multimodal data, so that the system can better parse and utilize multiple data sources, and provide more comprehensive and accurate information for subsequent test generation; the user specifies the generation conditions by entering the second instruction, and the system can display the generation conditions so that the user can clearly understand the generation conditions of the test. The requirements of the test data can be determined by the second input instruction input by the user, and the specific test question generation conditions can be determined, so that the subsequent system can generate test question data that meets the user's requirements based on the generation conditions specified by the user; the target test question data is generated by inputting the target text data into the test question generation large model. The test question data generated by the test question generation large model can ensure the quality and diversity of the test questions and improve the efficiency of generating test question data. Through the above steps, the user can quickly generate diverse and high-quality test question data through simple input instructions, improve the efficiency and diversity of generating test question data, and thus solve the technical problems of low efficiency and diversity in generating test question data in related technologies.

[0074] In the above-mentioned embodiment of the present application, in response to the conversion instruction acting on the operation interface, the method also includes: performing format conversion on the target document to obtain a converted document; extracting the first modal data in the converted document to obtain first text data, wherein the first modal data includes one of the following: text data, table data; performing at least one round of parsing on the second modal data in the converted document using a document parsing model to obtain second text data, wherein the second modal data includes: image data; summarizing the first text data and the second text data to obtain target text data.

[0075] The converted document may refer to a document in a format different from that of the target document and easier to understand than the target document.

[0076] Optionally, the target document format can usually be Hypertext Markup Language (HTML for short). In this case, the target document format can be converted to obtain a converted document. The converted document can be a markup language (Markdown for short). HTML syntax is relatively complex and relatively difficult to read. Compared with the target document in HTML format, the Markdown format is easier to read and edit. Converting the original document to Markdown format can improve the readability and editability of the document. The Markdown format is more suitable for version control and collaborative editing. Markdown documents can be easily managed and collaboratively edited through a version control system, while HTML documents may have problems such as format confusion and conflicts. Converting the original document to Markdown format can better support team collaboration and version control; Markdown format is more suitable for publishing to different platforms. Markdown documents can be easily converted to HTML, PDF and other formats, which are suitable for publishing to different platforms and devices. Markdown documents are also easier to display and share on web pages. Converting the original document, such as HTML format, to a converted document in a format that is easier to understand than the target document, for example, Markdown format can improve the readability, editability and applicability of the document, and facilitate the subsequent efficient generation of test data. The format of the target document and the converted document can also be determined according to actual needs, which is not limited here.

[0077] In an optional embodiment, in response to the user inputting a conversion instruction on the operation interface, the system can perform a format conversion operation on the target document according to the conversion instruction input by the user to obtain a converted document, and extract first modal data from the converted document, which may include text data and table data. The text data in the extracted first modal data can be integrated to obtain first text data; the system can use a document parsing model to perform a parsing operation on the second modal data in the converted document. The second modal data may be image data. The system can identify and extract the image data through a parsing model. After at least one round of parsing, the system can obtain second text data, which includes text content extracted from the image data. Finally, the first text data and the second text data can be summarized and integrated to obtain target text data. Through the above steps, the user can obtain complete text data, which may include at least text data, table data and image data extracted from the original document, thereby ensuring a full understanding and processing of multimodal data in the target document and ensuring the integrity and richness of the content of the subsequently generated test question data.

[0078] In the above embodiment of the present application, the second modal data in the converted document is parsed at least once using a large document parsing model to obtain second text data, including: responding to a parsing instruction on an operation interface, performing at least once parsing the second modal data using a large document parsing model through an asynchronous call to obtain second text data.

[0079] The above-mentioned parsing instructions may refer to instructions for controlling at least one round of parsing of the second modal data to obtain second text data, which are generated when the user interacts with the operating interface through clicks, selections or language. The form of the parsing instructions can be determined according to the specific interaction method between the user and the operating interface, and is not limited here.

[0080] The above-mentioned asynchronous call can refer to a programming mode in which the execution of a function does not block the running of the program, but is executed in the background, while allowing the program to continue to perform other operations. The asynchronous call can use a callback function or object (Promise) to process the result after the function is executed. The use of asynchronous calls can improve the performance and response speed of the program, and can be applied to scenarios that require network requests or time-consuming operations.

[0081] In an optional embodiment, a large document parsing model can be used to quickly and accurately obtain text data and convert it into structured data. In the process of generating test data, there will be some second modal data, such as pictures, videos, audio, etc., which need to be parsed into text data before further processing. Specifically, in response to the user issuing instructions through elements such as buttons or input boxes on the operation interface, the system executes parsing of the second modal data. The parsing instructions can be passed to the background processing program and can be processed through asynchronous calls. Based on the asynchronous call method, there is no need to wait for the result to be returned after the request is issued, but other tasks can be continued. The result can be obtained after the processing is completed. The system can pass the second modal data to the large document parsing model for processing, which can at least include data format conversion, preprocessing and other operations. , to ensure that the document parsing model can process the data correctly; the document parsing model can parse the second modal data and extract the text information therein, which can include text recognition, voice recognition and other technologies to convert pictures, audio, video and other data into processable text data; after the parsing is completed, the obtained second text data is returned to the front-end interface, and the user can view the parsed text data through the interface and further process or analyze the data. Through the above process, the second modal data can be parsed and converted into text data, providing a basis for subsequent test data generation. The system calls the corresponding parsing module to parse the second modal data according to the instructions input by the user. The system can generate data that meets the requirements in real time according to the user's operation, thereby improving the user experience and the intelligence of the system.

[0082] In the above embodiment of the present application, the second modal data is parsed at least once using a document parsing big model to obtain second text data, including: classifying the second modal data using the document parsing big model to determine the type of the second modal data; when the type of the second modal data is a preset type, parsing the text data in the second modal data using the document parsing big model to obtain the second text data.

[0083] The above-mentioned preset type may refer to a preset data content type that can be used when generating subsequent test question data. The specific content of the preset type may be determined according to actual needs and is not limited here.

[0084] Optionally, the second modal data includes image data of various content types. The second modal data may include flow chart images, table images and an operating interface for interaction. Considering that the operating interface for interaction may not contain data content that can be used in the subsequent generation of test data, the preset type can be set to flow chart images, table images, etc. that contain data content that can be used in the subsequent generation of test data. Therefore, based on the preset type, effective screening of image data of various content types in the second modal data can be achieved. The system can generate data that meets the requirements in real time according to the user's operations, thereby ensuring efficient processing of image data in the target document and improving the efficiency of generating test data.

[0085] In an optional embodiment, a document parsing large model can be used to automatically identify and extract text information. The present application can use the document parsing large model to classify the second modal data and determine the type of the second modal data. If the type of the second modal data is a preset type, that is, image data containing valid content, the document parsing large model can be used to parse the text data therein to obtain the second text data. If the type of the second modal data is not a preset type, that is, image data that does not contain valid content, for example, an operating interface for interaction, the corresponding second modal data can be directly discarded to avoid subsequent invalid processing, thereby saving system data processing resources and improving the efficiency of test data generation. Specifically, the image data can be discarded. The image data is input into the image classification model for identification to determine its type. If the second modal data is classified as image data containing valid content, the document parsing model can be used to parse the text data therein. The document parsing model uses natural language processing technology, such as text classification, entity recognition, relationship extraction, etc. After processing by the document parsing model, the second text data can be obtained, which contains text information extracted from the image data. This information may include titles, descriptions, keywords, etc., which are helpful for further data analysis and processing. Through the above steps, the document parsing model can be used to parse the text data in the second modal data to obtain the second text data, which provides a basis for subsequent test data generation.

[0086] In the above embodiment of the present application, when the generation conditions include a target dimension, the method also includes: responding to a conversion instruction acting on an operation interface, displaying multiple dimensions on the operation interface, wherein the multiple dimensions are the results of parsing the dimensions of the target text data using a dimensional parsing model; responding to a second input instruction acting on multiple dimensions, displaying the target dimension on the operation interface, wherein the target dimension is a dimension selected from the multiple dimensions based on the second input instruction.

[0087] In an optional embodiment, in response to a conversion instruction acting on an operation interface, the system can perform corresponding processing according to the user instruction. The system can use a dimensional analysis model to analyze multiple dimensions of the target text data to obtain the performance of the text data in different dimensions. These dimensions may include different aspects such as subject, emotion, time, etc., and the results of multiple dimensions can be displayed on the operation interface, so that the user can intuitively understand the performance of the text data in different dimensions; the user can select one of the dimensions as the target dimension based on the multiple dimensions displayed in the first step. The system receives the user's second input instruction and performs further processing based on the target dimension selected by the user. The system may highlight the information of the target dimension on the operation interface to allow the user to pay more attention to the content of the dimension. The above process uses the instructions and selections input by the user, combined with the dimensional analysis model, to analyze and display multiple dimensions of the text data, thereby helping users to better understand and process text data, and can help users generate test data more efficiently, and can perform personalized processing based on user needs and concerns.

[0088] The back-end program of the system of the present application can use the big model to analyze and disassemble the source text, and extract the knowledge points in the text. These knowledge points can be concepts, definitions, theorems, examples, etc. involved in the text. The back-end program will return the points to the front-end according to the extracted knowledge points for the user to choose. The user can choose one of the knowledge points to make questions according to his or her needs and teaching direction. For example, the back-end program can receive the source text input by the user, analyze and disassemble the text through the big model, and extract the knowledge points in the text. The extracted knowledge points can be returned to the front-end to form a list of knowledge points for the user to choose. The user can see the list of knowledge points on the operation interface and select one of the knowledge points according to his or her needs. The front-end can send a request to the back-end according to the knowledge points selected by the user, requesting the generation of relevant test question data. The back-end can generate corresponding test question data according to the knowledge points selected by the user, and the generated test question data is returned to the front-end, and the user can view and edit the test questions. Through the above process, the user can select specific knowledge points to make questions according to the needs, which improves the pertinence and effectiveness of the test question data. The back-end program uses the big model to perform in-depth analysis of the text, and provides users with a more intelligent test question generation service.

[0089] In the above embodiment of the present application, the method further includes: parsing the dimensions of the target text data using a large dimension parsing model through an asynchronous call to obtain multiple dimensions.

[0090] In an optional embodiment, in the process of generating test data, the dimension parsing big model can be used to parse the dimension of the target text data to obtain multiple dimensions, and the efficiency and speed of data processing can be improved by asynchronous calling. The construction of the dimension parsing big model can be realized by machine learning algorithm or deep learning algorithm, and multiple dimensions of information can be extracted from the target text data, such as the theme, difficulty, knowledge points, etc. Specifically, the target text data can be passed into the dimension parsing big model for processing by asynchronous calling. The asynchronous call can process multiple data at the same time, improving the efficiency and speed of data processing; the dimension parsing big model can be used to process the target text data to obtain multiple dimensions of information, for example, the theme of each sentence, the difficulty of each word, etc. can be obtained; according to the obtained multi-dimensional information, various types of test questions can be generated by the generation algorithm. For example, multiple-choice questions can be generated according to the theme, fill-in-the-blank questions can be generated according to the difficulty, etc. Through the above steps, the dimension parsing big model can be used to parse the dimension of the target text data by asynchronous calling, and multiple dimensions can be obtained, thereby realizing the generation of test data, which can improve the efficiency and quality of test generation, and is suitable for the scenario of large-scale test generation.

[0091] In the above-mentioned embodiment of the present application, in response to the generation instruction acting on the operation interface, the method also includes: obtaining a preset prompt word template corresponding to the test question generation model; determining the prompt word information matching the generation conditions; injecting the prompt word information into the preset slot corresponding to the generation conditions in the preset prompt word template to generate target prompt word information; inputting the target text data into the test question generation model, and using the target prompt word information to guide the test question generation model to generate target test question data.

[0092] The above-mentioned preset prompt word template may refer to a pre-designed template for guiding the test question generation model to generate test question data. It may include some preset prompt words or keywords for guiding the generation conditions and test question content. The preset prompt word template can help the test question generation model to better understand the generation conditions, and at the same time provide some key information to make the generated test question data more accurate and meet the requirements. The preset prompt word template can be determined according to actual needs and is not limited here. Through the use of the preset prompt word template, the test question generation process can be better controlled to ensure that the generated test question data meets the requirements and improve the generation efficiency and accuracy.

[0093] The above-mentioned preset slots may refer to slots at some preset positions contained in the preset prompt word template, which are used to fill in keyword information related to the generation conditions. After determining the prompt word information that matches the generation conditions, the prompt word information can be injected into the corresponding slots in the preset prompt word template to form target prompt word information, which is convenient for subsequently inputting the target text data into the test question generation model, and using the target prompt word information to guide the model to generate test question data.

[0094] In an optional embodiment, the test question generation model used can be determined. The test question generation model can be a rule-based model or a machine learning-based model such as a generative adversarial network. It is necessary to obtain the corresponding preset prompt word template according to the selected model. The template contains the basic structure and format of test question generation. According to the conditions and requirements of test question generation, prompt word information matching the generation conditions is determined. These prompt word information can be keywords, difficulty requirements, etc. related to the test question content. The determined prompt word information can be injected into the preset slots corresponding to the generation conditions in the preset prompt word template to generate target prompt word information. This can be achieved through natural language processing technology, inserting keywords into corresponding positions in the preset template. Finally, the target text data can be input into the test question generation model, and the target prompt word information can be used to guide the test question generation model to generate target test question data. According to the input text data and prompt word information, test question data that meets the requirements can be generated while maintaining the logic and rationality of the test question. The above process can help to achieve automatic generation of various types of test question data by determining the generation conditions, extracting key prompt word information, injecting preset templates, and guiding the model to generate test question data.

[0095] In the above embodiment of the present application, in response to a first input instruction acting on the operation interface, the method further includes: determining a target document identifier corresponding to the first input instruction; and acquiring the target document from a preset document library based on the target document identifier.

[0096] The above-mentioned preset document library may refer to a document library used to obtain target documents. The preset document library may contain various documents, such as textbooks, reference books, papers, etc. The preset document library may adopt a knowledge base system (KBS for short). The knowledge base system may be a system built on computer technology for collecting, organizing, storing and retrieving knowledge, and may include various forms of information, such as text, images, audio and video, etc. The knowledge base system may be used to help users solve problems, obtain information or make decisions, and support users' learning and work by providing a centralized knowledge base. The preset document library may also be a rule-based expert system, a semantic-based knowledge graph system, or an intelligent system based on machine learning and natural language processing technology. The preset document library may be determined according to actual needs and is not limited here.

[0097] In an optional embodiment, the test content that meets the requirements can be automatically generated by a computer program according to certain rules and requirements. Specifically, when determining the target document identifier corresponding to the first input instruction, the target document can be identified by some keywords or instructions. For example, the category of the target document can be identified by the keywords or phrases in the input instruction, and then the identifier of the target document can be determined according to these categories. The identifier of the target document can also be determined by a specific format or structure in the input instruction. For example, the instruction may contain information such as the document number or name. The specific process of obtaining the target document from the preset document library based on the target document identifier can include: according to the information in the first input instruction, identifying the identifier of the target document, such as the category, number, name, etc. of the document; matching the identifier of the target document with the documents in the preset document library to find a document that matches the target document identifier; after finding a matching document, the content of the target document can be obtained from the preset document library, which can be achieved by reading the document file or accessing the data in the document library. After obtaining the target document content, the document content can be further processed, such as extracting key information, filtering useful content, etc. Through the above steps, the process of obtaining the target document from the preset document library can be achieved, providing the necessary materials and information for the subsequent generation of test question data.

[0098] In the above-mentioned embodiment of the present application, after the target test question data is displayed on the operation interface, the method further includes: responding to a modification instruction acting on the target test question data, displaying the modified test question data on the operation interface, wherein the modified test question data is the test question data after the target test question data is modified based on the modification instruction; responding to a confirmation instruction acting on the modified test question data, displaying the export result of the modified test question data on the operation interface, wherein the export result is the result obtained by exporting the modified test question data.

[0099] The above-mentioned modification instructions may refer to instructions for controlling the display of modified test data on the operation interface generated by the user interacting with the operation interface through clicks, selections or language. The form of the modification instructions can be determined according to the specific interaction method between the user and the operation interface, and is not limited here.

[0100] In an optional embodiment, the operation interface can be used to display and modify test question data. The user can input modification instructions through the operation interface, such as adding, editing, and deleting test question content. The modification instructions can act on the target test question data and modify the target test question data according to the instructions input by the user; the system can display the modified test question data on the operation interface according to the modification instructions input by the user, and the user can check whether the modified test question data meets the expectations. If further modification is required, the user can continue to input modification instructions for adjustment; when the user confirms that the modified test question data meets the requirements, the user can enter a confirmation instruction to perform an export operation. The system can export the modified test question data according to the confirmation instruction entered by the user, and display the export result on the operation interface. The export result can be a file or other form of output. The user can save or use these export results for subsequent operations. In the above process, the user can perform modification, confirmation and export operations as needed to ensure that satisfactory test question data is obtained. The user can conveniently generate test question data that meets the requirements through the operation interface, thereby improving the efficiency and accuracy of generating test question data.

[0101] In the above embodiment of the present application, after the target document corresponding to the input instruction is displayed on the operation interface, the method also includes: determining whether the target document is a document stored in a preset document library; when the target document is a document stored in the preset document library, obtaining multiple test question data corresponding to the target document from the cache, wherein different test question data are input into the test question generation model by inputting the target text data, and using different generation conditions to guide the data generated by the test question generation model; in response to the generation instruction, displaying the target test question data on the operation interface, wherein the target test question data is the test question data corresponding to the generation instruction among the multiple test question data.

[0102] In an optional embodiment, it can be determined whether the target document is a document stored in a preset document library. This can be achieved by performing text matching or hash value comparison on the content of the target document. If the target document already exists in the preset document library, multiple test question data corresponding to the target document can be obtained from the cache. The process of obtaining multiple test question data corresponding to the target document can include: inputting the target text data into a test question generation model: inputting the text data of the target document into the test question generation model to generate corresponding test question data; using different generation conditions to guide the test question generation model to generate data: by setting different generation conditions, for example, different model parameters can be selected, different generation rules can be set, etc., to guide the test question generation model to generate multiple different test data; according to different generation conditions, the test question generation model will generate multiple test question data related to the target document; in response to the generation instruction, the process of displaying the target test question data on the operation interface includes the following steps: the system receives the user's generation instruction and needs to generate test question data related to the target document. The system can obtain multiple test question data corresponding to the target document from the cache according to the storage status of the target document in the preset document library. The system can also display the target test question data corresponding to the generation instruction on the operation interface, allowing users to view and use these test question data. Through the above process, the system can generate test question data related to the target document and display this data on the operation interface, providing users with more convenient test question generation and management functions.

[0103] In the above embodiment of the present application, the method also includes: regularly monitoring whether new documents appear in the preset document library, wherein the preset document library is used to store target documents; when new documents appear in the preset document library, using the document parsing big model to parse the new documents to obtain new text data; inputting the new text data into the test question generation big model, using different generation conditions to guide the test question generation big model to generate different test question data; storing different test question data in the cache.

[0104] In an optional embodiment, during the generation of test question data, the specific process of periodically monitoring whether new documents appear in the preset document library may include determining the preset document library that stores the target document, setting a scheduled task, and regularly checking whether there are new documents in the preset document library. The automatic monitoring function can be implemented through programming, and the files in the document library are scanned regularly to detect whether there are new documents. When new documents appear in the preset document library, the document parsing model can be used to parse the new documents. After being processed by the document parsing model, new text data can be obtained, and the new text data can be input into the test question generation model. The obtained new text data is input into the test question generation model. The test question generation model can be a model based on machine learning, which is used to generate text data based on the input text data. According to the generated test questions; different generation conditions can be used to guide the test question generation big model to generate different test question data: different generation conditions can be set in the test question generation big model, and test question data of different types or difficulties can be generated according to needs. For example, you can set it to generate test questions of different question types such as multiple-choice questions, fill-in-the-blank questions, and short-answer questions; different test question data can be stored in the cache, and the generated different test question data can be stored in the cache for subsequent use. The cache can be a storage medium such as a database and a file system, which is convenient for managing and retrieving test question data. Through the above steps, it is possible to periodically monitor whether new documents appear in the preset document library, and use the document parsing big model and the test question generation big model to generate different types of test question data, providing an effective solution for scenarios such as the education field that require a large number of test questions.

[0105] The intelligent question-setting system proposed in this application adopts a technical solution of multimodal understanding and multi-round intelligent generation. It processes the complex content in the target document, such as pictures, tables, etc., through a multimodal model, and designs a multi-task combination to break down knowledge points and customize questions to ensure that the generated questions can meet the needs of different difficulty levels and question types. In addition, asynchronous tasks and caching mechanisms can be used to optimize response time and improve the real-time performance and user experience of the system. It can not only adapt to the training needs of different fields, but also significantly improve the accuracy, coverage and diversity of questions, and transform the traditional question-setting process into a more intelligent and efficient standard. The traditional question-setting process can be transformed into a more intelligent and efficient future standard. It only needs to input relevant specifications to replace a large number of manual operations, and solves the problems of time and quality. It can be derived into various educational fields to provide strong support for the comprehensive improvement of education quality.

[0106] The technical solution proposed in this application can be applied to the supply team responsible for customer service personnel training resources. The work content of the application party may include learning and interpreting the latest application rules, and then converting these contents into various training resources such as slide presentations, knowledge points, test questions and practical operations. From determining the training direction, to learning and interpreting the rules, to the output and distribution of training content, the technical solution proposed in this application can be based on the good article comprehension ability of the Tongyi Qianwen model (for example, qwen1_5_110B). The system can assist in improving the production efficiency of training content. Through this system, not only can the test question generation process be greatly accelerated, but the design task can also be converted into an audit task, so that the output quality of the test questions is significantly improved, thereby optimizing the overall training process.

[0107] The content of the questions in this application can be derived from KBS documents. KBS documents contain a lot of complex and cumbersome application knowledge and rules, and there is no unified specification for the writing of KBS document formats. The text is within complex HTML tags, and at the same time there are contents in various formats such as tables, flow charts, slide presentations, screenshots, etc. If such data is not processed and directly input into the large model, it will greatly affect the production effect. Since KBS is a page document, the original data is in HTML format. This application can convert HTML tags into converted documents that are easy for the large model to understand through program packaging files (for example, jar package com.kotcrab.remark), such as Markdown format, etc. This application can also add support for rich text format or document files, portable document formats and other formats. Through corresponding data analysis and conversion cleaning, valid data is obtained for large model input. The pictures, screenshots, etc. in the text can be classified using a multimodal model (for example, qwen-vl-max), and the classification finds those containing valid knowledge theory information for a second call interpretation.

[0108] The source data used to generate test data in this application can be derived from offline data of the KBS system. The KBS system can include the ID, title and main content of the KBS document, which can be in HTML format. Through the query service in the user service (UserService), the user service (User Service) can be a service provision model that provides users with specific services or functions. Through the user service (User Service), users can find and obtain the required services through a certain platform or system. This model usually covers various fields, such as online shopping, social media, financial services, etc. Users can use the user service (User Service) to query the user service (User Service). The KBS system can be used to easily find and order the required services, thereby improving the accessibility and availability of the services. The original material files needed by the trainer can be found by the document number (id) and name in the KBS system. In addition to the queryable original data of the KBS document, it can also support the customized input of the source text of the question knowledge on the page. When the source text of the question is obtained and the user clicks the format conversion button, the back-end program can automatically and asynchronously call the large model to disassemble the knowledge points according to the source text while the format conversion is turned on. When the large model completes the knowledge point parsing, the back-end program can return the knowledge points to the front-end. The trainer can choose one of the knowledge points to set questions, or customize the input knowledge points.

[0109] Figure 3 is a flowchart of an optional method for generating test question data according to an embodiment of the present application, such as Figure 3 As shown, in step S302, a target document is obtained from a preset document library based on a document identifier or a document name through a query function in a user service.

[0110] Step S304: convert the format of the target document to obtain a converted document.

[0111] For example, HTML tags may be converted into a converted document that is easy for a large model to understand, such as a Markdown format, through a program packaging file (eg, a jar package com.kotcrab.remark).

[0112] Step S306, extracting the first modal data in the converted document to obtain first text data, wherein the first modal data includes one of the following: text data, table data; using the document parsing large model to classify the second modal data to determine the type of the second modal data, the second modal data includes: image data; when the type of the second modal data is a preset type, using the document parsing large model to parse the text data in the second modal data to obtain second text data.

[0113] For example, the text data in the first modal data may be directly processed, and the tags and attributes in the HTML tag content in the table data in the first modal data that are not related to the table content (for example, the tags and attributes may be , and ) is deleted; the process of processing the image data in the second modal data can be, for example, judging the content of the image data, taking into account that the image data of the interactive operation interface may not contain data content that can be used in the subsequent generation of test data, the preset type can be set to a flowchart image, a table image, etc., which contains data content that can be used in the subsequent generation of test data, so that based on the preset type, effective screening of image data of various content types in the second modal data can be achieved, and if the type of the second modal data is not the preset type, that is, image data that does not contain valid content, for example, an interactive operation interface, etc., the corresponding second modal data can be directly discarded here to avoid subsequent invalid processing, which can save the system's data processing resources and improve the efficiency of test data generation. The system can generate data that meets the requirements in real time according to the user's operation, ensure efficient processing of image data in the target document, and improve the efficiency of test data generation.

[0114] Step S308, summarizing the first text data and the second text data to obtain target text data; and displaying generation conditions on the operation interface, wherein the generation conditions include at least one of the following: target type, target difficulty level, and target dimension.

[0115] Step S310, in response to the generation instruction on the operation interface, the target test question data is displayed on the operation interface, wherein the target test question data is the data generated by inputting the target text data into the test question generation model and guiding the test question generation model using the generation conditions.

[0116] Figure 4 is a schematic diagram of configuring a window in another optional operation interface according to an embodiment of the present application, such as Figure 4 As shown, in the configuration window, from top to bottom, there can be set up a test source selection module, a test type selection module, a test difficulty selection module, a test point selection module and a generated question selection button. Among them, the test source selection module can be set with "KBS link" and "manual input" options for users to choose. When "KBS link" is selected, the user can select or fill in the KBS knowledge base address, and set whether to "parse KBS pictures" option. There is a prompt message on the page that "picture parsing will take a long time, it is recommended to choose whether to turn it on according to the kbs content", and there are also clickable options for "view conversion results" and "reconvert" options; the test type selection module can be set with "single choice" and "multiple choice" options for users to choose; the test difficulty selection module can be set with " Easy" and "Difficult" options. After clicking "Easy", a prompt such as "Definition of Difficulty: Usually a simple scenario is given directly, and the customer is asked to choose the correct answer. The correct answer can be directly selected by recalling the knowledge points. Requirements: Understand the basic concepts of application scenarios, and be familiar with service standards, application rules, compensation standards, applicable consumer membership levels, member tags, and applicable order statuses. Be familiar with the basic operations, operation entrances, and interface information of the system. Examination method: usually direct questions. " can be set in the examination point selection module with "Select" and "Custom" options for users to choose from. After clicking "Select", examination points such as "Account Switching Operation Process" can be displayed for users to choose from. After selecting the above configuration items, users can click "Generate Questions" to automatically generate target test data with one click.

[0117] Figure 5 is a flowchart of an optional method for generating test question data according to an embodiment of the present application, such as Figure 5 As shown, the specific steps may include the following:

[0118] Step S502: The user can input the link, number or name of the KBS document in the configuration window in the operation interface, or make a custom input.

[0119] Step S504, in response to the user clicking the image analysis operation, the system can execute the image analysis function. At this time, the user usually needs to wait for 15 to 30 seconds. If the user does not need to parse the image content in the KBS document, the system can directly discard the image content in the KBS, and the user only needs to wait for 1 second.

[0120] Step S506, after the conversion is completed, the user can view and confirm the conversion result, or choose to re-convert;

[0121] Step S508, after the user confirms the conversion result, the user can configure the difficulty of the test questions, select the knowledge points to be tested, etc. in the configuration window in the operation interface, or the user can customize the input.

[0122] Step S510, after the user selects the operation of generating test questions, the system waits for 5 to 10 seconds and then outputs the generated test question data. The user can view the generated test question data and modify the generated test question data based on the modification instructions. Finally, the target test question data that meets the user's needs can be generated.

[0123] Figure 6 is a schematic diagram of a system timing of an optional method for generating test question data according to an embodiment of the present application, such as Figure 6 As shown, users can use the front-end <1> Submit KBS document link / ID / name or enter the document directly. <2> Choose whether to parse the image. The front end can be based on the model training module. <3> Request to parse KBS documents or input, the model training module can be based on user services <4> The model training module can search for KBS HTML data by KBS ID name. <5> Convert HTML to Markdown, process table tags, and when (alt) needs to parse images, the model training module can parse large models through documents <6> Analyze the image (10s~30s), <7> Return results; model training module can be based on memory database <8> Store the parsing results in the cache, and the front end can train the module based on the model <9> Polling to obtain the parsed MarkDown text; the asynchronous execution steps may include: the model training module can generate a large model based on the test questions <10> Asynchronous request analysis knowledge points, <11> Parse knowledge points (5s-15s), the model training module can be based on the in-memory database <12> Store the parsing results in the cache, and the front end can train the module based on the model <13> Polling to get the parsed knowledge points, the front end can <14> Load it into the knowledge point selection drop-down box of the question configuration. The front end can <15> Renders the MarkDown results to the MarkDown editor, and users can use <16> Edit MarkDown documents, <17> Select the question configuration, the front end can be based on the model training module <18> Send out a question request, and the model training module can generate a large model based on the test questions. <19> Large model questions (5s to 15s), model training module can be based on in-memory database <20> The question results are stored in the cache, and the front end can be based on the model training module <21> Polling to get the question results, the front end can <22> Loaded into the topic editor, users can use the front-end <23> View / edit questions, <24> Adopt, the front end can <25> Load the test questions into the list, and users can use the front-end <26> Export to Excel / Import the question bank. Finally, the front end can train the module based on the model <27> The system timing process of storing the test questions and the method for generating the test question data can also be determined according to actual needs and is not limited here.

[0124] The above-mentioned model training module can adopt module (xixi_training), the above-mentioned user service can be user service (Uservice), the above-mentioned document parsing large model can adopt model (qwen_vl_max), the above-mentioned memory database can adopt memory database (Tair), and the above-mentioned test question generation large model can adopt (qwen1_5_110B). The above-mentioned parts can also be determined according to actual needs and are not limited here.

[0125] Figure 7 is a schematic diagram of the interaction of functional modules of an optional method for generating test question data according to an embodiment of the present application, such as Figure 7 As shown, the functional modules of the method for generating test question data proposed in the present application may include a KBS document processing module, a test question production and design module, a user interaction module and a cache optimization module, wherein the KBS document processing module can execute the HTML to MarkDown function, the multimodal model image classification function and the multimodal model image parsing function; the test question production and design module can execute the knowledge point disassembly function and the question generation function; the user interaction module can execute the confirmation and release function, the question type and difficulty selection function and the user upload document function; the cache optimization module can execute the scheduled task preprocessing function and the cache mechanism function, the KBS document processing module can obtain the KBS application document, the multimodal model diagram The film classification function and the multimodal model image analysis function can adopt the multimodal model (qwen_vl_max). In terms of model selection for the knowledge point decomposition function and the question generation function, the model (qwen1.5-110B) can be selected. The model (qwen2-72B-instruct) and the model (qwen-max) can be used in ultra-long contexts. The multiple-choice question type and difficulty function and the cache mechanism function can jump to the question generation function, and then jump to the confirmation and release function. The user upload document function can also jump to the HTML to MarkDown function. The interaction process of the above functional modules can also be determined according to actual needs and is not limited here.

[0126] This application takes into account that if the big model is directly asked to make questions based on the article, the big model will randomly make a question, and the training application has extremely high requirements for question making. It is necessary to ensure that the question types are rich, the difficulty is differentiated, and the knowledge points should be covered as much as possible. According to the design requirements of the test paper, the prompt words for generating questions can be designed to support question types, such as single-choice / multiple-choice customization, difficulty, such as simple / difficult customization, and asynchronous tasks can also be designed to disassemble the article knowledge points for optional configuration, and also support custom input of knowledge points. In the prompt word design concept of this application, the system prompt words are used to explain the task clearly, and the user prompt words are used to input task materials, question source text and specified knowledge points. This allocation can achieve better results. Regarding the task of disassembling knowledge points, it has been found through experiments that the big model can interpret the knowledge points suitable for question making from complex application documents very well. The method is to make it clear in the system prompt words that the task background is to make questions, and to stipulate and limit the emphasis on the output format so that the output can be parsed and returned by the program, and the source document is directly input in the user prompt words. The system prompt words can be as follows.

[0127] As the core big model calling task of the application, a variety of customized designs are implemented in the question-setting task. In addition to directly injecting the specified knowledge points into the user prompt word slot, the difficulty customization and question type customization supported by the front end are based on the user's different choices. The pre-written simple and difficult definitions, single-choice and multiple-choice definitions are selected and injected into the corresponding slot instead of injecting all of them. When injecting the simple definition, the big model does not know the difficult definition, which can effectively avoid the interference of the semantics of different requirement definitions on the task effect, that is, the difficulty (difficulty) prompt word and question type (questionType) prompt word, question type selection (questionTypeChoice) slot in the following prompt words. The specific slot definition injection is also given as follows. The output restriction of this task can be the requirement object representation (json) format, and a format example without content is given to avoid the interference of content examples on the generation effect. The prompt word configuration options are as follows:

[0128] For example, "Simple difficulty prompt words" can be configured as "Simple difficulty: \nDefinition: Simple difficulty questions usually directly give clear and simple scenarios, requiring customer service staff to be able to select the correct answer through basic recall or intuitive judgment. This type of question generally only involves a single knowledge point or operation step, and does not require complex logical reasoning or in-depth data analysis. \nRequirement: Customer service staff need to master the basic concepts, service standards, application rules, and compensation standards of application scenarios, including applicable consumer membership levels, member tags, order status, system operations, etc. This type of question mainly tests customer service staff's familiarity and understanding of basic application knowledge. \nTesting method: Test through direct questioning, basic data interpretation, or simple logical reasoning, such as directly asking about related application operations, rule selection, or situational judgment."

[0129] For example, "difficulty prompt words" can be configured as "Difficulty: \nDefinition: Difficulty questions are designed with complex application scenarios, containing interference information, multiple variables and possibly incomplete information. Customer service staff are required to be able to comprehensively apply multiple knowledge points, critical thinking and judgment, and solve problems through complex logical reasoning and high-level data analysis. \nRequirements: Customer service staff need to have a strong understanding of application rules, processing principles, processing ideas and judgment in complex and ambiguous situations. This includes being able to distinguish easily confused scenarios, comprehensively consider multiple information points, and use creative thinking and strategy formulation capabilities to find solutions. \nExamination method: Comprehensive examination through case analysis, complex data interpretation and multi-dimensional analysis, advanced logical reasoning, innovation and strategy formulation. For example, analyze the given consumer's buried consultation description, order information and service trajectory, and require solutions that comply with application rules and principles."

[0130] For another example, "Single-choice prompt" can be configured as "Single-choice question", "Multiple-choice prompt" can be configured as "Multiple-choice question (2 to 4 correct answers)", and "Fill-in selection prompt" can be configured as "5. Please note that you are going to give a multiple-choice question with 2 to 4 answers. When outputting the correct answers, use commas to separate them, such as "Correct answer"; "A,B,C,D"\n". The prompt configuration options can also be determined according to actual needs and are not limited here.

[0131] Regarding the response time of the system in this application, there can be a total of four calls in the entire process from the interpretation of the target document to the generation of the test content. The response time directly affects the user experience. This application proposes that the image interpretation task can be customized by the user. The image interpretation task generally takes more than ten seconds. If the user determines that the large model does not need to understand the image content, and the remaining text content is sufficient for the question, the knowledge points in the possible images can be discarded. The knowledge content of the images in most KBS documents is not very critical, and the questions that meet the requirements can be generated without images; the knowledge point disassembly task is initiated asynchronously and completed asynchronously when the user determines the input and configuration; considering that the source of the questions is often a newer KBS article, this application proposes that the scheduled task can pre-process these updated KBS as a cache every day, and the question results will be directly returned when the trainer uses it. The need for questioning can appear after adding new application knowledge, that is, adding KBS. The program can be used to monitor the addition and modification of KBS, and then use these new KBS documents to generate all possible combinations of question types and difficulties as caches, which can reduce user operation costs. At present, statistics show that more than half of the adopted questions are directly returned by the cache.

[0132] Regarding the model selection of this application, for such generative tasks with few calls and low response time requirements, the first thing to consider when selecting a model is the model's capabilities, so larger-scale models are generally preferred. The effects of each model in generating questions were observed, and the accuracy of the questions, the ability to understand the knowledge in the article, and the degree of compliance with the prompt word restrictions were compared. The models with better results can be used in turn. Model (qwen1.5-110B, qwen2-72B-instruct and qwen-max), this application can mainly call model (qwen1.5-110B), and the other two models (qwen2-72B-instruct and qwen-max) can be called as alternatives in longer contexts or when the model availability is reported. The multimodal model can use model (qwen-vl-max), which has a good ability to interpret picture-based knowledge content such as slides.

[0133] The technical solution proposed in this application has multimodal knowledge material support. Most of the products in the related technology do not support the processing of documents containing multimodal materials such as text, tables, flowcharts and screenshots, which will seriously limit the comprehensiveness and accuracy of the questions. This solution can accurately interpret the content of documents in various formats by using a multimodal model (for example, qwen-vl-max). Specifically, we first convert complex target documents, such as HTML tags, into easy-to-understand converted documents, such as Markdown format, and then classify and interpret the pictures and other multimodal content in the document using a multi-discussion prompt word project to ensure that the model can fully and accurately understand and generate high-quality test questions; it has customized support for knowledge points and difficulty. Products in the related technology often do not support customized test question generation for specific knowledge points and difficulty. This application supports users to customize the knowledge points and difficulty levels of questions by designing a flexible prompt word system. Specifically, the knowledge points in the article are disassembled through asynchronous tasks, and then single-choice questions, multiple-choice questions and questions of different difficulties are generated through prompt word configuration, ensuring that the generated test questions can cover specific knowledge points and meet the expected difficulty settings, thereby improving the pertinence and effectiveness of training.

[0134] The technical solution proposed in this application can realize the interpretation of multimodal materials by using a multimodal model (for example, qwen-vl-max) to process document contents in multiple formats including text, tables, flowcharts and screenshots; it can convert complex target documents, for example, HTML tags, into easy-to-understand converted documents, for example, Markdown format, and then classify and interpret the images and other multimodal content in the document using multiple rounds of prompt word engineering, thereby ensuring that the model can fully and accurately understand and generate high-quality test questions; customized knowledge points and difficulty support, through asynchronous tasks to disassemble the knowledge points in the article, provide basic data support for customized question setting, design prompt word engineering for specific knowledge points and difficulty levels, support single-choice questions, multiple-choice questions and customization of questions of different difficulty levels, and can use asynchronous tasks and caching mechanisms to pre-process and cache commonly used data, optimize response time, and improve the real-time performance and user experience of the system.

[0135] According to an embodiment of the present application, a method for generating test data is also provided. It should be noted that the steps shown in the flowchart in 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.

[0136] Figure 8 is a flow chart of a method for generating test question data according to an embodiment of the present application, such as Figure 8 As shown, the following steps may be specifically included:

[0137] Step S802: Acquire the target document by calling the first interface.

[0138] The first interface includes a first parameter, a parameter value of the first parameter includes a target document, and the target document contains multimodal data.

[0139] The above-mentioned first interface may refer to an interface or method that can obtain a target document. The first interface includes a first parameter. The parameter value of the first parameter may be a target document. The target document may contain multimodal data, that is, it may contain a variety of different types of data, such as text, pictures, audio, etc. The first interface may be determined according to actual needs and is not limited here.

[0140] In an optional embodiment, the present application can generate test content that meets specific requirements and rules through certain algorithms and technologies, which can be used in examinations, tests or teaching activities. In the process of generating test data, the first interface can be called through a program or code, and the relevant information of the target document can be passed into the first parameter. According to the first interface, according to the passed-in parameter value, the content of the target document is obtained, including multimodal data such as text, pictures, audio, and video. The target document containing multimodal data can be obtained by calling an interface or method. The target document can be a file downloaded from the Internet, or it can be data obtained from a database query, or other sources. In actual applications, the first interface can be an application interface, a web link, a database query, etc. Obtaining the target document by calling this interface is the first step in generating test data and is a key link in the entire data processing process. Through the above steps, the original data can be obtained, providing a basis for subsequent data processing and test generation.

[0141] Step S804, using the document parsing model to parse the multimodal data to obtain target text data.

[0142] In an optional embodiment, the document parsing model can refer to a model that uses a deep learning model to parse multimodal data and extract target text data from it. The trained document parsing model can be used to parse new multimodal data to extract target text data, and the target text data can be used to generate test data; the training process of the document parsing model can include preparing a data set containing multimodal data, such as a data set containing multiple data types such as text, pictures, and audio. It can be an existing data set or self-collected data. The prepared data is pre-processed, including data cleaning, data standardization, etc., to facilitate subsequent model training, select a suitable deep learning model to construct a document parsing model, and can process multiple data types at the same time to extract target text data from them. The constructed document parsing model is trained using the prepared data set. During the training process, the model will learn how to effectively extract the target text data from the multimodal data, so that the above-mentioned trained document parsing model can be obtained. Through the above steps, the multimodal data can be parsed using the document parsing model to effectively extract the target text data, which provides an effective method for generating test data.

[0143] Step S806, input the target text data into the test question generation model, and use the test question generation model to generate the target test question data.

[0144] In an optional embodiment, a deep learning model can be used as a large model for test question generation to generate test question data that meets specific requirements. A suitable deep learning model can be selected as the large model for test question generation, which can include a recurrent neural network, a long short-term memory network, a transformer, etc., and a suitable model structure can be selected according to actual needs and data characteristics. There is no limitation here. The target text data can be input into the large model for test question generation to generate target test question data. The generated test question data can be adjusted and screened as needed to meet actual needs; the generated test question data can be evaluated and optimized to verify its quality and accuracy. The test question data can be evaluated by manual evaluation, automatic evaluation, etc., and the model can be optimized and improved based on the evaluation results. Through the above steps, the large model for test question generation can be used to generate test question data that meets the requirements, providing more diversified and high-quality test question resources for education, examinations and other fields.

[0145] Step S808, outputting target test question data by calling the second interface.

[0146] The second interface includes a second parameter, and the parameter value of the second parameter includes target test question data.

[0147] In an optional embodiment, the target test question data can be output by calling a second interface, the second interface includes a second parameter, and the parameter value of the second parameter includes the target number of test questions. Specifically, the second interface can be called according to the needs of generating test question data, and the target number of test questions can be passed in as the parameter value of the second parameter. According to the parameter value of the second interface, a corresponding number of test question data can be generated, which can include operations such as randomly generating test question content, setting test question difficulty, and determining test question type. After the test question data is generated, the generated test question data can be output, which can be saved in a file, transmitted to other systems, or displayed to users. Through the above steps, the process of generating test question data according to the target number of test questions can be realized. This process can be customized according to specific needs to meet the test question generation needs in different scenarios.

[0148] According to an embodiment of the present application, a device for generating test question data for implementing the above test question data generation method is also provided. Fig. 9 is a schematic diagram of a device for generating test question data according to an embodiment of the present application, such as Fig. 9 As shown, the device includes: a first display module 902 , a second display module 904 , a third display module 906 , and a fourth display module 908 .

[0149] Among them, the first display module 902 is used to respond to the first input instruction acting on the operation interface, and display the target document corresponding to the input instruction on the operation interface, wherein the target document includes multimodal data; the second display module 904 is used to respond to the conversion instruction acting on the operation interface, and display the target text data corresponding to the conversion instruction on the operation interface, wherein the target text data is text data obtained by parsing the multimodal data using the document parsing large model; the third display module 906 is used to respond to the second input instruction acting on the operation interface, and display the generation conditions corresponding to the second input instruction on the operation interface, wherein the generation conditions include at least one of the following: target type, target difficulty level, target dimension; the fourth display module 908 is used to respond to the generation instruction acting on the operation interface, and display the target test question data on the operation interface, wherein the target test question data is the data generated by inputting the target text data into the test question generation large model and guiding the test question generation large model using the generation conditions.

[0150] Among them, the above-mentioned device also includes a conversion module, which is used to respond to the conversion instruction acting on the operation interface, convert the format of the target document to obtain a converted document; extract the first modal data in the converted document to obtain first text data, wherein the first modal data includes one of the following: text data, table data; use the document parsing large model to perform at least one round of parsing on the second modal data in the converted document to obtain second text data, wherein the second modal data includes: image data; summarize the first text data and the second text data to obtain the target text data.

[0151] Among them, the conversion module is also used to respond to the parsing instruction acting on the operation interface, and use the document parsing model to perform at least one round of parsing on the second modal data through asynchronous calls to obtain second text data.

[0152] Among them, the conversion module is also used to classify the second modal data using the document parsing large model to determine the type of the second modal data; when the type of the second modal data is a preset type, the document parsing large model is used to parse the text data in the second modal data to obtain second text data.

[0153] Among them, the fourth display module is also used to respond to the conversion instruction on the operation interface when the generation conditions include the target dimension, and display multiple dimensions on the operation interface, wherein the multiple dimensions are the results of parsing the dimensions of the target text data using the dimensional parsing model; respond to the second input instruction on the multiple dimensions, and display the target dimension on the operation interface, wherein the target dimension is the dimension selected from the multiple dimensions based on the second input instruction.

[0154] Among them, the above-mentioned device also includes a parsing module, which is used to parse the dimensions of the target text data using a large dimensional parsing model through asynchronous calling to obtain multiple dimensions.

[0155] Among them, the above-mentioned device also includes an acquisition module, which is used to respond to the generation instruction acting on the operation interface, obtain the preset prompt word template corresponding to the test question generation model; determine the prompt word information matching the generation conditions; inject the prompt word information into the preset slot corresponding to the generation conditions in the preset prompt word template to generate target prompt word information; input the target text data into the test question generation model, and use the target prompt word information to guide the test question generation model to generate target test question data.

[0156] The above-mentioned device further includes a determination module, which is used to determine a target document identifier corresponding to the first input instruction; and obtain the target document from a preset document library based on the target document identifier.

[0157] Among them, the fourth display module is also used to display the target test question data on the operation interface, and then respond to the modification instruction acting on the target test question data to display the modified test question data on the operation interface, wherein the modified test question data is the test question data after the target test question data is modified based on the modification instruction; respond to the confirmation instruction acting on the modified test question data to display the export result of the modified test question data on the operation interface, wherein the export result is the result obtained by exporting the modified test question data.

[0158] Among them, the fourth display module is also used to determine whether the target document is a document stored in a preset document library after the target document corresponding to the input instruction is displayed on the operation interface; when the target document is a document stored in the preset document library, multiple test question data corresponding to the target document are obtained from the cache, wherein different test question data are input into the test question generation model by inputting the target text data, and the data generated by the test question generation model are guided by different generation conditions; in response to the generation instruction, the target test question data is displayed on the operation interface, wherein the target test question data is the test question data corresponding to the generation instruction among the multiple test question data.

[0159] Among them, the above-mentioned device also includes a monitoring module, which is used to periodically monitor whether new documents appear in the preset document library, wherein the preset document library is used to store target documents; when new documents appear in the preset document library, the new documents are parsed using the document parsing model to obtain new text data; the new text data is input into the test question generation model, and different generation conditions are used to guide the test question generation model to generate different test question data; and different test question data are stored in the cache.

[0160] According to an embodiment of the present application, a device for generating test question data for implementing the above test question data generation method is also provided. Fig.10 is a schematic diagram of another optional test question data generating device according to an embodiment of the present application, such as Fig.10 As shown, the device includes: a calling module 1002 , a parsing module 1004 , an input module 1006 and an output module 1008 .

[0161] Among them, the calling module 1002 is used to obtain the target document by calling the first interface, wherein the first interface includes a first parameter, the parameter value of the first parameter includes the target document, and the target document includes multimodal data; the parsing module 1004 is used to parse the multimodal data using the document parsing big model to obtain the target text data; the input module 1006 is used to input the target text data into the test question generation big model, and generate the target test question data using the test question generation big model; the output module 1008 is used to output the target test question data by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes the target test question data.

[0162] The embodiment of the present application may provide an electronic device, which may be any electronic device in a group of electronic devices. Optionally, in this embodiment, the electronic device may also be replaced by a terminal device such as a mobile terminal.

[0163] Optionally, in this embodiment, the electronic device may be located in at least one network device among a plurality of network devices of a computer network.

[0164] In this embodiment, the electronic device can execute the program code in the method.

[0165] Optionally, Fig.11 is a structural block diagram of an electronic device according to an embodiment of the present application, such as Fig.11 As shown, the electronic device A may include: one or more (only one is shown in the figure) processors 1102, a memory 1104, a storage controller, and a peripheral interface, wherein the peripheral interface is connected to a radio frequency module, an audio module, and a display.

[0166] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implementing the methods in the above embodiments. The memory 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 instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal A via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0167] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: in response to a first input instruction on the operation interface, a target document corresponding to the input instruction is displayed on the operation interface, wherein the target document includes multimodal data; in response to a conversion instruction on the operation interface, target text data corresponding to the conversion instruction is displayed on the operation interface, wherein the target text data is text data obtained by parsing the multimodal data using a document parsing large model; in response to a second input instruction on the operation interface, a generation condition corresponding to the second input instruction is displayed on the operation interface, wherein the generation condition includes at least one of the following: target type, target difficulty level, target dimension; in response to a generation instruction on the operation interface, target test question data is displayed on the operation interface, wherein the target test question data is the data generated by inputting the target text data into the test question generation large model and guiding the test question generation large model using the generation condition.

[0168] Optionally, in response to a conversion instruction applied to an operation interface, the method further includes: performing format conversion on a target document to obtain a converted document; extracting first modal data from the converted document to obtain first text data, wherein the first modal data includes one of the following: text data, table data; performing at least one round of parsing on second modal data in the converted document using a document parsing model to obtain second text data, wherein the second modal data includes: image data; and aggregating the first text data and the second text data to obtain target text data.

[0169] Optionally, the second modal data in the converted document is parsed at least once using a large document parsing model to obtain second text data, including: responding to a parsing instruction on an operation interface, performing at least once parsing the second modal data using a large document parsing model through an asynchronous call to obtain second text data.

[0170] Optionally, the second modal data is parsed at least once using a large document parsing model to obtain second text data, including: classifying the second modal data using a large document parsing model to determine the type of the second modal data; when the type of the second modal data is a preset type, parsing the text data in the second modal data using the large document parsing model to obtain the second text data.

[0171] Optionally, before responding to a generation instruction on the operation interface, the method also includes: responding to a second input instruction on the operation interface, and displaying generation conditions corresponding to the second input instruction on the operation interface, wherein the generation conditions include at least one of the following: target type, target difficulty level, and target dimension; responding to a generation instruction on the operation interface, and displaying target test question data on the operation interface, wherein the target test question data is data generated by inputting target text data into a test question generation model and using the generation conditions to guide the test question generation model.

[0172] Optionally, when the generation conditions include a target dimension, the method further includes: responding to a conversion instruction acting on an operation interface, displaying multiple dimensions on the operation interface, wherein the multiple dimensions are results of parsing the dimensions of the target text data using a dimensional parsing model; responding to a second input instruction acting on multiple dimensions, displaying the target dimension on the operation interface, wherein the target dimension is a dimension selected from the multiple dimensions based on the second input instruction.

[0173] Optionally, the method further includes: parsing the dimensions of the target text data using a large dimension parsing model through an asynchronous call to obtain multiple dimensions.

[0174] Optionally, in response to a generation instruction on the operation interface, the method also includes: obtaining a preset prompt word template corresponding to the test question generation model; determining prompt word information matching the generation conditions; injecting the prompt word information into a preset slot corresponding to the generation conditions in the preset prompt word template to generate target prompt word information; inputting the target text data into the test question generation model, and using the target prompt word information to guide the test question generation model to generate target test question data.

[0175] Optionally, in response to a first input instruction on the operation interface, the method further includes: determining a target document identifier corresponding to the first input instruction; and acquiring the target document from a preset document library based on the target document identifier.

[0176] Optionally, after displaying the target test question data on the operation interface, the method further includes: responding to a modification instruction acting on the target test question data, displaying the modified test question data on the operation interface, wherein the modified test question data is the test question data after the target test question data is modified based on the modification instruction; responding to a confirmation instruction acting on the modified test question data, displaying an export result of the modified test question data on the operation interface, wherein the export result is a result obtained by exporting the modified test question data.

[0177] Optionally, after displaying the target document corresponding to the input instruction on the operation interface, the method also includes: determining whether the target document is a document stored in a preset document library; when the target document is a document stored in the preset document library, obtaining multiple test question data corresponding to the target document from the cache, wherein different test question data are obtained by inputting target text data into a test question generation model, and using different generation conditions to guide the data generated by the test question generation model; in response to the generation instruction, displaying the target test question data on the operation interface, wherein the target test question data is the test question data corresponding to the generation instruction among the multiple test question data.

[0178] Optionally, the method also includes: regularly monitoring whether new documents appear in a preset document library, wherein the preset document library is used to store target documents; when new documents appear in the preset document library, parsing the new documents using a document parsing big model to obtain new text data; inputting the new text data into a test question generation big model, using different generation conditions to guide the test question generation big model to generate different test question data; and storing different test question data in a cache.

[0179] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0180] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0181] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD), including several instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0182] The embodiment of the present application also provides a computer-readable storage medium. Optionally, in this embodiment, the computer-readable storage medium can be used to store program codes executed by the following method.

[0183] In response to a first input instruction on the operation interface, a target document corresponding to the input instruction is displayed on the operation interface, wherein the target document includes multimodal data; in response to a conversion instruction on the operation interface, target text data corresponding to the conversion instruction is displayed on the operation interface, wherein the target text data is text data obtained by parsing multimodal data using a document parsing large model; in response to a second input instruction on the operation interface, a generation condition corresponding to the second input instruction is displayed on the operation interface, wherein the generation condition includes at least one of the following: target type, target difficulty level, target dimension; in response to a generation instruction on the operation interface, target test question data is displayed on the operation interface, wherein the target test question data is data generated by inputting the target text data into a test question generation large model and guiding the test question generation large model using the generation condition.

[0184] Optionally, in response to a conversion instruction applied to an operation interface, the method further includes: performing format conversion on a target document to obtain a converted document; extracting first modal data from the converted document to obtain first text data, wherein the first modal data includes one of the following: text data, table data; performing at least one round of parsing on second modal data in the converted document using a document parsing model to obtain second text data, wherein the second modal data includes: image data; and aggregating the first text data and the second text data to obtain target text data.

[0185] Optionally, the second modal data in the converted document is parsed at least once using a large document parsing model to obtain second text data, including: responding to a parsing instruction on an operation interface, performing at least once parsing the second modal data using a large document parsing model through an asynchronous call to obtain second text data.

[0186] Optionally, the second modal data is parsed at least once using a large document parsing model to obtain second text data, including: classifying the second modal data using a large document parsing model to determine the type of the second modal data; when the type of the second modal data is a preset type, parsing the text data in the second modal data using the large document parsing model to obtain the second text data.

[0187] Optionally, before responding to a generation instruction on the operation interface, the method also includes: responding to a second input instruction on the operation interface, and displaying generation conditions corresponding to the second input instruction on the operation interface, wherein the generation conditions include at least one of the following: target type, target difficulty level, and target dimension; responding to a generation instruction on the operation interface, and displaying target test question data on the operation interface, wherein the target test question data is data generated by inputting target text data into a test question generation model and using the generation conditions to guide the test question generation model.

[0188] Optionally, when the generation conditions include a target dimension, the method further includes: responding to a conversion instruction acting on an operation interface, displaying multiple dimensions on the operation interface, wherein the multiple dimensions are results of parsing the dimensions of the target text data using a dimensional parsing model; responding to a second input instruction acting on multiple dimensions, displaying the target dimension on the operation interface, wherein the target dimension is a dimension selected from the multiple dimensions based on the second input instruction.

[0189] Optionally, the method further includes: parsing the dimensions of the target text data using a large dimension parsing model through an asynchronous call to obtain multiple dimensions.

[0190] Optionally, in response to a generation instruction on the operation interface, the method also includes: obtaining a preset prompt word template corresponding to the test question generation model; determining prompt word information matching the generation conditions; injecting the prompt word information into a preset slot corresponding to the generation conditions in the preset prompt word template to generate target prompt word information; inputting the target text data into the test question generation model, and using the target prompt word information to guide the test question generation model to generate target test question data.

[0191] Optionally, in response to a first input instruction on the operation interface, the method further includes: determining a target document identifier corresponding to the first input instruction; and acquiring the target document from a preset document library based on the target document identifier.

[0192] Optionally, after displaying the target test question data on the operation interface, the method further includes: responding to a modification instruction acting on the target test question data, displaying the modified test question data on the operation interface, wherein the modified test question data is the test question data after the target test question data is modified based on the modification instruction; responding to a confirmation instruction acting on the modified test question data, displaying an export result of the modified test question data on the operation interface, wherein the export result is a result obtained by exporting the modified test question data.

[0193] Optionally, after displaying the target document corresponding to the input instruction on the operation interface, the method also includes: determining whether the target document is a document stored in a preset document library; when the target document is a document stored in the preset document library, obtaining multiple test question data corresponding to the target document from the cache, wherein different test question data are obtained by inputting target text data into a test question generation model, and using different generation conditions to guide the data generated by the test question generation model; in response to the generation instruction, displaying the target test question data on the operation interface, wherein the target test question data is the test question data corresponding to the generation instruction among the multiple test question data.

[0194] Optionally, the method also includes: regularly monitoring whether new documents appear in a preset document library, wherein the preset document library is used to store target documents; when new documents appear in the preset document library, parsing the new documents using a document parsing big model to obtain new text data; inputting the new text data into a test question generation big model, using different generation conditions to guide the test question generation big model to generate different test question data; and storing different test question data in a cache.

[0195] The embodiment of the present application further provides a computer program product. Optionally, in this embodiment, the computer program product may include a computer program, and the computer program implements the method provided in the embodiment when executed by a processor.

[0196] The above-mentioned computer program product may refer to a software program that has been written, tested and released and can be run on a computer or other device. The computer program product may include an application, an operating system, tool software, etc., which is used to implement specific functions or solve specific problems.

[0197] The embodiments of the present application also provide a computer program product. Optionally, the computer program product may include a non-volatile computer-readable storage medium, which may be used to store a computer program, and when the computer program is executed by a processor, the method provided in the embodiments is implemented.

[0198] The above-mentioned non-volatile computer-readable storage medium may refer to a medium for storing data. The non-volatile computer-readable storage medium can keep the data from being lost when the power is off, and can be used to store long-term data, such as operating systems, applications, and user files. The non-volatile storage medium may include hard disk drives, solid-state drives, optical disks, and flash memory storage devices, etc.

[0199] The embodiment of the present application further provides a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, the method provided in the above embodiment is implemented.

[0200] The above-mentioned computer program may refer to a collection of instructions used to tell a computer to perform a specific task or operation. A computer program may be written by a programmer using a specific programming language and may include algorithms, data structures, logic, and control flows. A computer program may be used for a variety of purposes, including application software, operating systems, and the like.

[0201] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0202] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0203] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0204] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0205] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0206] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for generating test question data, characterized in that: include: In response to a first input instruction acting on an operation interface, displaying a target document corresponding to the input instruction on the operation interface, wherein the target document includes multimodal data; In response to a conversion instruction acting on the operation interface, target text data corresponding to the conversion instruction is displayed on the operation interface, wherein the target text data is text data obtained by parsing the multimodal data using a document parsing model; In response to a second input instruction acting on the operation interface, a generation condition corresponding to the second input instruction is displayed on the operation interface, wherein the generation condition includes at least one of the following: target type, target difficulty level, and target dimension; In response to the generation instruction acting on the operation interface, the target test question data is displayed on the operation interface, wherein the target test question data is the data generated by inputting the target text data into the test question generation model and guiding the test question generation model to generate data using the generation conditions.

2. The method according to claim 1, characterized in that: In response to a conversion instruction acting on the operation interface, the method further includes: Performing format conversion on the target document to obtain a converted document; Extracting first modal data from the converted document to obtain first text data, wherein the first modal data includes one of the following: text data, table data; Using the document parsing macromodel, at least one round of parsing is performed on the second modality data in the converted document to obtain second text data, wherein the second modality data includes: image data; The first text data and the second text data are aggregated to obtain the target text data.

3. The method according to claim 2, characterized in that The step of performing at least one round of parsing on the second modal data in the converted document using the document parsing macro model to obtain second text data includes: In response to the parsing instruction on the operation interface, the second modal data is parsed for at least one round using the document parsing model through asynchronous calling to obtain the second text data.

4. The method according to claim 2, characterized in that: The step of performing at least one round of parsing on the second modal data using the document parsing macromodel to obtain the second text data includes: Classifying the second modal data using the document parsing macromodel to determine a type of the second modal data; In the case where the type of the second modal data is a preset type, the text data in the second modal data is parsed using the document parsing large model to obtain the second text data.

5. The method according to claim 1, characterized in that In the case where the generation condition includes the target dimension, the method further includes: In response to a conversion instruction acting on the operation interface, a plurality of dimensions are displayed on the operation interface, wherein the plurality of dimensions are the result of parsing the dimensions of the target text data using a dimensional parsing macromodel; In response to the second input instruction acting on the multiple dimensions, the target dimension is displayed on the operation interface, wherein the target dimension is a dimension selected from the multiple dimensions based on the second input instruction.

6. The method according to claim 5, characterized in that The method further comprises: The dimension parsing big model is used to parse the dimension of the target text data through an asynchronous calling method to obtain the multiple dimensions.

7. The method according to claim 1, characterized in that In response to the generating instruction acting on the operation interface, the method further comprises: Obtaining a preset prompt word template corresponding to the test question generation model; Determining prompt word information matching the generation condition; Injecting the prompt word information into a preset slot corresponding to the generation condition in the preset prompt word template to generate target prompt word information; The target text data is input into the test question generation model, and the target prompt word information is used to guide the test question generation model to generate the target test question data.

8. The method according to claim 1, characterized in that: In response to a first input instruction acting on the operation interface, the method further includes: Determining a target document identifier corresponding to the first input instruction; The target document is acquired from a preset document library based on the target document identifier.

9. The method according to any one of claims 1 to 8, characterized in that After the target test question data is displayed on the operation interface, the method further includes: In response to a modification instruction acting on the target test question data, displaying modified test question data on the operation interface, wherein the modified test question data is test question data after the target test question data is modified based on the modification instruction; In response to a confirmation instruction acting on the modified test question data, an export result of the modified test question data is displayed on the operation interface, wherein the export result is a result obtained by exporting the modified test question data.

10. The method according to any one of claims 1 to 8, characterized in that After displaying the target document corresponding to the input instruction on the operation interface, the method further includes: Determining whether the target document is a document stored in a preset document library; In the case where the target document is a document stored in the preset document library, a plurality of test question data corresponding to the target document is obtained from the cache, wherein different test question data are generated by inputting the target text data into the test question generation model and guiding the test question generation model to generate data using different generation conditions; In response to the generating instruction, the target test question data is displayed on the operation interface, wherein the target test question data is the test question data corresponding to the generating instruction among the plurality of test question data.

11. The method according to claim 10, characterized in that The method further comprises: Regularly monitoring whether a new document appears in a preset document library, wherein the preset document library is used to store the target document; When the newly added document appears in the preset document library, the newly added document is parsed using the document parsing model to obtain newly added text data; Inputting the newly added text data into the test question generation model, and using the different generation conditions to guide the test question generation model to generate the different test question data; The different test question data are stored in the cache.

12. A method for generating test data, characterized in that: include: Acquire a target document by calling a first interface, wherein the first interface includes a first parameter, a parameter value of the first parameter includes the target document, and the target document includes multimodal data; Parsing the multimodal data using a large document parsing model to obtain target text data; Inputting the target text data into a test question generation model, and using the test question generation model to generate target test question data; The target test question data is output by calling a second interface, wherein the second interface includes a second parameter, and a parameter value of the second parameter includes the target test question data.

13. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 12 when running.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 12.

15. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 12.

Images