Question bank generation method and device, equipment and storage medium

Abstract

This invention provides a question bank generation method, apparatus, device, and storage medium. The method includes: acquiring knowledge points; determining the content, knowledge type, and cognitive level of the knowledge points; based on the knowledge type and cognitive level of the knowledge points, and according to preset question type matching rules, determining the question types and corresponding question structures for the knowledge points; using natural language generation technology, constructing the question stems for the knowledge points based on the content of the knowledge points; verifying the questions for the knowledge points; and generating a question bank after successful verification. The technical solution of this application, by determining the knowledge type and cognitive level of the knowledge points; and by determining the question types and corresponding question structures for the knowledge points based on the knowledge type and cognitive level of the knowledge points and according to preset question type matching rules, enables questions to accurately match the examination objectives, thereby improving user experience.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a question bank generation method, apparatus, device, and storage medium. Background Technology

[0002] In related technologies, question bank generation largely relies on manual question creation, which is inefficient and costly, making it difficult to meet educational needs. Automated generation technologies are mostly based on simple templates or rules, resulting in similar question types that fail to reflect the cognitive dimensions, logical structures, and other multidimensional characteristics of knowledge points. They also lack sufficient in-depth exploration of the connections between knowledge points, leading to poor matching between questions and knowledge points and an inability to accurately assess learners' knowledge mastery. Summary of the Invention

[0003] The technical problem to be solved by this invention is that, in related technologies, the matching between questions and knowledge points is poor, making it impossible to accurately assess the learner's level of knowledge mastery. In response to the deficiencies in the existing technology, this invention provides a question bank generation method, apparatus, device, and storage medium.

[0004] To address the aforementioned technical problems, this invention provides a question bank generation method, comprising: Acquire knowledge points; Determine the content, knowledge type, and cognitive level of the knowledge points; The cognitive hierarchy is determined according to Bloom's taxonomy. Based on the knowledge type and cognitive level of the knowledge point, and according to the preset question type matching rules, the question type of the knowledge point and the question structure corresponding to the question type are determined; Using natural language generation technology, construct the question stems for the knowledge points based on their content; The questions on the knowledge points are validated, and a question bank is generated after the validation is passed.

[0005] In one implementation, determining the cognitive level of the knowledge point includes: classifying the cognitive level of the knowledge point into: memory, understanding, application, analysis, evaluation, or creation according to Bloom's Taxonomy.

[0006] In one implementation, the knowledge types include: factual knowledge, conceptual knowledge, and procedural knowledge.

[0007] In one implementation, the question types include: multiple choice questions, true / false questions, and short answer questions; The question structure of the multiple-choice question includes: a stem and multiple options; The question structure for judging it includes: the question stem and the judgment area; The structure of the short answer questions includes: the question stem and the answer area.

[0008] In one implementation, after generating the question bank, the method further includes: using a clustering algorithm to perform clustering calculations on the questions in the question bank, identifying duplicate questions, and deleting the duplicate questions.

[0009] In one implementation, each topic is assigned a search tag to support user searches.

[0010] In one implementation, upon first use, a user profile is determined through test questions, and the user's ability level is automatically assessed. Based on the stated ability level, determine the question type and question structure; The user profile is updated in real time during the question-and-answer process; New questions are generated based on the updated user profile.

[0011] Secondly, a question bank generation device includes: The acquisition module is used to acquire knowledge points; The first determining module is used to determine the content, knowledge type, and cognitive level of the knowledge point; The cognitive hierarchy is determined according to Bloom's taxonomy. The second determining module is used to determine the question type of the knowledge point and the question structure corresponding to the question type based on the knowledge type and cognitive level of the knowledge point and according to the preset question type matching rules. The question stem construction module is used to construct question stems for the knowledge points based on the content of the knowledge points using natural language generation technology. The verification and generation module is used to verify the questions on the knowledge points. After the verification is successful, a question bank is generated.

[0012] Thirdly, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the question bank generation method as described in any of the preceding claims.

[0013] Fourthly, a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the steps of the question bank generation method described in any of the preceding claims.

[0014] The technical solution of this application has the following beneficial effects: by determining the content, knowledge type, and cognitive level of the knowledge point; based on the knowledge type and cognitive level of the knowledge point, and according to the preset question type matching rules, the question type of the knowledge point and the question structure corresponding to the question type are determined, so that the questions accurately match the examination targets, improve the user experience, facilitate intelligent analysis of the multi-dimensional characteristics of the knowledge point, and dynamically generate high-quality and diversified test questions to meet the needs of accurately assessing learners' knowledge mastery. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the question bank generation method in an embodiment of the present invention; Figure 2 This is a schematic diagram of a webpage illustrating a question-setting client according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the question bank generation device shown in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device shown in an embodiment of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0017] like Figure 1 As shown in the figure, an embodiment of the present invention provides a question bank generation method, including: In step S102, knowledge points are acquired.

[0018] Knowledge point acquisition and parsing involves extracting raw knowledge content from multiple sources such as documents and databases, and then performing in-depth analysis using natural language processing techniques. Through models such as named entity recognition, core concepts and key entities in the text are accurately identified and extracted.

[0019] In step S104, the content, knowledge type, and cognitive level of the knowledge point are determined; the cognitive level is determined according to Bloom's Taxonomy.

[0020] Knowledge point characteristic analysis and annotation are the foundation for dynamically adapting to question types and generating content. The analyzed knowledge points undergo multi-dimensional characteristic analysis and annotation, primarily including: Cognitive dimension classification: Based on Bloom's Taxonomy, knowledge points are divided into levels such as memorization, understanding, application, analysis, evaluation, and creation. Logical structure analysis: Determining the inherent type of knowledge points, such as factual knowledge, conceptual knowledge, and procedural knowledge.

[0021] In Bloom's hierarchy of cognitive levels, the memory level focuses on memorizing key concepts, such as the names of core bicycle parts or the first step in chain maintenance. The comprehension level assesses understanding of the meaning of concepts, such as the specific logic behind the energy-saving principle of a bicycle or why mountain bikes are suitable for off-road riding. The application level focuses on practically applying knowledge to solve problems, such as using troubleshooting steps to locate the fault in a bicycle with brake failure or choosing the appropriate type of bicycle based on road conditions. Bloom's hierarchy of cognitive levels sets standards and levels for learning requirements, making the assessment of knowledge points more precise. Users can filter questions according to their cognitive goals; for example, beginners can practice memory / comprehension questions, while advanced learners can practice application questions, resulting in a more practical question bank. Different cognitive levels correspond to different assessment objectives: the memory level tests whether the knowledge is memorized (e.g., bicycle part names, true / false questions); the comprehension level tests whether the knowledge is understood (e.g., energy-saving principle, multiple-choice questions); and the application level tests whether the knowledge can be applied (e.g., troubleshooting, short-answer questions), ensuring that questions are relevant to the assessment needs.

[0022] In step S106, based on the knowledge type and cognitive level of the knowledge point, and according to the preset question type matching rules, the question type of the knowledge point and the question structure corresponding to the question type are determined.

[0023] Based on the above characteristic analysis, the system dynamically matches or generates the most suitable question type and structure for each knowledge point from a pre-set question type rule base. For example, for memory-based knowledge points, it automatically generates fill-in-the-blank or multiple-choice questions. For application-based knowledge points, it automatically generates short-answer or case analysis questions.

[0024] Table 1. Two-Dimensional Question Determination Table Each question is tagged with its knowledge point type and cognitive level. The server has a built-in combination rule table. When generating a question, the two tags are automatically read and matched with the corresponding question type and structure, without the need for manual intervention.

[0025] In step S108, natural language generation technology is used to construct the question stem of the knowledge point based on the content of the knowledge point.

[0026] Utilizing natural language generation technology, it constructs semantically clear and grammatically correct question stems based on knowledge points. It generates definitive answers for objective questions and reference answer points or scoring criteria for subjective questions. For multiple-choice questions, it employs various strategies such as synonym substitution and conceptual confusion to generate effective distractors.

[0027] In step S110, the questions for the knowledge points are verified, and a question bank is generated after the verification is passed.

[0028] In this embodiment, a rule engine is used to verify the consistency of the answers and the rationality of the logic.

[0029] After the question bank is generated, it is output and managed, and the final optimized questions are imported into the question bank. It also supports multi-dimensional filtering and searching based on knowledge point characteristics. It supports exporting to multiple formats, such as JSON, DOCX, and XLSX, for easy integration into learning and training systems.

[0030] The technical solution of this application determines the content, knowledge type, and cognitive level of the knowledge point; based on the knowledge type and cognitive level of the knowledge point, and according to the preset question type matching rules, determines the question type of the knowledge point and the question structure corresponding to the question type, so that the question accurately matches the examination target and improves the user experience.

[0031] In some embodiments, the question types include: multiple-choice questions, true / false questions, and short-answer questions. The multiple-choice question structure includes: a stem and multiple options. The true / false question structure includes: a stem and a judgment area. The short-answer question structure includes: a stem and an answer area.

[0032] In some embodiments, after generating the question bank, the method further includes: using a clustering algorithm to perform clustering calculations on the questions in the question bank, identifying duplicate questions, and deleting the duplicate questions.

[0033] In this embodiment, a clustering algorithm is used to deduplicate the generated questions, ensuring the diversity and uniqueness of the question bank.

[0034] In some embodiments, see Appendix Figure 2 Each topic is assigned search tags to support user searches. The search tags include: whether a 3D model exists and the equipment type.

[0035] The following explanation uses user-answered questions in the question-answering client as an example. Of course, the same applies to user-created questions in the question-creating client.

[0036] On the quiz client's page, users can search in the search bar by entering keywords, including but not limited to "equipment type" (e.g., tanks, drones, bicycles, televisions, etc.). Keywords can also include "3D model," allowing users to view the 3D model after selection. Additionally, keywords can include "three-view drawings," which users can view after selection.

[0037] Specifically, the quiz client sends a search request to the server. Upon receiving the request, the server searches its question bank based on the search keywords. If the tags in the question bank match the search keywords, the search is successful. For example, if the user enters the keyword "drone," the search message is sent to the server. The server, based on the received message, searches its question bank for drones. If the question bank contains a "drone" tag, the search is successful, and the server sends drone-related questions from the question bank to the quiz client. The user's quiz client will then display these drone-related questions on its display page.

[0038] On the user's answer client page, each keyword can be set as a drop-down list. For example, in the equipment type search bar, a drop-down list can be set. When a user clicks the drop-down list, specific equipment will be displayed, including but not limited to drones, missiles, and ships. The user can then select specific equipment from the drop-down list.

[0039] The equipment types mentioned above belong to the visible light template category. Alongside the visible light template, infrared templates and radar SAR templates can also be set. The equipment in the visible light template is all visible light imaging equipment; the infrared template can also include various subjects captured by infrared light, and the knowledge points for these subjects can be set in a question bank for retrieval.

[0040] In some embodiments, the generated questions also include three-view drawings of a 3D model of the subject involved in the question. Users can click to view the images on the answering client to see the three-view drawings. For example, when answering a multiple-choice question about fighter jets, after displaying the answer, users can click to view the images to see the fighter jet's three-view drawings.

[0041] In some embodiments, the following attribute parameters may also be generated for each question, including but not limited to: number of references, score, number of test papers, and difficulty.

[0042] In a question bank setting, these three metrics are the core dimensions for measuring the value and usage of questions.

[0043] The score indicates the weight of a question in the exam paper, directly determining its impact on the total score. For example, in a 10-point word problem and a 2-point multiple-choice question, a higher score indicates greater importance of the question in the exam; usually, difficult questions and questions covering core knowledge points are worth more points. The score is used to control the overall score structure of the exam paper during compilation, such as the distribution of points between single-choice, multiple-answer, and subjective questions, and also affects students' time allocation when answering questions.

[0044] Citation count refers to the number of times a question has been cited, that is, the cumulative number of times the question has been officially used, such as being included in exam papers or question bank collections. The more times a question is cited, the more popular it is. For example, if a math problem is cited in 100 sets of exam papers, it indicates that its examination angle and difficulty level are widely recognized. The purpose is to select high-quality questions; highly cited questions are usually more relevant to the exam points, avoiding the use of obscure or controversial questions.

[0045] The number of times a question has been used in test paper creation refers to the cumulative number of times a question has been successfully included in the final test paper through a test paper creation system, such as teacher-created test papers or system-generated test papers. It is similar to the number of citations, but focuses more on actual test paper creation usage. Some questions may be saved but not actually used in test papers; for example, if a question is used by 50 teachers in their test paper creation, the number of test paper creations is 50. This reflects the frequency of the question's use in actual teaching and assessment, helping users quickly identify frequently used questions.

[0046] Difficulty level is a key indicator for question bank compilation and student learning analysis. It includes three types: general, easy, and difficult. The basic question bank contains low-difficulty questions, while the advanced question bank contains high-difficulty questions.

[0047] When editing questions in the question-creating client, the aforementioned attribute parameters can be displayed on the client's page to assist the question creator. For any given question, the client also provides a delete button for deleting the question and an edit button for modifying it. Clicking "edit" allows for further editing of the question, including editing the question stem and each option.

[0048] In some embodiments, in addition to combining the two dimensions mentioned above, questions can also be generated in real time based on user profiles. Specifically, when a user uses the question bank for the first time, the user profile is determined through test questions, and the user's ability level is automatically determined; based on the ability level, the question type and question structure are determined; during the answering process, the user profile is updated in real time; and new questions are generated based on the updated user profile.

[0049] In this embodiment, the user's profile tags can take the following form: Ability level tags: Beginner, Intermediate, Expert.

[0050] Learning objective tags: basic consolidation, focusing on memorization / understanding; ability improvement, focusing on application; exam preparation, focusing on comprehensiveness. Answering habit tags, optional, to improve the experience: Prefer multiple choice questions, prefer short answer questions, short answer time.

[0051] The following table uses conceptual knowledge and application layer analysis of SAR imaging principles as an example, and shows the adaptation table for different user profiles: Table 2

[0052] When a user uses the question bank for the first time, they are given a test of 3-5 questions. The system automatically determines the user's profile, which includes their ability level. If the user answers all questions incorrectly, their ability level is determined to be beginner. If they answer all questions correctly, their ability level is determined to be expert. If a user answers some questions correctly, their ability level can be determined based on the percentage of correct answers. This advancement can be further divided into three levels: Level 1, Level 2, and Level 3. The specific level can be determined by setting a threshold for the percentage of correct answers.

[0053] During the user's quiz, the system updates the user profile tags in real time, such as improving accuracy and progressing from beginner to advanced level. When generating questions, the system first reads user profile tags, then combines the knowledge point type and cognitive level, calls the corresponding adjustment rules, and dynamically generates suitable questions.

[0054] During the quiz, the user profile is updated in real time; new questions are generated based on the updated user profile. For example, with 100 questions, if a user has answered a predetermined percentage threshold of questions (e.g., 0.6, though this can be flexibly set), and the user has answered 60 questions, their profile is updated based on their answers to those 60 questions. For instance, if all 60 questions are answered correctly, the user's ability level is adjusted from beginner to expert. The remaining 40 questions will then be progressively more difficult.

[0055] For example, if we discover that users have specific answering habits when answering these 60 questions, such as a preference for questions with short stems, we can update the user's habits by setting the stems for the remaining 40 questions to be shorter. Alternatively, we can increase the length of the stems to correct the user's answering habits and prevent them from struggling to adapt to real exam questions.

[0056] In some embodiments, in addition to the two dimensions and user profile, a question scenario tag is added. For the same knowledge point, cognitive level, and user, questions are generated in different scenarios to fit the purpose of the scenario, avoiding the use of the same set of questions for all scenarios.

[0057] In this embodiment, question-setting scenario tags are defined, and the scenarios include, but are not limited to, one or more of the following: Classroom practice scenario: The goal is to quickly consolidate knowledge points, focusing on simple and fast-paced answers; Homework scenario: The purpose is to deepen understanding, focusing on medium difficulty and a small number of subjective questions; Exam and assessment scenarios: The purpose is to differentiate abilities, with a focus on comprehensive coverage and design including common mistakes; Practical training scenario: The aim is to improve practical skills, focusing on step details and case-based questions.

[0058] See Table 3 for the correspondence between the question scenarios and the questions.

[0059] Table 3 In this embodiment, when generating questions, the question-generating scenario is first selected, such as the user selecting an exam or assessment. The scenario tags, user profile, and dual-dimensional tags are read, and the corresponding scenario's question type rules are invoked to generate questions. The number of questions, answering time, and parsing display permissions are set for different scenarios; for example, the parsing is hidden in the exam scenario and displayed in the practice scenario.

[0060] In some embodiments, determining the cognitive level of a knowledge point can be achieved using a machine learning model. The machine learning model can employ a Transformer, which implements the determination of the cognitive level of a knowledge point based on Bloom's Taxonomy. The core of this approach is to transform the level determination into a text classification task, using the Transformer's Encoder layer to capture the semantic features of the knowledge point text, and then mapping them to Bloom's six cognitive levels through the classification head.

[0061] Bloom's Taxonomy has six cognitive levels as follows: Memory: Identifying and recalling knowledge points, such as stating the content of the Pythagorean theorem.

[0062] Understanding: Explaining and summarizing knowledge points, such as explaining the Pythagorean theorem in your own words.

[0063] Application: Apply the knowledge points to new scenarios, such as using the Pythagorean theorem to calculate the side length of a right triangle.

[0064] Analysis: Break down and compare the knowledge points, such as analyzing the differences between the Pythagorean theorem and the law of cosines.

[0065] Evaluation: Judge and demonstrate the value of knowledge points, such as evaluating the role of the Pythagorean theorem in geometric proofs.

[0066] Creation: Generate new content based on knowledge points, such as designing a right triangle teaching aid using the Pythagorean theorem.

[0067] It is evident that each of the above cognitive levels has specific semantic features. Cognitive levels are determined through these specific semantic features.

[0068] Data Preparation. Collect samples, each containing input text and a label; the label represents the cognitive level to which the input text belongs. The sample size should be at least several thousand samples, with a label assigned to each sample. The labels should be evenly distributed across the overall sample set, for example, 1000 samples across 6 cognitive levels out of 6000 samples. This facilitates comprehensive model training and avoids cognitive errors in the model.

[0069] Text preprocessing adapts the input format to Transformer, converting natural language text into vectors that Transformer can recognize.

[0070] The input layer receives the preprocessed vector; The Encoder layer is used to extract semantic features from the vector; The classification head is used to determine the cognitive level based on semantic features and output the cognitive level. It outputs the probabilities of 6 levels and takes the level with the highest probability as the final result.

[0071] For example, if the input knowledge point is: Analyze the internal logic of Newton's three laws of motion, the model will output that this knowledge point belongs to the "Analysis" level.

[0072] The training process is as follows: input the training set samples into the model to obtain the predicted hierarchical probabilities, use cross-entropy loss to calculate the difference between the predicted probabilities and the true labels, backpropagate to update all parameters of the model, including the Transformer Encoder layer and the classification head, evaluate the accuracy with the validation set after each round of training, and retain the model with the highest accuracy.

[0073] Secondly, this application proposes a question bank generation device, see appendix. Figure 3 ,include: Module 21 is used to retrieve knowledge points; The first determining module 22 is used to determine the content, knowledge type, and cognitive level of the knowledge point; The cognitive hierarchy is determined according to Bloom's taxonomy. The second determining module 23 is used to determine the question type of the knowledge point and the question structure corresponding to the question type based on the knowledge type and cognitive level of the knowledge point and according to the preset question type matching rules; The question stem construction module 24 is used to construct the question stem of the knowledge point based on the content of the knowledge point using natural language generation technology; The verification generation module 25 is used to verify the questions on the knowledge points. After the verification is passed, a question bank is generated.

[0074] In some embodiments, a deduplication module is also included, which is used to verify that after the generation module 25 generates the question bank, a clustering algorithm is used to perform clustering calculations on the questions in the question bank to identify duplicate questions and delete the duplicate questions.

[0075] In some embodiments, a tagging module is also included, which sets search tags for each topic to support user searches.

[0076] In some embodiments, a judgment module is also included, which, upon first use, determines the user profile through test questions and automatically determines the user's ability level. The third determining module is used to determine the question type and question structure based on the ability level. The update module is used to update the user profile in real time during the question-answering process; and generate new questions based on the updated user profile.

[0077] Thirdly, see appendix. Figure 4 This application proposes an electronic device including a memory 32, a processor 31, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the question bank generation method as described in any of the preceding claims.

[0078] The aforementioned electronic devices can be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. These electronic devices may include, but are not limited to, processors and memory. Those skilled in the art will understand that the figures are merely examples of electronic devices and do not constitute a limitation on the electronic devices. They may include more or fewer components than illustrated, or combine certain components, or different components. For example, the aforementioned electronic devices may also include input / output devices, network access devices, buses, etc.

[0079] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0080] Fourthly, this application proposes a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the steps of the question bank generation method described in any of the preceding claims.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating a question bank, characterized in that, include: Acquire knowledge points; Determine the content, knowledge type, and cognitive level of the knowledge points; The cognitive hierarchy is determined according to Bloom's taxonomy. Based on the knowledge type and cognitive level of the knowledge point, and according to the preset question type matching rules, the question type of the knowledge point and the question structure corresponding to the question type are determined; Using natural language generation technology, construct the question stems for the knowledge points based on their content; The questions on the knowledge points are validated, and a question bank is generated after the validation is passed.

2. The question bank generation method according to claim 1, characterized in that, Determining the cognitive levels of the knowledge points includes: classifying the cognitive levels of the knowledge points into: memory, understanding, application, analysis, evaluation, or creation according to Bloom's Taxonomy.

3. The question bank generation method according to claim 1, characterized in that, The knowledge types mentioned include: factual knowledge, conceptual knowledge, and procedural knowledge.

4. The question bank generation method according to claim 1, characterized in that, The question types include: multiple choice, true / false, and short answer questions; The question structure of the multiple-choice question includes: a stem and multiple options; The question structure for judging it includes: the question stem and the judgment area; The structure of the short answer questions includes: the question stem and the answer area.

5. The question bank generation method according to claim 1, characterized in that, After generating the question bank, the method further includes: using a clustering algorithm to perform clustering calculations on the questions in the question bank, identifying duplicate questions, and deleting the duplicate questions.

6. The question bank generation method according to claim 1, characterized in that, Each question is assigned a search tag to support user searches.

7. The question bank generation method according to claim 6, characterized in that, Upon first use, a user profile is established through a test, and the user's ability level is automatically determined. Based on the stated ability level, determine the question type and question structure; The user profile is updated in real time during the question-and-answer process; New questions are generated based on the updated user profile.

8. A question bank generation device, characterized in that, include: The acquisition module is used to acquire knowledge points; The first determining module is used to determine the content, knowledge type, and cognitive level of the knowledge point; The cognitive hierarchy is determined according to Bloom's taxonomy. The second determining module is used to determine the question type of the knowledge point and the question structure corresponding to the question type based on the knowledge type and cognitive level of the knowledge point and according to the preset question type matching rules. The question stem construction module is used to construct question stems for the knowledge points based on the content of the knowledge points using natural language generation technology. The verification and generation module is used to verify the questions on the knowledge points. After the verification is successful, a question bank is generated.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the question bank generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the program instructions are executed by the processor, they implement the steps of the question bank generation method according to any one of claims 1 to 7.

Images