A heuristic question bank and game generation method, device and electronic equipment

Abstract

This application discloses a heuristic question bank and game generation method, device, and electronic device, relating to the field of computer technology. This application pre-builds a heuristic question bank architecture containing nested questions across multiple hierarchical structures. It generates level content through bidirectional parsing of tabular and plain text question bank files, coupled with hierarchical recursive answering interaction logic, ultimately generating two types of answer result files simultaneously. The tabular question bank file reduces the manpower required for teachers to build the question bank, while the hierarchical structure allows for personalized heuristic training targeting weak knowledge points, avoiding the problem of repetitive questions on already mastered knowledge points. Simultaneously, both answer result files can completely retain all user answer data, facilitating teachers to directly perform learning statistics based on the tabular answer result files. The question bank structure and question-and-answer content can be edited and interactively run offline without relying on a database, adapting to practical training scenarios.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a heuristic question bank and game generation method, apparatus and electronic device. Background Technology

[0002] When using gamified educational interactive methods to increase learning interest, integrating knowledge-based training into game levels requires teachers to consume a significant amount of time and energy to create question banks. Furthermore, the questions often lack interrelationships, progressive levels, and complex cyclical structures. Current methods use question banks to store a large number of unstructured documents, making hierarchical question management impossible. Even when hierarchical question management is employed, it only differentiates questions by difficulty level, failing to categorize and classify them according to the various knowledge points within the teaching objectives. The question generation mode is also random or sequential, leading to the repetition of questions on already mastered knowledge points and difficulty in tailoring questions on unmastered knowledge points to the specific teaching objectives, making it unsuitable for personalized enhancement of knowledge mastery. Summary of the Invention

[0003] This application provides a heuristic question bank and game generation method, device, and electronic device to at least solve the problems in related technologies where gamified educational interaction methods require teachers to spend a lot of time and energy to set up question banks. The questions are usually not interconnected, lack progressive and complex looping methods, which leads to the repetition of questions on knowledge points that have been mastered, and the difficulty in adapting and refining questions on knowledge points that have not been mastered according to teaching objectives, making it unsuitable for personalized enhancement of knowledge point mastery.

[0004] This application provides a heuristic question bank and game generation method, including: The structure of the heuristic question bank includes multiple game levels, each game level contains multiple knowledge points, each knowledge point corresponds to multiple sets of questions, and each set of questions contains multi-level nested questions that are related in a tree structure. Edit a tabular question bank file containing heuristic question bank content. When the tabular question bank file is obtained, convert the tabular question bank file into a plain text question bank file. Parse the plain text question bank file to obtain each set of questions corresponding to each knowledge point in each game level, and extract at least one question from each layer of the corresponding tree structure in each set of questions to form the level content. When interaction with the level content is detected, the questions at the top level of the corresponding tree structure are displayed. The answers to each question are collected and it is determined whether the answer is correct. If the answer is correct, the knowledge point corresponding to the current question is skipped. If the answer is incorrect, the next level of questions is displayed. If all questions at the current level are answered correctly, the knowledge point corresponding to the current level is skipped. Otherwise, the next level of questions is displayed, and so on, until all questions at the bottom level are answered correctly, at which point the knowledge point corresponding to the current question is skipped. When the interaction of the level ends, the difficulty level of the level is obtained. If the difficulty level of the level is easy, the user is prompted with a success message and final score based on the number of questions answered and the total time consumed. If the difficulty level of the level is medium or hard, and the user fails to pass all levels smoothly, the user is returned to failure, and questions are drawn again to form a new level and a new round of challenges until all levels can be completed smoothly in one go. The total time consumed and the number of questions answered are then displayed as the final score.

[0005] Furthermore, each group of questions contains multi-level nested questions arranged in a tree structure, including: Each group of questions contains at least multiple top-level questions, multiple sub-level questions corresponding to each top-level question, and multiple sub-sub-level questions corresponding to each sub-level question; The top-level questions are placed at the top level of the tree structure, the sub-level questions are placed at the second-to-top level of the tree structure, the second-to-second sub-level questions are placed at the second-to-last level of the tree structure, and so on. Questions belonging to the same knowledge point are linked together according to adjacent levels to form a group of questions corresponding to the same knowledge point.

[0006] Furthermore, the step of converting the tabular question bank file into a plain text question bank file includes: Obtain the tabular question bank file, and read the global variable row and question content row line by line according to the cell structure of the tabular question bank file to obtain the tabular structured data; Based on the structured data in the table, the content of each cell is converted into the corresponding XML node content according to the preset XML tag mapping rules to obtain a plain text format question bank file; The tabular question bank file and the plain text question bank file support bidirectional lossless conversion. When the plain text question bank file is obtained, the XML node content is reverse-parsed according to the XML tag mapping rules and written into the corresponding cell to obtain the tabular question bank file.

[0007] Furthermore, the tabular question bank file containing the heuristic question bank content includes: Obtain the subject name, chapter information, students' basic level, and teaching materials, and generate a question bank generation request based on the preset heuristic question bank prompt strategy; The question bank generation request is sent to the large language model, and the combination of question stem, answer, distractor and solution content returned by the large language model is obtained to obtain the original question bank data; Based on the original question bank data, a tabular question bank file is generated by organizing the data according to the structure template of the tabular file.

[0008] Furthermore, the method also includes: The tabular question bank file is manually verified. The manually verified tabular question bank file is converted into a plain text question bank file. The question hierarchy structure, question stem description, answer and distractor configuration in the plain text question bank file are extracted as fine-tuning training samples. The large language model is fine-tuned and trained based on the fine-tuning training samples to obtain a fine-tuned model that adapts to the requirements of heuristic question bank generation. When a command to generate question bank data is received, the fine-tuned model is invoked to generate target question bank data that matches the teacher's question-setting style and knowledge point coverage requirements, and a tabular question bank file is generated based on the target question bank data.

[0009] Furthermore, when interacting with the level content, it also includes: When a student exits the level interaction page, the current answering progress data is obtained. The answering progress data includes the current question position, the answering results of questions already answered, and the status of uncompleted knowledge points. The answer progress data is uploaded and stored in the server database and associated with the corresponding student's account identifier; When a student is detected to have logged in again or reopened the check-in page, the system retrieves the answer progress data corresponding to the account identifier from the server database and restores the student to the answer state at the time of the last logout based on the answer progress data.

[0010] Furthermore, when interacting with the level content, it also includes: When a student is detected to have exited the level interaction page, the IP address, MAC address or student ID stored in the browser's long-term cookie of the device used by the student to log in is obtained, and the IP address, MAC address and student ID are associated and recorded in the login device log of the corresponding student account. If it is detected that the student is using a browser instead of a client application and the MAC address cannot be obtained, the student's first login ID is saved in a permanent cache as an alternative identification information for the MAC address. Regularly compare the login device logs corresponding to different student accounts. When multiple different student accounts are detected with the same IP address or the same MAC address and student ID, mark the multiple different student accounts as suspected boosting accounts. A list of suspected account boosting services is generated and pushed to the teacher's backend for manual verification by the teacher based on the students' daily performance.

[0011] Furthermore, the step of extracting at least one question from each layer of the corresponding tree structure in each group of questions to form the level content also includes: Retrieve multiple parallel questions at the same level within each group of questions, and retrieve the group identifier and the list of questions at each level of the corresponding tree structure for each group of questions; For questions at the same level under the same group identifier, at least one question is selected as the question to be displayed in the current level, according to a random selection strategy or a sequential selection strategy.

[0012] This application also provides a heuristic question bank and game generation device, including: The question bank structure setting module is used to set the structure of the heuristic question bank, which includes multiple game levels, each game level contains multiple knowledge points, each knowledge point corresponds to multiple sets of questions, and each set of questions contains multi-level nested questions that are related in a tree structure. The level content generation module is used to edit a tabular question bank file with heuristic question bank content. When the tabular question bank file is obtained, it is converted into a plain text question bank file. The plain text question bank file is parsed to obtain each set of questions corresponding to each knowledge point in each game level. At least one question is extracted from each layer of the corresponding tree structure in each set of questions to form the level content. The interactive question-answering module is used to display the questions at the top level of the corresponding tree structure when interaction with the level content is detected. It collects the answers to each question and determines whether the answer is correct. If the answer is correct, it skips the knowledge point corresponding to the current question. If the answer is incorrect, it displays the next level of questions. If all questions at the current level are answered correctly, it skips the knowledge point corresponding to the current level. Otherwise, it displays the next level of questions, and so on, until all questions at the bottom level are answered correctly, at which point it skips the knowledge point corresponding to the current question. The answer result feedback module is used to obtain the difficulty level of the level content when the interaction of the level content ends. If the difficulty level of the level content is easy, it will prompt the user to pass the level and the final score based on the number of questions answered and the total time spent. If the difficulty level of the level content is medium or hard, and the user fails to pass all levels smoothly, it will return to the level failure, draw new questions to form new level content and a new round of challenge, until the user can pass all levels smoothly in one go. The module will then prompt the user to use the total time spent and the number of questions answered as the final score.

[0013] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described heuristic question bank and game generation methods.

[0014] This application utilizes a pre-built heuristic question bank architecture containing nested questions within a multi-layered tree structure. It generates level content through bidirectional parsing of both tabular and plain text question bank files, coupled with hierarchical recursive interactive logic. Ultimately, it simultaneously generates two types of answer result files. The tabular format reduces the manpower required for teachers to build the question bank, while the hierarchical tree structure enables personalized, targeted training for weak knowledge points, avoiding repetitive questions on already mastered concepts. Furthermore, both answer result files completely retain all user answer data, allowing teachers to directly perform learning statistics based on the tabular format. The question bank structure and question-and-answer content can be edited and interactively managed offline without relying on a database, making it suitable for practical training scenarios. Attached Figure Description

[0015] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the heuristic question bank and game generation method in one embodiment of this application; Figure 2 This is a logic diagram of the heuristic question bank and game generation method in one embodiment of this application; Figure 3 This is a schematic diagram of the in-game question extraction process in one embodiment of this application; Figure 4 This is a schematic diagram of the interaction process of level content in one embodiment of this application; Figure 5 This is a flowchart illustrating the process of selecting difficulty levels in a level, as shown in one embodiment of this application. Figure 6 This is a structural block diagram of a heuristic question bank and game generation device in one embodiment of this application; Figure 7 This is an internal structural diagram of a computer device in one embodiment of this application; Figure 8 This is a top view of the question-answering interaction device in one embodiment of this application; Figure 9 This is a side view of a question-answering interaction device in one embodiment of this application; Figure 10 This is an internal structural diagram of the question-answering interaction device in one embodiment of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0018] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0019] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] Based on the background technology, existing patents of the same type have multiple defects. General game education question bank patents only set up game modules with independent question banks, and only support level-based, random or sequential question generation. They do not support bidirectional conversion between tabular and plain text question bank files, lack top-level (main) / sub-level question linkage logic, and lack a sample ID extraction grouping question mechanism, making them unsuitable for question banks of STEM majors such as microcontrollers, analog electronics, and digital electronics. Patents related to layered question banks only add reading aids to the subject headings, lack standardized external file import and conversion logic, and lack multi-level recursive heuristic answering logic. Embedded microcontroller teaching and answering systems only store simple text question banks, lack custom hierarchical XML structures, and lack one-click conversion function for tabular question bank files. All existing solutions lack an integrated combination of large language model-generated question banks and fine-tuning of large language models based on structured question banks, offline flat question bank storage without database, multi-level recursive heuristic answering, multi-dimensional anti-cheating, PK battle, game animation linkage, and other integrated solutions.

[0021] The overall shortcomings of existing technologies can be summarized as follows: teachers have a large workload in building question banks; questions lack hierarchical relationships, making it impossible to achieve heuristic teaching; question bank storage relies on databases, resulting in high deployment costs; the question generation method is monotonous, making it easy for students to memorize options and pass the level by guessing; there is a lack of large language models to assist in question generation; there is no perfect anti-cheating mechanism; and the game interaction is monotonous.

[0022] This application provides a heuristic question bank and game generation method. Through games, it helps teenagers quickly grasp the correspondence between concepts and terminology. In particular, the visual understanding of terms and rules is key to students' knowledge acquisition. This skill development logic corresponds inversely to the training principles of artificial intelligence machine learning, abstracting terms and knowledge into vectors. Through repeated training with multi-layered recursive question-and-answer sessions, a stable vector association between terms and knowledge is established in the human brain, forming a conditioned reflex-like ability to skillfully apply knowledge, laying the foundation for classroom teaching content. This invention relies on tabular and plain text format question bank files to achieve database-free question bank storage, supporting lossless bidirectional conversion. It uses a large language model to automatically generate a hierarchical heuristic question bank and fine-tune the model in a closed loop. It employs a multi-layered tree structure nested questions to achieve personalized training by recursively displaying lower-level heuristic sub-questions when answers are incorrect. It is equipped with PK battles, leaderboards, animation linkage, and multiple anti-cheating mechanisms, adapting to teaching scenarios requiring conceptual understanding and proficiency, such as 51 / STM32 microcontrollers, analog and digital electronics, and English.

[0023] like Figure 1 , Figure 2 As shown, embodiments of this application provide a heuristic question bank and game generation method, including the following steps: Step S1: Set the structure of the heuristic question bank to include multiple game levels, each game level contains multiple knowledge points, each knowledge point corresponds to multiple sets of questions, and each set of questions contains multi-level nested questions that are related in a tree structure. Step S2: Edit a tabular question bank file containing heuristic question bank content. When the tabular question bank file is obtained, convert the tabular question bank file into a plain text question bank file, parse the plain text question bank file to obtain each set of questions corresponding to each knowledge point in each game level, and extract at least one question from each layer of the corresponding tree structure in each set of questions to form the level content. Step S3: When interaction with the level content is detected, the questions at the top level of the corresponding tree structure are displayed. The answers to each question are collected and it is determined whether the answer is correct. If the answer is correct, the knowledge point corresponding to the current question is skipped. If the answer is incorrect, the next level of questions is displayed. If all questions at the current level are answered correctly, the knowledge point corresponding to the current level is skipped. Otherwise, the next level of questions is displayed, and so on, until all questions at the bottom level are answered correctly, at which point the knowledge point corresponding to the current question is skipped. Step S4: When the interaction of the level content ends, obtain the difficulty level of the level content. If the difficulty level of the level content is easy, based on the number of questions answered by the current user and the total time consumed, indicate that the level has been successfully completed and the final score is displayed. If the difficulty level of the level content is medium or hard, and all levels cannot be passed smoothly, return to the level failure state, draw questions again to form level content and a new round of level challenge, until all levels can be successfully completed in one go, and indicate the total time consumed and the number of questions answered as the final score.

[0024] The final score is generated by recording the current user's answer data in the plain text format question bank file to form a plain text format answer result file, and then exporting the plain text format answer result file into a tabular format answer result file according to the tabular format question bank file.

[0025] If questions or knowledge points require multimedia files or JS scripts, simply save the filenames and open attributes of the referenced files as plain text in the corresponding question bank. Store these files and the XML file in the same path, and the software can then call and render the relevant content when executed.

[0026] During the construction of the heuristic question bank structure, the technical staff divides the game levels according to the course teaching logic. Each individual game level is bound to several teaching knowledge points, and multiple sets of independent questions are configured for each knowledge point. Each set of questions constructs a complete multi-level structure tree. The structure tree forms a fixed heuristic and guiding relationship from top to bottom. The questions at the top of the structure tree are the core assessment questions of the corresponding knowledge point. The questions at the bottom of the structure tree are the explanatory texts, step-by-step sub-questions, mnemonic devices, and other questions with heuristic and guiding functions that accompany the questions at the top level. This forms a complete question structure with multiple nested relationships. Taking the knowledge point of microcontroller number conversion as an example, the corresponding game level sets up the knowledge point of binary to hexadecimal conversion. This knowledge point is equipped with multiple sets of top-level questions. Each set of top-level questions is accompanied by multiple levels of reading-type sub-questions. The sub-questions contain heuristic content such as conversion calculation steps, numerical weight mnemonics, and number memory rules.

[0027] like Figure 2As shown, in the stage of processing the tabular question bank file and generating level content, the instructor uses Excel software to compile a standardized tabular question bank file. After the system obtains the tabular question bank file, it converts the tabular question bank file into a plain text question bank file in XML format according to the preset XML tag mapping rules. The system parses the global variable configuration line, question level line, and question content line stored in the plain text question bank file line by line, and extracts all the data corresponding to each game level, each knowledge point, each group of questions, and each level of questions. For each level in the structure tree, the system selects at least one question by combining the sample ID group identifier, random sampling strategy, or sequential sampling strategy. It integrates all the extracted question content to generate complete level content. The generated level content will completely inherit all global configuration parameters such as animation parameters, PK battle parameters, single question answering time limit, and level difficulty pre-configured in the tabular question bank file.

[0028] like Figure 2 As shown, during the interactive stage of the level content, after the system starts the complete level interaction process, it will first display the top-level question at the top of the tree structure, simultaneously collect the user's input answer and complete the correctness judgment. If the user's answer is determined to be correct, it will skip all levels of questions corresponding to that knowledge point and directly enter the answer process corresponding to the next set of knowledge points. If the user's answer is determined to be incorrect, it will jump to display the next level of questions corresponding to the current question. The system will continuously execute the level answer judgment logic in a loop. After all questions within the same level are answered correctly, the system determines that the knowledge content corresponding to that level has been mastered by the user and stops displaying all remaining questions at that level. If there are questions answered incorrectly within a level, the system will continue to retrieve questions from deeper levels and push them to the user until all questions at the bottom of the tree structure corresponding to the current knowledge point are answered correctly. Only then will the system determine that the user has fully mastered the knowledge point and skip that knowledge point.

[0029] like Figure 2 As shown, after all the interactive processes of the level content are completed, the system reads the standardized storage format of the plain text format question bank file, and synchronously collects complete answer data such as all user answer records, correct and incorrect mark of questions, answer time of a single question, total time to complete the level, and cumulative number of kills. Based on all the collected answer data, a plain text format answer result file is generated. The system reads the cell structure, column field layout, and global variable header format of the tabular format question bank file, and reverse-converts the plain text format answer result file to generate a tabular format answer result file that is completely consistent with the original tabular format question bank file. The instructor can directly open the file to view the class's answer status and perform data statistics.

[0030] In this embodiment, each group of questions contains multi-level nested questions arranged in a structured hierarchy, including: Each group of questions contains at least multiple top-level (main) questions, multiple sub-level questions corresponding to each top-level question, and multiple sub-sub-level questions corresponding to each sub-level question; The top-level questions are placed at the top level of the tree structure, the sub-level questions are placed at the second-to-top level of the tree structure, the second-to-second sub-level questions are placed at the second-to-last level of the tree structure, and so on. Questions belonging to the same knowledge point are linked together according to adjacent levels to form a group of questions corresponding to the same knowledge point.

[0031] Specifically, when interacting with the content of the level, a top-level question is displayed in each group of questions. The answers to each top-level question are collected and it is determined whether the answer is correct. If the answer is correct, the knowledge point corresponding to the current top-level question is skipped. If the answer is incorrect, the sub-questions of the current top-level question are displayed. If all the sub-questions of the current top-level question are answered correctly, the knowledge point corresponding to the current top-level question is skipped. Otherwise, the sub-sub-questions of the current sub-question are displayed, and so on, until all the questions at the bottom level are answered correctly. If all the questions at the bottom level are answered correctly, the knowledge point corresponding to the current question is skipped. Otherwise, it is recorded that the knowledge point corresponding to the current question has not been mastered.

[0032] Taking the knowledge points related to number system conversion in microcontrollers as an example, the top-level (main) questions are set as multiple-choice questions on binary to hexadecimal conversion, the sub-level (sub) questions are set as questions on reading and explaining the conversion steps, and the next sub-level (sub-sub) questions are set as short questions on memorizing numerical weight formulas. During the interactive content flow of the level, the system prioritizes loading all top-level (main) questions within each group of questions, collects the user's submitted answers for each question, and performs a correct / incorrect judgment. If any top-level (main) question is judged to be answered incorrectly, the system will immediately retrieve all sub-level (sub) questions bound to that top-level (main) question and push them to the user. If all sub-level (sub) questions are answered correctly, the system determines that the user has mastered the basic knowledge point corresponding to that top-level (main) question, and automatically jumps back to the top-level (main) question to re-examine the answer. If there are incorrect answers within a sub-level (sub) question, the system will continue to retrieve the next sub-level (sub-sub) questions and push them to the user. The system continues to retrieve deeper-level questions until all questions at the bottom of the knowledge point's tree structure are answered correctly. At this point, the system determines that the knowledge point is fully mastered. If there are still incorrect answers in the bottom-level questions, the system will mark the knowledge point as not mastered and simultaneously write it into a plain text format answer result file for storage.

[0033] The hierarchical recursive judgment logic in this embodiment can accurately locate students' weak knowledge points, and push basic auxiliary questions layer by layer to guide students to think and deduce independently, realizing personalized teaching by tracing the source of wrong questions, and avoiding the low learning efficiency caused by general repetitive practice. After all the interactive process is completed, the system automatically marks the knowledge points that have not been mastered. Instructors can directly select students' weak knowledge points from the tabular answer result file and adjust the focus of classroom teaching accordingly, effectively improving the accuracy of classroom teaching. The multi-level decomposition method can flatten the complex knowledge points in the fields of microcontrollers, analog and digital electronics into multiple progressive questions, lowering the entry threshold for students to understand the complex professional logic, and is highly adaptable to the actual use scenarios of practical training.

[0034] like Figure 3 As shown, Figure 3 The process for extracting questions in the game involves first determining whether the question group is out of order. After entering the question group, it is determined whether the top-level (main) questions are out of order. It is then determined whether there are subgroups at this level. If so, questions are extracted from the subgroups. Finally, the sub-level questions are determined and extracted.

[0035] In this embodiment, converting the tabular question bank file into a plain text question bank file includes: Obtain the tabular question bank file, and read the global variable row and question content row line by line according to the cell structure of the tabular question bank file to obtain the tabular structured data; Based on the structured data in the table, the content of each cell is converted into the corresponding XML node content according to the preset XML tag mapping rules to obtain a plain text format question bank file; The tabular question bank file and the plain text question bank file support bidirectional lossless conversion. When the plain text question bank file is obtained, the XML node content is reverse-parsed according to the XML tag mapping rules and written into the corresponding cell to obtain the tabular question bank file.

[0036] The purpose of bidirectional lossless conversion is that a tabular question bank is more suitable for manually edited questions, while the game relies on an XML file. During game operation, if errors are found in questions or answers, teachers can directly modify the XML file to correct the errors (clicking the "Correct" button in the game interface allows quick location of the incorrect question in the XML file). The modified XML file can then be converted back to a tabular question bank file.

[0037] The first three lines of the tabular question bank file are global variable configuration lines, which include the file version number ExcelVer, the order of appearance of top-level questions MainSeq, the level name EpisodeName, the level information EpisodeInfo, the level cover image EpisodePic, the visual layout Layout, the PK scheme settings, the animation scheme Animation, the interactive function SocialFeature, the time limit for multiple-choice questions Time4Single, the time limit for fill-in-the-blank questions Time4Texting, the time limit for reading comprehension questions Time4Reading, and the difficulty gradient Difficulty; While parsing the plain text format question bank file to obtain each set of questions corresponding to each knowledge point in each game level, the global variable configuration line is parsed simultaneously to obtain global configuration parameters, and the display method and interaction rules of the level are controlled according to the global configuration parameters.

[0038] After obtaining the tabular question bank file with the .xlsx extension, the system reads all cell contents line by line, distinguishing between the first three lines containing global variable configurations and the question content lines starting from the fifth line. It extracts all text, image names, and numerical parameters from each cell, integrating them into a complete tabular structured data. The system uses pre-defined fixed XML tag mapping rules, ensuring that each column in the tabular question bank file corresponds to a unique XML node tag. The system encapsulates all global parameters, question levels, question stems, standard answers, distractors, single-question time limits, and image storage paths within XML nodes, generating a plain text question bank file in XML format. During the reverse conversion process, the system reads all XML node content from the plain text question bank file, matches the pre-defined XML tag mapping rules, and fills the corresponding cells in the Excel file, completely preserving all parameters, question text, and image path information. This achieves bidirectional lossless conversion between the tabular and plain text question bank files.

[0039] The first three lines of the tabular question bank file store global variable configuration information. The file version number (ExcelVer) distinguishes different versions of the question bank for backward compatibility. The top-level question appearance order (MainSeq) configures whether the top-level (main) questions are displayed randomly or sequentially. The level name (EpisodeName), level information (EpisodeInfo), and level cover image (EpisodePic) configure the level name, level description, and level cover image respectively. The visual layout (Layout) configures the software interface layout style. The PK scheme settings control whether the battle mode is on or off. The animation scheme (Animation) configures the animations linked to shooting-type questions. The system includes three interactive features: SocialFeature controls the leaderboard and class interaction functions; Time4Single for multiple-choice questions, Time4Texting for fill-in-the-blank questions, and Time4Reading for reading comprehension questions; and Difficulty gradient sets the overall training difficulty level. The system extracts all global configuration parameters while parsing the plain text question bank file and automatically matches these parameters during the interactive operation phase of the level content. These parameters control all interactive rules, including the order of question appearance, countdown timer duration, animation effects, PK battle on / off status, and level difficulty coefficient.

[0040] A two-way lossless conversion mechanism between tabular and plain text question bank files allows instructors to edit question banks anytime using an Excel visual interface, eliminating the need to manually write XML code and significantly lowering the barrier to entry for professional question bank editing. All level configuration parameters are stored uniformly within the question bank file, eliminating the need for a separate backend database. Each question bank file has independent and complete operational capabilities, and offline local programs can directly load all level configuration information and run normally without a cloud server. Parameters such as level difficulty gradients, single-question answering time limits, animation effects, and PvP battles are centrally configured. Different difficulty levels can be quickly switched within the same question bank file without modifying the question content, adapting to the teaching needs of students at three different levels: beginners, intermediate learners, and advanced learners.

[0041] In this embodiment, the tabular question bank file containing heuristic question bank content includes: Obtain the subject name, chapter information, students' basic level, and teaching materials, and generate a question bank generation request based on the preset heuristic question bank prompt strategy; The question bank generation request is sent to the large language model, and the combination of question stem, answer, distractor and solution content returned by the large language model is obtained to obtain the original question bank data; Based on the original question bank data, a tabular question bank file is generated by organizing the data according to the structure template of the tabular file.

[0042] Leveraging a large language model, this system automatically generates a standardized, hierarchical, heuristic question bank, replacing the manual input of questions into Excel files by instructors. This significantly reduces the time spent on question bank creation. Dozens of questions within a single game level can be quickly generated by artificial intelligence, effectively reducing the workload for teachers. A dedicated prompting strategy constrains the output content of the large language model to perfectly match the overall architecture of the multi-level question bank. The AI-generated questions come with a top-level (main), sub-level (sub) hierarchical structure, extracted number (SampleID) grouping identifiers, and question type classification information, eliminating the need for instructors to further split the question hierarchy and ensuring a unified and standardized structure for all question bank files. The entire process is adaptable to the question generation needs of microcontrollers, analog and digital electronics, and English. The large language model can automatically generate professional question types such as number system conversion, C51 keywords, register definitions, English word root analysis, and verb tenses. Simultaneously, it provides mnemonic devices, calculation steps, and English word root evolution as sub-level reading questions, closely matching the actual needs of practical training scenarios.

[0043] In this embodiment, the method further includes: The tabular question bank file is manually verified. The manually verified tabular question bank file is converted into a plain text question bank file. The question hierarchy structure, question stem description, answer and distractor configuration in the plain text question bank file are extracted as fine-tuning training samples. The large language model is fine-tuned and trained based on the fine-tuning training samples to obtain a fine-tuned model that adapts to the requirements of heuristic question bank generation. When a command to generate question bank data is received, the fine-tuned model is invoked to generate target question bank data that matches the teacher's question-setting style and knowledge point coverage requirements, and a tabular question bank file is generated based on the target question bank data.

[0044] Instructors manually verify the tabular question bank files automatically generated by the large language model, correcting errors in question stems, standard answers, and hierarchical divisions. After manual verification, the tabular question bank files are converted to plain text format by the system. The system then traverses all XML nodes within the plain text question bank file, extracting hierarchical relationships, STEM terminology, distractor logic, reading comprehension explanation templates, and microcontroller-specific question structures. This extracted content is then uniformly packaged into fine-tuning training samples, continuously improving the accuracy of the professional question bank content generated by the large language model and reducing the workload for instructors in later corrections. The fine-tuned large language model is familiar with microcontroller, analog and digital electronics terminology, circuit calculation logic, C51 programming syntax, and word root composition, avoiding errors in the professional knowledge content output by general large language models, thus significantly improving the professional matching degree of the question bank.

[0045] In this embodiment, when interacting with the level content, the following is also included: Obtain the answering time limit parameter for the current question, and calculate the movement speed multiplier of the moving target in the animation area based on the answering time limit parameter; The movement trajectory of the target in the screen animation area is controlled according to the movement speed multiplier. When a timeout signal is detected, the target is controlled to move away from the hitable area. Collect answer result signals. When the answer result signal is correct, display a hit animation effect; when the answer result signal is incorrect, display a miss animation effect.

[0046] This system deeply integrates the time limit for answering individual questions, the correctness of answers, and the shooting game animations, creating a visual game feedback mechanism that enhances the fun of the quiz and attracts students to actively participate in the training. The animation movement speed dynamically adjusts based on the time limit for each question; shorter time limits for specialized multiple-choice questions correspond to targets with faster movement speeds, forcing students to master basic knowledge points and strengthening their quick reaction abilities. The animations accompanying the correct answers provide intuitive feedback on the results. Combined with the game's cumulative kill count statistics, this fully leverages the competitive spirit among students to improve their focus and enhance the overall learning atmosphere in the classroom.

[0047] In this embodiment, when interacting with the level content, the following is also included: If a command to enter battle mode is received, the list of participating students is obtained, and the same question sequence is randomly selected from the same heuristic question bank based on the list of participating students to obtain a unified battle question set for each participating student. Simultaneously launch the level interaction process for each participating student, and record the completion time and cumulative kill count for each participating student; When all participants have completed the level or the time limit has expired, the completion times of each participant are compared, and the participant who completes the level first is determined to be the winner.

[0048] The multiplayer PK battle mode creates a classroom-based competitive training environment, fully leveraging students' sense of honor to motivate them to repeatedly practice and increase the overall frequency of knowledge point training. All participating students use the same question sequence, eliminating unfairness caused by differences in random question selection from the question bank, and accurately comparing the mastery of different students' knowledge points. The entire battle process supports two operating modes: teachers initiating class PK tournaments and students freely matching opponents, adapting to both classroom group training and after-class independent practice.

[0049] In this embodiment, when interacting with the level content, the following is also included: Obtain the completion records of multiple students and extract the fastest completion time and the highest number of kills for each student; The participants are sorted in ascending order based on their fastest completion time to obtain a completion speed leaderboard. The students are sorted in descending order based on the highest number of kills to obtain a kill count leaderboard. After the level interaction ends, the ranking of the level completion speed and the ranking of the number of kills will be displayed.

[0050] After each student completes the individual challenge and the entire PK battle, the terminal interface simultaneously displays two leaderboards, intuitively showing the differences in training levels among students within the class. The leaderboards quantify students' mastery of knowledge points based on two dimensions: completion speed and accuracy. Instructors can easily identify students with weak knowledge and those with strong grasp of the material. The server permanently stores students' complete challenge history, allowing instructors to track students' progress. The public display of the leaderboards stimulates students' competitive spirit, motivating them to repeatedly attempt challenges to optimize completion time and increase the number of correct answers, thus reinforcing the practical effect of repeated practice.

[0051] In this embodiment, when interacting with the level content, the following is also included: When a student exits the level interaction page, the current answering progress data is obtained. The answering progress data includes the current question position, the answering results of questions already answered, and the status of uncompleted knowledge points. The answer progress data is uploaded and stored in the server database and associated with the corresponding student's account identifier; When a student is detected to have logged in again or reopened the check-in page, the system retrieves the answer progress data corresponding to the account identifier from the server database and restores the student to the answer state at the time of the last logout based on the answer progress data.

[0052] This implementation avoids the cheating loophole of students repeatedly restarting the game and refreshing levels to practice easy questions to avoid weak knowledge points. The system automatically saves the complete answer progress when the student exits the page, preventing students from skipping unmastered knowledge points by restarting the page. The entire process supports segmented challenge training; students can exit the challenge midway and continue training after class or the next day, adapting to fragmented learning scenarios. Answer progress data is linked to the student's account, allowing instructors to retrieve any student's answer records in the backend, providing a complete understanding of the distribution of incorrect answers and knowledge gaps.

[0053] In this embodiment, when interacting with the level content, the following is also included: When a student is detected to have exited the level interaction page, the IP address, MAC address or student ID stored in the browser's long-term cookie of the device used by the student to log in is obtained, and the IP address, MAC address and student ID are associated and recorded in the login device log of the corresponding student account. If it is detected that the student is using a browser instead of a client application and the MAC address cannot be obtained, the student's first login ID is saved as an alternative identification information for the MAC address using permanent cache data (cookie with an expiration time set to 1 year). Regularly compare the login device logs corresponding to different student accounts. When multiple different student accounts are detected with the same IP address or the same MAC address and student ID, mark the multiple different student accounts as suspected boosting accounts. A list of suspected account boosting services is generated and pushed to the teacher's backend for manual verification by the teacher based on the students' daily performance.

[0054] This system provides an automated detection solution for cheating in after-school homework scenarios, using device hardware identification to identify abnormal behavior such as multiple accounts sharing the same device, thus addressing the vulnerabilities of the single anti-cheating mechanism of resuming answering after breaks. The system automatically generates a list of suspected cheating accounts, reducing the workload of teachers manually verifying cheating behavior. The system also retains complete login device logs in the backend, allowing for complete tracing of cheating activities. The resuming answering after breaks and the device identification detection mechanism form a dual anti-cheating system, significantly improving the authenticity of game-based quiz results and ensuring that training scores objectively reflect students' true knowledge mastery.

[0055] In this embodiment, when interacting with the level content, the following is also included: Obtain the question and answer data from the plain text format question bank file, encrypt the question and answer data, and obtain the encrypted question bank data; The encrypted question bank data is transmitted to the terminal for parsing and display, and screenshotting and copying of the terminal page are prohibited; Obtain the answering time limit parameter for the current question, and control the answering time for a single question according to the answering time limit parameter. If the time limit is exceeded, the current question is judged as a failure.

[0056] The questions are set with a QuestionType attribute, which includes multiple choice, fill-in-the-blank, and reading comprehension questions. Different question types correspond to different time limits and interaction methods. Multiple choice questions display the question stem and multiple options for students to choose from. Fill-in-the-blank questions display the question stem for students to enter their answers. Reading comprehension questions display the explanation of knowledge points for students to read before proceeding to the next stage.

[0057] Encrypting the question bank data and prohibiting page copying and screenshotting prevents plug-ins from capturing plaintext standard answers, avoiding online cheating by searching for answers. Combined with a timed answering mechanism, this further reduces the time students have to search for and answer questions, ensuring they can complete practice questions independently. The system standardizes the interaction logic for three question types, automatically matching the corresponding answering time limit, eliminating the need for instructors to configure countdown times for each question, thus simplifying the overall workload of question bank configuration. Reading comprehension sub-questions are only used for explaining knowledge points and do not have interactive answering functions, strictly distinguishing between explanatory content and core assessment questions, making the tiered teaching logic clear and easy to understand.

[0058] Please see Figure 3 In this embodiment, the step of extracting at least one question from each layer of the corresponding tree structure in each group of questions to form level content further includes: Retrieve multiple parallel questions at the same level within each group of questions, and retrieve the group identifier and the list of questions at each level of the corresponding tree structure for each group of questions; For questions at the same level under the same group identifier, at least one question is selected as the question to be displayed in the current level, according to a random selection strategy or a sequential selection strategy.

[0059] Among them, the random question selection mechanism addresses the loophole that students can pass the level by rote memorization of question options. Multiple parallel questions for the same knowledge point are randomly displayed, and students can only answer the questions correctly if they fully grasp the calculation principle, thus avoiding the need to rely on memorizing options to pass the level.

[0060] The heuristic question bank employs a flat question bank storage method, independent of database storage, directly listing all question content through tabular and plain text files. Multiple sets of questions corresponding to the same knowledge point are recorded in tabular files in parallel rows, distinguished by hierarchy and group identifier fields. All questions are stored in a flat, table-row format, allowing instructors to intuitively modify, add, and delete question content in batches, significantly simplifying question bank maintenance compared to the complex editing interfaces of database backends.

[0061] like Figure 5 As shown, Figure 5This flowchart illustrates the process of selecting difficulty levels for each level, outlining the logic for setting the game's difficulty. For the same question, setting the difficulty level to Easy, Middle, or Hard (20%~50%) will change the game's overall difficulty without altering other question banks. The game difficulty can be selected by students before entering the game or set by teachers in the backend.

[0062] In the aforementioned heuristic question bank and game generation method, a heuristic question bank architecture containing nested questions in a multi-layered tree structure is pre-built. Level content is generated through bidirectional parsing of tabular and plain text question bank files. Combined with hierarchical recursive answer interaction logic, two types of answer result files are generated simultaneously. The tabular question bank file reduces the manpower required for teachers to build the question bank, while the hierarchical answer structure enables personalized heuristic training for weak knowledge points, avoiding the problem of repetitive questions on already mastered knowledge points. Simultaneously, both answer result files can completely retain all user answer data, allowing teachers to directly perform learning statistics based on the tabular answer result file. The question bank structure and question-and-answer content can be edited and interactively played offline without relying on a database, making it suitable for practical training scenarios.

[0063] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0064] In one embodiment, such as Figure 6 As shown, a heuristic question bank and game generation device are provided, including: a question bank structure setting module, a level content generation module, an interactive question-answering module, and a question result feedback module.

[0065] The question bank structure setting module is used to set the structure of the heuristic question bank, which includes multiple game levels, each game level contains multiple knowledge points, each knowledge point corresponds to multiple sets of questions, and each set of questions contains multi-level nested questions that are related in a tree structure. The level content generation module is used to edit a tabular question bank file with heuristic question bank content. When the tabular question bank file is obtained, it is converted into a plain text question bank file. The plain text question bank file is parsed to obtain each set of questions corresponding to each knowledge point in each game level. At least one question is extracted from each layer of the corresponding tree structure in each set of questions to form the level content. The interactive question-answering module is used to display the questions at the top level of the corresponding tree structure when interaction with the level content is detected. It collects the answers to each question and determines whether the answers are correct. If the answers are correct, the knowledge points corresponding to the current question are skipped. If the answers are incorrect, the next level of questions is displayed. If all questions at the current level are answered correctly, the knowledge points corresponding to the current level are skipped. Otherwise, the next level of questions is displayed, and so on, until all questions at the bottom level are answered correctly, at which point the knowledge points corresponding to the current question are skipped. The answer result feedback module is used to obtain the difficulty level of the level content when the interaction of the level content ends. If the difficulty level of the level content is easy, it will prompt the user to pass the level and the final score based on the number of questions answered and the total time consumed. If the difficulty level of the level content is medium or hard, and the user fails to pass all levels smoothly, it will return to the level failure state, draw questions again to form level content and a new round of challenge, until the user can pass all levels smoothly in one go. The module will then prompt the user to use the total time consumed and the number of questions answered as the final score.

[0066] In this embodiment, each group of questions contains multi-level nested questions arranged in a structured hierarchy, including: Each group of questions contains at least multiple top-level (main) questions, multiple sub-level questions corresponding to each top-level question, and multiple sub-sub-level questions corresponding to each sub-level question; The top-level questions are placed at the top level of the tree structure, the sub-level questions are placed at the second-to-top level of the tree structure, the second-to-second sub-level questions are placed at the second-to-last level of the tree structure, and so on. Questions belonging to the same knowledge point are linked together according to adjacent levels to form a group of questions corresponding to the same knowledge point.

[0067] Specifically, when interacting with the content of the level, a top-level question is displayed in each group of questions. The answers to each top-level question are collected and it is determined whether the answer is correct. If the answer is correct, the knowledge point corresponding to the current top-level question is skipped. If the answer is incorrect, the sub-questions of the current top-level question are displayed. If all the sub-questions of the current top-level question are answered correctly, the knowledge point corresponding to the current top-level question is skipped. Otherwise, the sub-sub-questions of the current sub-question are displayed, and so on, until all the questions at the bottom level are answered correctly. If all the questions at the bottom level are answered correctly, the knowledge point corresponding to the current question is skipped. Otherwise, it is recorded that the knowledge point corresponding to the current question has not been mastered.

[0068] In this embodiment, converting the tabular question bank file into a plain text question bank file includes: Obtain the tabular question bank file, and read the global variable row and question content row line by line according to the cell structure of the tabular question bank file to obtain the tabular structured data; Based on the structured data in the table, the content of each cell is converted into the corresponding XML node content according to the preset XML tag mapping rules to obtain a plain text format question bank file; The tabular question bank file and the plain text question bank file support bidirectional lossless conversion. When the plain text question bank file is obtained, the XML node content is reverse-parsed according to the XML tag mapping rules and written into the corresponding cell to obtain the tabular question bank file.

[0069] The first three lines of the tabular question bank file are global variable configuration lines, which include the file version number ExcelVer, the order of appearance of top-level questions MainSeq, the level name EpisodeName, the level information EpisodeInfo, the level cover image EpisodePic, the visual layout Layout, the PK scheme settings, the animation scheme Animation, the interactive function SocialFeature, the time limit for multiple-choice questions Time4Single, the time limit for fill-in-the-blank questions Time4Texting, the time limit for reading comprehension questions Time4Reading, and the difficulty gradient Difficulty; While parsing the plain text format question bank file to obtain each set of questions corresponding to each knowledge point in each game level, the global variable configuration line is parsed simultaneously to obtain global configuration parameters, and the display method and interaction rules of the level are controlled according to the global configuration parameters.

[0070] In this embodiment, the tabular question bank file containing heuristic question bank content includes: Obtain the subject name, chapter information, students' basic level, and teaching materials, and generate a question bank generation request based on the preset heuristic question bank prompt strategy; The question bank generation request is sent to the large language model, and the combination of question stem, answer, distractor and solution content returned by the large language model is obtained to obtain the original question bank data; Based on the original question bank data, a tabular question bank file is generated by organizing the data according to the structure template of the tabular file.

[0071] In this embodiment, as Figure 6 As shown, the heuristic question bank and game generation device further includes a question bank optimization module, which is used for: The tabular question bank file is manually verified. The manually verified tabular question bank file is converted into a plain text question bank file. The question hierarchy structure, question stem description, answer and distractor configuration in the plain text question bank file are extracted as fine-tuning training samples. The large language model is fine-tuned and trained based on the fine-tuning training samples to obtain a fine-tuned model that adapts to the requirements of heuristic question bank generation. When a command to generate question bank data is received, the fine-tuned model is invoked to generate target question bank data that matches the teacher's question-setting style and knowledge point coverage requirements, and a tabular question bank file is generated based on the target question bank data.

[0072] In this embodiment, when interacting with the level content, the interactive question-answering module 3 is further used to: Obtain the answering time limit parameter for the current question, and calculate the movement speed multiplier of the moving target in the animation area based on the answering time limit parameter; The movement trajectory of the target in the screen animation area is controlled according to the movement speed multiplier. When a timeout signal is detected, the target is controlled to move away from the hitable area. Collect answer result signals. When the answer result signal is correct, display a hit animation effect; when the answer result signal is incorrect, display a miss animation effect.

[0073] In this embodiment, when interacting with the level content, the interactive question-answering module 3 is further used to: If a command to enter battle mode is received, the list of participating students is obtained, and the same question sequence is randomly selected from the same heuristic question bank based on the list of participating students to obtain a unified battle question set for each participating student. Simultaneously launch the level interaction process for each participating student, and record the completion time and cumulative kill count for each participating student; When all participants have completed the level or the time limit has expired, the completion times of each participant are compared, and the participant who completes the level first is determined to be the winner.

[0074] In this embodiment, when interacting with the level content, the interactive question-answering module is further used to: Obtain the completion records of multiple students and extract the fastest completion time and the highest number of kills for each student; The participants are sorted in ascending order based on their fastest completion time to obtain a completion speed leaderboard. The students are sorted in descending order based on the highest number of kills to obtain a kill count leaderboard. After the level interaction ends, the ranking of the level completion speed and the ranking of the number of kills will be displayed.

[0075] In this embodiment, when interacting with the level content, the following is also included: When a student exits the level interaction page, the current answering progress data is obtained. The answering progress data includes the current question position, the answering results of questions already answered, and the status of uncompleted knowledge points. The answer progress data is uploaded and stored in the server database and associated with the corresponding student's account identifier; When a student is detected to have logged in again or reopened the check-in page, the system retrieves the answer progress data corresponding to the account identifier from the server database and restores the student to the answer state at the time of the last logout based on the answer progress data.

[0076] In this embodiment, when interacting with the level content, the interactive question-answering module 3 is further used to: When a student is detected to have exited the level interaction page, the IP address, MAC address or student ID stored in the browser's long-term cookie of the device used by the student to log in is obtained, and the IP address, MAC address and student ID are associated and recorded in the login device log of the corresponding student account. If it is detected that the student is using a browser instead of a client application and the MAC address cannot be obtained, the student's first login ID is saved as an alternative identification information for the MAC address using permanent cache data (cookie with an expiration time set to 1 year). Regularly compare the login device logs corresponding to different student accounts. When multiple different student accounts are detected with the same IP address or the same MAC address and student ID, mark the multiple different student accounts as suspected boosting accounts. A list of suspected account boosting services is generated and pushed to the teacher's backend for manual verification by the teacher based on the students' daily performance.

[0077] In this embodiment, when interacting with the level content, the interactive question-answering module 3 is further used to: Obtain the question and answer data from the plain text format question bank file, encrypt the question and answer data, and obtain the encrypted question bank data; The encrypted question bank data is transmitted to the terminal for parsing and display, and screenshotting and copying of the terminal page are prohibited; Obtain the answering time limit parameter for the current question, and control the answering time for a single question according to the answering time limit parameter. If the time limit is exceeded, the current question is judged as a failure.

[0078] The questions are set with a QuestionType attribute, which includes multiple choice, fill-in-the-blank, and reading comprehension questions. Different question types correspond to different time limits and interaction methods. Multiple choice questions display the question stem and multiple options for students to choose from. Fill-in-the-blank questions display the question stem for students to enter their answers. Reading comprehension questions display the explanation of knowledge points for students to read before proceeding to the next stage.

[0079] In this embodiment, the step of extracting at least one question from each layer of the corresponding tree structure in each group of questions to form the level content further includes: Retrieve multiple parallel questions at the same level within each group of questions, and retrieve the group identifier and the list of questions at each level of the corresponding tree structure for each group of questions; For questions at the same level under the same group identifier, at least one question is selected as the question to be displayed in the current level, according to a random selection strategy or a sequential selection strategy.

[0080] The heuristic question bank adopts a flat question bank storage method, which does not rely on database storage. All question content is directly listed through tabular files and plain text files. Multiple sets of questions corresponding to the same knowledge point are recorded in the tabular file in the form of parallel rows. The level and group of the question are distinguished by the level field and the group identifier field.

[0081] The aforementioned heuristic question bank and game generation device, by pre-building an overall architecture of a heuristic question bank containing nested questions in a multi-layered tree structure, generates level content through bidirectional parsing of tabular and plain text question bank files. Combined with hierarchical recursive answering interaction logic, it ultimately generates two types of answer result files simultaneously. This approach reduces the manpower required for teachers to build the question bank by using tabular question bank files, and enables personalized heuristic training for weak knowledge points through hierarchical answering based on the tree structure, avoiding the problem of repeatedly generating questions for already mastered knowledge points. The question bank structure and question-and-answer content can be edited and interactively challenged without relying on a database, making it suitable for practical training and teaching scenarios.

[0082] For a description of the features in the embodiments corresponding to the heuristic question bank and game generation device, please refer to the relevant descriptions in the embodiments corresponding to the heuristic question bank and game generation method, which will not be repeated here.

[0083] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above embodiments of the heuristic question bank and game generation method.

[0084] In one embodiment, the electronic device may be a server, and its internal structure diagram may be as follows: Figure 7As shown, this electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores heuristic question banks and game generation data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a heuristic question bank and game generation method.

[0085] like Figure 8 , Figure 9 , Figure 10 As shown, the electronic device also includes a question-and-answer interaction device, which is connected to the processor through the network interface or system bus. The question-and-answer interaction device includes a game controller 1, a microphone 2, a press-to-talk (PTT) button 3, a microphone indicator 4, a confirmation button 5, and a selection button 6. The microphone 2 is fixed in the middle of the game controller 1. A set of selection buttons 6 is provided on both sides of the front of the game controller 1. Each set of selection buttons 6 includes four directional control keys: up, down, left, and right. A microphone indicator 4 is provided between the two sets of selection buttons 6. The microphone 2 is located on the top side of the microphone indicator 4. The press-to-talk (PTT) button 3 and the confirmation button 5 are provided on the top side of the game controller 1. The microphone 2 is located between the press-to-talk (PTT) button 3 and the confirmation button 5. The gun microphone 2 includes a gun microphone dustproof mesh 20, non-equally spaced slots 21 on the inner wall of the interference tube, a microphone support sponge 22, and a condenser microphone 23. The extension direction of the gun microphone 2 is perpendicular to the front of the game controller 1. The extension direction of the gun microphone 2 is hollow and has a sound cavity. The sound cavity opening is equipped with a gun microphone dustproof mesh 20. The inner side wall of the sound cavity is equipped with non-equally spaced slots 21 on the inner wall of the interference tube. The bottom of the sound cavity is equipped with a condenser microphone 23 and a microphone support sponge 22.

[0086] When in use, after pressing the PTT (Public Speaking) button 3, the gamepad's microphone circuit is enabled, and the microphone indicator 4 lights up. The student can then read the reading comprehension questions aloud into the microphone or use the computer's voice input function to enter text. Pressing the PTT button 3 and the microphone indicator 4 ensures student privacy by pausing the vibration motor inside the gamepad 1 to prevent interference. The enhanced directivity of the microphone 2 further prevents voice interference between students. Compared to a mouse, the gamepad's buttons or gyroscope allow for quick selection of answers, and the confirmation button allows for quick confirmation of answers and switching to the next question.

[0087] In addition, the game controller 1 can have a built-in identity ID. When playing online or pressing the confirmation button, the identity ID will be sent to the server. Together with the user ID stored in the browser's permanent cookie, it will serve as an alternative identity recognition scheme for the MAC address.

[0088] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above embodiments of the heuristic question bank and game generation method at runtime.

[0089] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0090] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the heuristic question bank and game generation method.

[0091] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above embodiments of the heuristic question bank and game generation method.

[0092] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0093] The foregoing has provided a detailed description of the heuristic question bank, game generation method, apparatus, and electronic device provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only intended to aid in understanding the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A heuristic question bank and game generation method, characterized in that, include: The structure of the heuristic question bank includes multiple game levels, each game level contains multiple knowledge points, each knowledge point corresponds to multiple sets of questions, and each set of questions contains multi-level nested questions that are related in a tree structure. Edit a tabular question bank file containing heuristic question bank content. When the tabular question bank file is obtained, convert the tabular question bank file into a plain text question bank file. Parse the plain text question bank file to obtain each set of questions corresponding to each knowledge point in each game level, and extract at least one question from each layer of the corresponding tree structure in each set of questions to form the level content. When interaction with the level content is detected, the questions at the top level of the corresponding tree structure are displayed. The answers to each question are collected and it is determined whether the answer is correct. If the answer is correct, the knowledge point corresponding to the current question is skipped. If the answer is incorrect, the next level of questions is displayed. If all questions at the current level are answered correctly, the knowledge point corresponding to the current level is skipped. Otherwise, the next level of questions is displayed, and so on, until all questions at the bottom level are answered correctly, at which point the knowledge point corresponding to the current question is skipped. When the interaction of the level ends, the difficulty level of the level is obtained. If the difficulty level of the level is easy, the user is prompted with a success message and final score based on the number of questions answered and the total time consumed. If the difficulty level of the level is medium or hard, and the user fails to pass all levels smoothly, the user is returned to failure, and questions are drawn again to form a new level and a new round of challenges until all levels can be completed smoothly in one go. The total time consumed and the number of questions answered are then displayed as the final score.

2. The heuristic question bank and game generation method according to claim 1, characterized in that, Each group of questions contains multiple nested questions arranged in a tree structure, including: Each group of questions contains at least multiple top-level questions, multiple sub-level questions corresponding to each top-level question, and multiple sub-sub-level questions corresponding to each sub-level question; The top-level questions are placed at the top level of the tree structure, the sub-level questions are placed at the second-to-top level of the tree structure, the second-to-second sub-level questions are placed at the second-to-last level of the tree structure, and so on. Questions belonging to the same knowledge point are linked together according to adjacent levels to form a group of questions corresponding to the same knowledge point.

3. The heuristic question bank and game generation method according to claim 1, characterized in that, The step of converting the tabular question bank file into a plain text question bank file includes: Obtain the tabular question bank file, and read the global variable row and question content row line by line according to the cell structure of the tabular question bank file to obtain the tabular structured data; Based on the structured data in the table, the content of each cell is converted into the corresponding XML node content according to the preset XML tag mapping rules to obtain a plain text format question bank file; The tabular question bank file and the plain text question bank file support bidirectional lossless conversion. When the plain text question bank file is obtained, the XML node content is reverse-parsed according to the XML tag mapping rules and written into the corresponding cell to obtain the tabular question bank file.

4. The heuristic question bank and game generation method according to claim 1, characterized in that, The tabular question bank file containing heuristic question bank content includes: Obtain the subject name, chapter information, students' basic level, and teaching materials, and generate a question bank generation request based on the preset heuristic question bank prompt strategy; The question bank generation request is sent to the large language model, and the combination of question stem, answer, distractor and solution content returned by the large language model is obtained to obtain the original question bank data; Based on the original question bank data, a tabular question bank file is generated by organizing the data according to the tabular file structure template.

5. The heuristic question bank and game generation method according to claim 4, characterized in that, The method further includes: The tabular question bank file is manually verified. The manually verified tabular question bank file is converted into a plain text question bank file. The question hierarchy structure, question stem description, answer and distractor configuration in the plain text question bank file are extracted as fine-tuning training samples. The large language model is fine-tuned and trained based on the fine-tuning training samples to obtain a fine-tuned model that adapts to the requirements of heuristic question bank generation. When a command to generate question bank data is received, the fine-tuned model is invoked to generate target question bank data that matches the teacher's question-setting style and knowledge point coverage requirements, and a tabular question bank file is generated based on the target question bank data.

6. The heuristic question bank and game generation method according to claim 1, characterized in that, Interacting with the level content also includes: When a student exits the level interaction page, the current answering progress data is obtained. The answering progress data includes the current question position, the answering results of the questions, and the status of the unfinished knowledge points. The answer progress data is uploaded and stored in the server database and associated with the corresponding student's account identifier; When a student is detected to have logged in again or reopened the check-in page, the system retrieves the answer progress data corresponding to the account identifier from the server database and restores the student to the answer state at the time of the last logout based on the answer progress data.

7. The heuristic question bank and game generation method according to claim 1, characterized in that, When interacting with the level content, it also includes: When a student is detected to have exited the level interaction page, the IP address, MAC address or student ID stored in the browser's long-term cookie of the device used by the student to log in is obtained, and the IP address, MAC address and student ID are associated and recorded in the login device log of the corresponding student account. If it is detected that the student is using a browser instead of a client application and the MAC address cannot be obtained, the student's first login ID is saved as an alternative identification information for the MAC address using permanent cache data (cookie with an expiration time set to 1 year). Regularly compare the login device logs corresponding to different student accounts. When multiple different student accounts are detected with the same IP address or the same MAC address and student ID, mark the multiple different student accounts as suspected boosting accounts. A list of suspected account boosting services is generated and pushed to the teacher's backend for manual verification by the teacher based on the students' daily performance.

8. The heuristic question bank and game generation method according to claim 1, characterized in that, The step of extracting at least one question from each layer of the corresponding tree structure in each group of questions to form the level content also includes: Get the multiple parallel questions at the same level in each group of questions, and get the group identifier of each group of questions and the list of questions at each level of the corresponding tree structure. For questions at the same level under the same group identifier, at least one question is selected as the question to be displayed in the current level, according to a random selection strategy or a sequential selection strategy.

9. A heuristic question bank and game generation device, characterized in that, include: The question bank structure setting module is used to set the structure of the heuristic question bank, which includes multiple game levels, each game level contains multiple knowledge points, each knowledge point corresponds to multiple sets of questions, and each set of questions contains multi-level nested questions that are related in a tree structure. The level content generation module is used to edit a tabular question bank file with heuristic question bank content. When the tabular question bank file is obtained, it is converted into a plain text question bank file. The plain text question bank file is parsed to obtain each set of questions corresponding to each knowledge point in each game level. At least one question is extracted from each layer of the corresponding tree structure in each set of questions to form the level content. The interactive question-answering module is used to display the questions at the top level of the corresponding tree structure when interaction with the level content is detected. It collects the answers to each question and determines whether the answer is correct. If the answer is correct, it skips the knowledge point corresponding to the current question. If the answer is incorrect, it displays the next level of questions. If all questions at the current level are answered correctly, it skips the knowledge point corresponding to the current level. Otherwise, it displays the next level of questions, and so on, until all questions at the bottom level are answered correctly, at which point it skips the knowledge point corresponding to the current question. The answer result feedback module is used to obtain the difficulty level of the level content when the interaction of the level content ends. If the difficulty level of the level content is easy, it will prompt the user to pass the level and the final score based on the number of questions answered and the total time spent. If the difficulty level of the level content is medium or hard, and the user fails to pass all levels smoothly, it will return to the level failure, draw new questions to form new level content and a new round of challenge, until the user can pass all levels smoothly in one go. The module will then prompt the user to use the total time spent and the number of questions answered as the final score.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the heuristic question bank and game generation method as described in any one of claims 1 to 8 when executing the computer program.

Images