An AI learning device-based individualized test paper generation and printing control method

Abstract

The application relates to the technical field of AI learning devices and education data processing, and provides a personalized test paper generation and printing control method based on an AI learning device. The method collects student homework images through an image acquisition module of the learning device, performs character recognition and standard answer comparison on the homework images, recognizes and marks all wrong question contents in the homework, extracts knowledge points corresponding to each wrong question, analyzes error causes of each wrong question, generates a learning condition label corresponding to the wrong question, calls a large model to generate multiple test questions associated with the wrong question knowledge points, formats the multiple test questions into an electronic test paper according to a preset format, acquires a connection type and format specification of a connectable printing device, adjusts an output format of the electronic test paper according to the connection type and format specification of the printing device, and sends a printing control instruction to the printing device to output a paper test paper. The whole process of wrong question consolidation is automatically executed.

Description

Technical Field

[0001] This application relates to the field of AI learning devices and educational data processing technology, and in particular to a method for personalized test paper generation and printing control based on AI learning devices. Background Technology

[0002] Existing intelligent learning devices can complete homework correction and error identification through image acquisition, but the error correction and knowledge point consolidation stages are disconnected. The generated consolidation questions have a low degree of matching with the test points and reasons for errors, failing to achieve accurate training for applying knowledge to new situations. At the same time, existing learning devices cannot automatically adapt the test paper output format according to the connection type of the printing device. The printing operation requires manual adjustment or switching of terminals, failing to form a complete processing link from error collection to paper test paper output. The operation of error consolidation is cumbersome and inefficient. Summary of the Invention

[0003] This application provides a personalized test paper generation and printing control method based on AI learning devices, aiming to solve the problem that existing intelligent learning devices can complete homework correction and error identification through image acquisition, but the error correction and knowledge point consolidation are disconnected, and the generated consolidation test questions have a low matching degree with the test points and error causes of the errors, making it impossible to achieve accurate training for applying knowledge to new situations.

[0004] In a first aspect, embodiments of this application provide a method for personalized test paper generation and printing control based on an AI learning device, the method comprising: The learning device's image acquisition module captures images of student assignments. Text recognition is performed on these images, comparing them with standard answers to identify and mark all incorrect answers. This process includes: activating the image acquisition module to capture a complete image of the assignment page; performing tilt correction and noise removal on the complete image; performing text recognition on the processed image to obtain the assignment answers; comparing the assignment answers with pre-stored standard answers question by question to identify incorrect answers and marking the corresponding areas in the assignment image; and then using a large model to analyze all incorrect answers, extracting the knowledge points corresponding to each incorrect answer, analyzing the reasons for each incorrect answer, and generating learning tags corresponding to the incorrect answers. Based on the extracted knowledge points and the analyzed reasons for errors, a large model is invoked to generate multiple test questions associated with the knowledge points of the incorrect questions. This includes: inputting the extracted knowledge points and the analyzed reasons for errors into the large model; matching test materials consistent with the test points of the incorrect questions through the large model; generating multiple test questions according to different examination angles and question types; controlling the examination scope of the test questions to cover all knowledge points corresponding to the incorrect questions; and formatting the multiple test questions into an electronic test paper according to a preset format. Obtain the connection type and format specifications of the connectable printing devices, adjust the output format of the electronic exam paper according to the connection type and format specifications of the printing devices, and send print control commands to the printing devices to output the paper exam paper.

[0005] In some embodiments, the step of calling the large model to analyze all incorrect questions, extracting the knowledge points corresponding to each incorrect question, and analyzing the reasons for each incorrect question includes: extracting the question stem and answer content of each incorrect question and inputting them into the large model; matching the knowledge system database through the large model to obtain the knowledge point entries corresponding to each incorrect question; and comparing the differences between the answer content and the correct answer through the large model to determine the type of error reason corresponding to each incorrect question.

[0006] In some embodiments, generating learning labels corresponding to incorrect questions includes: counting the number of incorrect questions for the same knowledge point, combining the error reason type of each incorrect question to generate a mastery level label and error reason distribution label for the corresponding knowledge point, and summarizing all the labels to form a learning label set for the corresponding student.

[0007] In some embodiments, the step of formatting multiple test questions into an electronic test paper according to a preset layout includes: classifying and sorting all test questions according to question type, setting a blank answer area for each question, adjusting the spacing between questions and font size to match the size parameters of standard paper, and generating an electronic test paper file that conforms to printing specifications.

[0008] In some embodiments, obtaining the connection type and format specifications of the connectable printing devices, and adjusting the output format of the electronic exam paper according to the connection type and format specifications of the printing devices, includes: scanning all currently connectable printing devices, reading the connection method and supported paper size and resolution parameters of each printing device, adjusting the page size, resolution and layout of the electronic exam paper according to the parameters of the selected printing device, and generating an output file adapted to the selected printing device.

[0009] In some embodiments, the method further includes: acquiring images of paper test papers completed by students, identifying and correcting the test paper images to obtain new incorrect questions, comparing the overlap of knowledge points between the new incorrect questions and the original incorrect questions, and updating the corresponding student's learning tag set.

[0010] In some embodiments, the method further includes: when no directly connectable printing device is detected, sending an electronic test paper file and a print trigger command to a pre-bound mobile terminal, connecting to the corresponding printing device through the mobile terminal, and performing the test paper printing operation.

[0011] In some embodiments, the method further includes: after identifying and marking all incorrect questions, calling a large model to generate explanation content for the knowledge points corresponding to each incorrect question. The explanation content sequentially includes the explanation of the core concepts of the knowledge point, the location of the error cause of the corresponding incorrect question, the breakdown of the standard problem-solving ideas and steps, and the summary of general problem-solving techniques for similar question types; displaying the explanation content of each incorrect question through the display module of the learning device, and displaying the mastery level feedback option after the explanation of each incorrect question is completed, receiving the mastery level feedback submitted by the student, which is divided into levels of complete understanding, basic understanding, and still having questions; combining the student's mastery level feedback with the generated learning labels to jointly adjust the number and difficulty of the subsequently generated test questions: for knowledge points where the feedback is "still having questions", increasing the number of corresponding test questions generated, while reducing the difficulty of the questions and increasing the proportion of basic questions; for knowledge points where the feedback is "completely understood", reducing the number of corresponding test questions generated, while increasing the difficulty of the questions and increasing the proportion of extended questions.

[0012] This application constructs a complete processing chain, from homework image acquisition, error identification, error knowledge point analysis, personalized test paper generation to adaptive printing output, to achieve fully automated execution of the error consolidation process, significantly improving the efficiency of error consolidation. Based on the knowledge points and reasons for the errors, it generates associated test questions, ensuring a high degree of matching between the consolidation questions and the test points of the errors, achieving precise learning and application training effects, and improving the targeting of error consolidation. It can automatically adjust the test paper output format according to the connection type and format specifications of the printing device, adapting to different printing scenarios, simplifying printing operations, and improving the compatibility between learning devices and various printing devices.

[0013] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic flowchart illustrating the steps of a personalized test paper generation and printing control method based on an AI learning device, provided in one embodiment of this application. Figure 2 This is a schematic diagram illustrating the principle of a personalized test paper generation and printing control method based on an AI learning device, provided in one embodiment of this application. Figure 3This is a schematic block diagram of a personalized test paper generation and printing control system based on an AI learning device, provided in one embodiment of this application. Figure 4 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0016] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit 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, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0019] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0020] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0021] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0022] Existing intelligent learning devices can complete homework correction and error identification through image acquisition, but the error correction and knowledge point consolidation stages are disconnected. The generated consolidation questions have a low degree of matching with the test points and reasons for errors, failing to achieve accurate training for applying knowledge to new situations. At the same time, existing learning devices cannot automatically adapt the test paper output format according to the connection type of the printing device. The printing operation requires manual adjustment or switching of terminals, failing to form a complete processing link from error collection to paper test paper output. The operation of error consolidation is cumbersome and inefficient.

[0023] Please refer to Figure 1 and Figure 2 This embodiment provides a personalized test paper generation and printing control method based on an AI learning device. It is applied to a computer device with data processing capabilities, which can be integrated into the AI ​​learning device or deployed on a separate server device. The computer device interacts with the AI ​​learning device and printing device via wired or wireless communication. The AI ​​learning device has a built-in image acquisition module and display output module, providing functions such as homework correction, error correction reinforcement, and test paper printing for student learning scenarios. All user data and image information involved in this method are collected and processed with the user's authorization and consent, and the entire process complies with relevant data security and personal information protection regulations.

[0024] Combination Figure 1 and Figure 2 As shown, this method consists of three core stages executed sequentially: image acquisition of assignments and identification of incorrect questions and generation of learning labels; personalized test paper assembly and layout; and printing device adaptation and printing control. Each stage is connected sequentially to form a complete chain for consolidating and processing incorrect questions.

[0025] The provided method for personalized test paper generation and printing control based on AI learning devices includes steps S101 to S103. Details are as follows: Step S101. Acquire student homework images through the image acquisition module of the learning device, perform text recognition on the homework images and compare them with the standard answers, identify and mark all wrong questions in the homework, including: start the image acquisition module to acquire the complete image of the homework page, perform tilt correction and noise removal on the complete image, perform text recognition on the processed image to obtain the homework answers, compare the homework answers with the pre-stored standard answers question by question, determine the wrong questions, mark the wrong question area in the corresponding position in the homework image; call the large model to analyze all wrong questions, extract the knowledge points corresponding to each wrong question, analyze the reasons for each wrong question, and generate learning tags corresponding to the wrong questions.

[0026] Specifically, this step is the initial data collection and analysis stage of the entire method. It mainly completes four tasks: acquiring homework images, automatically identifying wrong questions, in-depth analysis of wrong questions, and generating learning tags, providing a data foundation for subsequent personalized test paper generation.

[0027] In practice, a capture trigger command is first sent to the image acquisition module of the AI ​​learning device to control the module to start capturing images of the complete homework page. During the capture process, the shooting angle can be adjusted using the posture detection unit built into the AI ​​learning device to ensure that the entire homework page is within the capture frame. The captured raw images are then transmitted to the device's built-in image processing unit for preprocessing.

[0028] The image preprocessing stage specifically includes two operations: tilt correction and noise removal. For the deformation of the work page caused by the tilt of the shooting angle, the four boundaries of the work page are identified by the edge detection algorithm, the tilt angle of the page is calculated, and the image is corrected by affine transformation so that the work page presents a standard rectangle. For the noise caused by uneven lighting in the shooting environment and paper texture, the image is smoothed by the filtering algorithm to eliminate interfering pixels and improve the accuracy of subsequent text recognition.

[0029] After image preprocessing, text recognition processing is performed on the corrected assignment images. An optical character recognition algorithm is used to extract the text content line by line, distinguishing between the question stem area and the student's answer area to obtain the complete assignment answer text. During the recognition process, layout analysis algorithms can be used to identify layout elements such as the assignment question number, question type, and answer area. The answer content is then bound and stored with the corresponding question number, forming a structured assignment answer dataset.

[0030] The structured homework answers are then compared question-by-question with a pre-stored dataset of standard answers. This dataset is pre-stored in a structured manner by subject, chapter, and question number, containing the correct answer, scoring criteria, and related knowledge points for each question. During the comparison, for objective questions, the consistency between the answers and the standard answers is directly compared to determine correctness. For subjective questions, key scoring points are extracted from the answers and matched with the scoring points in the standard answers; the degree of matching determines the correctness and severity of the error. After comparing all questions, all incorrectly answered questions are marked as incorrect. Simultaneously, a visual marker box is generated at the location of each incorrect question in the pre-processed homework image, creating a homework image with error markers, which can be output to the display module for students to view.

[0031] After identifying incorrect questions, the stems of all incorrect questions, student responses, and correct answers are input into a pre-trained large language model. The model then performs in-depth analysis of each incorrect question. The large model has a built-in knowledge system database covering the entire academic stage of the corresponding subject. This database is stored hierarchically by subject, chapter, knowledge point, and sub-knowledge point. Each knowledge point is associated with a corresponding test type, common test methods, and typical error causes. The large model uses semantic matching to map the incorrect question stems to knowledge point entries in the knowledge system database, extracting one or more knowledge point entries corresponding to each incorrect question. Simultaneously, it compares the differences between the student's response and the correct answer, and matches this with a typical error cause database to analyze the error type for each incorrect question. These error cause types include four categories: conceptual misunderstanding, calculation error, misreading the question, and lack of mastery of problem-solving methods.

[0032] Based on the extraction of knowledge points and analysis of error causes, learning labels corresponding to the incorrect answers in this assignment are generated. Specifically, the number of incorrect answers for each knowledge point is counted, and combined with the types of errors, the student's mastery of the corresponding knowledge point is calculated according to the preset scoring rules, generating a mastery level label for that knowledge point. The mastery level can be divided into four levels from high to low. At the same time, different error causes under the same knowledge point are statistically analyzed to generate error cause distribution labels, recording the frequency of each type of error cause under the corresponding knowledge point. The mastery level labels and error cause distribution labels for all knowledge points are summarized to form a set of student learning labels corresponding to this assignment. This set of learning labels will serve as the core basis for generating subsequent personalized test questions.

[0033] The learning device's image acquisition module captures images of student assignments, performs text recognition on these images, compares them with standard answers, identifies and marks all incorrect answers in the assignments, and achieves this through the following process: First, it receives the correction command triggered by the user, starts the image acquisition module of the AI ​​learning device, guides the user to adjust the position of the assignment through the preview screen, and automatically triggers the shooting when the assignment page is completely in the center area of ​​the preview screen and the clarity meets the recognition requirements, thus acquiring the complete original image of the assignment page.

[0034] The acquired raw images are first converted to grayscale, transforming the color images into single-channel grayscale images to reduce the computational load of subsequent processing. Then, edge detection is performed, extracting line contours from the image using edge detection operators, identifying the outer border contour of the work page, calculating the coordinates of the four vertices of the page contour, and performing tilt and distortion corrections on the image through perspective transformation, thus correcting the tilted work page to a standard rectangular page with a normal viewing angle.

[0035] After correction, noise removal is performed. The median filtering algorithm is used to eliminate salt and pepper noise in the image, and the Gaussian filtering is used to smooth the subtle noise caused by the paper texture. At the same time, the image contrast is enhanced to improve the contrast between the text area and the background area, thereby further improving the accuracy of text recognition.

[0036] The preprocessed image is subjected to layout analysis. The row and column regions in the image are divided by projection segmentation. The position of the question number of each question is identified by combining the question number features. The question area and answer area of ​​each question are divided accordingly. The question stem text in the question area and the student's answer text in the answer area are extracted respectively. The text is converted into text content by optical character recognition algorithm and stored in a structured manner according to the question number order, forming structured homework data containing question number, question stem and answer content.

[0037] The standard answer dataset corresponding to this assignment is retrieved. This dataset stores the correct answers and scoring rules for each question in the same question number order. The structured assignment data is then matched and compared with the standard answer dataset question by question number: For objective questions (multiple-choice and true / false), the character consistency between the answer text and the standard answer is directly compared; a complete match indicates a correct answer, otherwise an incorrect answer. For fill-in-the-blank questions, multiple answer matching is supported; as long as the answer matches any one of the standard answers, it is considered correct. For subjective questions (calculation and application problems), key steps and the final result are extracted from the answer and matched with the step scoring points and the final result of the standard answer. A score is calculated based on the matching degree; a score below a preset threshold indicates an incorrect answer.

[0038] After determining the correctness of all questions, all questions judged as incorrect are extracted and marked as wrong questions. At the same time, in the preprocessed homework image, a rectangular marking box is generated for the question area corresponding to each wrong question. The marking box is marked with a distinguishable line style. Finally, a homework image with wrong question markings is generated and can be output to the display module for students to intuitively see the location of wrong questions.

[0039] Step S102. Based on the extracted knowledge points and the analyzed reasons for the errors, call the large model to generate multiple test questions associated with the knowledge points of the incorrect questions. This includes: inputting the extracted knowledge points and the analyzed reasons for the errors into the large model; matching test materials consistent with the test points of the incorrect questions through the large model; generating multiple test questions according to different examination angles and question types; controlling the examination scope of the test questions to cover all knowledge points corresponding to the incorrect questions; and formatting the multiple test questions into an electronic test paper according to a preset format.

[0040] Specifically, this step is the core generation stage of the method. Based on the wrong knowledge points, error reasons and learning tags obtained in step S101, personalized test questions with the effect of learning by analogy are generated through the large model, and the standardized layout of the test paper is completed to form an electronic test paper file that can be printed.

[0041] In practice, all knowledge point entries extracted in step S101, the error reason type corresponding to each incorrect question, and the generated set of learning information tags are input into the large model, triggering the large model to execute the personalized test question generation task. The large model has a built-in structured test question material library, which is classified and stored according to five dimensions: subject, knowledge point, question type, difficulty, and examination angle. Each test question is associated with a corresponding knowledge point tag and error reason matching tag.

[0042] After receiving the input parameters, the large model first determines the number of test questions to be generated for each knowledge point based on the mastery level labels in the learning progress label set: the lower the mastery level of a knowledge point, the more test questions are generated to ensure that weak knowledge points are fully consolidated. Then, for each knowledge point, based on the corresponding error reason type, test materials are matched from the test material library with consistent test points and targeted examination angles to address the error reasons. Simultaneously, to achieve a consolidation effect of learning by analogy, multiple test questions with different examination formats and question angles are generated for the same knowledge point, covering common examination methods and common mistakes of that knowledge point. This ensures that the scope of the test questions fully covers all knowledge points corresponding to the incorrect questions and can specifically fill in students' knowledge gaps.

[0043] During the test question generation process, the large model can adjust parameters and rewrite question stems on the original test question materials to avoid generating test questions that are completely identical to the original questions, thus ensuring training effectiveness. After generating all test questions, all test questions are deduplicated to eliminate questions with duplicate content or highly overlapping test points, ensuring the validity of the test paper content.

[0044] After the test questions are generated, all test questions are formatted to form standardized electronic test papers. The formatting process begins by categorizing and sorting all test questions by question type, following the standard order of multiple choice, fill-in-the-blank, and problem-solving questions. Within each question type, questions are further sorted by difficulty level from lowest to highest, conforming to the formatting logic of standard test papers. Next, a corresponding answer space is provided for each question, with the size adjusted according to the question type: smaller answer spaces for objective questions and ample writing space for subjective problem-solving questions.

[0045] During the typesetting process, the system matches the size parameters of standard printing paper, defaulting to commonly used paper sizes. It adjusts the page margins, line spacing, font size, and question spacing according to the paper's width and height to ensure all questions are fully arranged within the paper's dimensions and the layout is clear and legible. After typesetting, a fixed-format electronic exam paper file is generated. This electronic exam paper file uses a universal document format, compatible with the reading and printing needs of various printing devices.

[0046] The process of generating multiple test questions associated with the knowledge points extracted and the reasons for errors analyzed, based on the extracted knowledge points and the analysis of the errors, is specifically implemented through the following steps: The compiled list of knowledge points, the mastery level label for each knowledge point, and the error distribution label are input into the large model. At the same time, the preset test paper generation rule parameters are also input. The test paper generation rule parameters include three categories: total number of test questions, requirement for the proportion of question types, and requirement for difficulty distribution.

[0047] The large model first allocates the number of test questions for each knowledge point based on the mastery level label: the lower the mastery level of the knowledge point, the more test questions are allocated to ensure the training amount of weak knowledge points; for the error reasons with the highest proportion in the error reason distribution, more targeted test questions are allocated accordingly.

[0048] The large model then selects test questions from its built-in question bank based on the principles of matching knowledge points and adapting error causes. For each knowledge point and its corresponding error cause, it selects questions whose examination angle matches that error cause. For example, for errors caused by conceptual confusion, it selects questions that focus on conceptual analysis; for errors caused by calculation mistakes, it selects similar questions that include calculation steps.

[0049] To achieve the effect of consolidating knowledge by analogy, multiple test questions with different question types, different question formats, and different contexts are selected for the same knowledge point. This examines students' mastery of the knowledge point from multiple dimensions and avoids rote memorization of answers. After obtaining the initial set of test questions, the difficulty of the questions is checked to ensure that the difficulty matches the students' level of mastery, avoiding the negative impact of questions that are too difficult or too easy on the training effect.

[0050] For the selected test materials, the large model can rewrite and adjust the question parameters and scenario descriptions as needed to generate entirely new test content, avoiding duplication with the original questions or questions that students have already answered. Ultimately, multiple test questions are generated that meet the requirements for quantity, question type, and difficulty, and the scope of all test questions fully covers the knowledge points corresponding to all incorrect questions, achieving targeted reinforcement.

[0051] Step S103. Obtain the connection type and format specifications of the connectable printing device, adjust the output format of the electronic test paper according to the connection type and format specifications of the printing device, and send a print control command to the printing device to output the paper test paper.

[0052] Specifically, this step is the output execution stage of the method. The core implementation is automatic recognition of the printing device, adaptive adjustment of the format, and issuance of printing instructions, ensuring that the generated electronic test paper can be adapted to printing devices with different connection methods to complete the output of the paper test paper.

[0053] In practice, the device scanning function is first activated to scan all connectable printing devices in the current environment. The scanning process simultaneously supports device detection using multiple connection protocols, including local wired connections, LAN wireless connections, and Bluetooth connections, identifying the device identifiers and connection types of all connectable printing devices.

[0054] After obtaining a list of available printing devices, the user can select a target printing device, or the default printing device can be automatically selected according to preset priorities. After selecting a target printing device, a parameter reading command is sent to the target printing device to read the format specification parameters supported by the corresponding printing device, including four types of parameter information: supported paper size, print resolution, color mode, and supported file formats.

[0055] Based on the format specifications of the target printing device, the output format of the electronic exam paper generated in step S102 is adjusted. Specifically, this includes: adjusting the page size and margins of the electronic exam paper according to the paper sizes supported by the printing device to ensure the exam content perfectly matches the paper size; adjusting the resolution of text and lines within the exam paper according to the resolution parameters of the printing device to ensure clear printed output; and converting the file format of the electronic exam paper according to the file formats supported by the printing device to ensure the printing device can read the file correctly.

[0056] After format adaptation and adjustment, the final print output file is generated. A print control command is sent to the target printing device, transmitting the print output file to the target printing device and controlling the printing device to execute the printing operation, outputting the paper exam paper. During the printing process, the system can receive real-time printing status information from the printing device, including three types of information: printing progress, printing completion status, and paper outage fault. This status information is also synchronously output to the display module of the AI ​​learning device for user viewing.

[0057] In some embodiments, the method further includes: after identifying and marking all incorrect questions, calling a large model to generate explanation content for the knowledge points corresponding to each incorrect question. The explanation content sequentially includes the explanation of the core concepts of the knowledge point, the location of the error cause of the corresponding incorrect question, the breakdown of the standard problem-solving ideas and steps, and the summary of general problem-solving techniques for similar question types; displaying the explanation content of each incorrect question through the display module of the learning device, and displaying the mastery level feedback option after the explanation of each incorrect question is completed, receiving the mastery level feedback submitted by the student, which is divided into levels of complete understanding, basic understanding, and still having questions; combining the student's mastery level feedback with the generated learning labels to jointly adjust the number and difficulty of the subsequently generated test questions: for knowledge points where the feedback is "still having questions", increasing the number of corresponding test questions generated, while reducing the difficulty of the questions and increasing the proportion of basic questions; for knowledge points where the feedback is "completely understood", reducing the number of corresponding test questions generated, while increasing the difficulty of the questions and increasing the proportion of extended questions.

[0058] This embodiment takes the paper-based homework for the seventh-grade mathematics chapter on "Linear Equations in One Variable" as the processing object. Based on an AI learning tablet equipped with a high-definition image acquisition module, an NPU image processing unit, and a lightweight large language model, it fully executes the entire process of personalized test paper generation and printing control. The three core stages are sequentially connected to form a complete processing link. The specific implementation process is as follows: Phase 1: Homework Image Acquisition, Error Recognition, and Learning Tag Generation (corresponding to step S101): After students complete the homework consisting of 10 questions (6 objective questions and 4 subjective calculation questions), the tablet's "Homework Correction" function is triggered, activating the rear 13-megapixel image acquisition module to output a real-time preview. The tablet's built-in posture detection unit identifies the four boundaries of the homework page in real time. When it detects that the A4 homework paper is completely in the center of the image, the tilt angle is less than 5°, and the clarity meets the OCR recognition threshold, it automatically triggers the shutter to acquire the original color image with a resolution of 4160×3120 pixels.

[0059] The raw image is transferred to the NPU image processing unit for preprocessing: Grayscale conversion: Converting a three-channel color image into a single-channel 8-bit grayscale image reduces the amount of data required for subsequent calculations; Tilt correction: The Canny edge detection operator is used to extract the page outline and the page tilt angle is calculated to be 3.2°. Affine correction is performed through the perspective transformation matrix to correct the tilted page to a standard A4 rectangle ratio. Noise Removal and Enhancement: Salt and pepper noise is eliminated by 3×3 kernel median filtering, paper texture noise is smoothed by 5×5 kernel Gaussian filtering, and then the contrast between text and background is improved by histogram equalization. The text recognition accuracy can be improved by more than 12% after preprocessing.

[0060] Layout analysis and character recognition are performed on the preprocessed images: the rows and columns are divided by horizontal and vertical projection methods, and the question stems and answer areas of the 10 questions are located by combining the features of question numbers such as "Question X" and "(X)". The question stem text and student answer text of each question are extracted by the optical character recognition engine and stored in a structured manner according to the question number as a homework answer dataset.

[0061] The pre-stored standard answer dataset was retrieved and compared question by question to determine incorrect answers: Questions 1-3 (multiple choice): In question 2, the student's answer "B" did not match the standard answer "C", so it was determined to be an incorrect question; Questions 4-6 (fill-in-the-blank): In question 5, the student's calculation result "x=4" did not match the standard answer "x=6", so it was determined to be an incorrect question; Questions 7-10 (calculation application questions): The key steps and final results of the answers were extracted and matched with the standard answer's scoring points. In question 8, the student omitted the step of changing the sign and the result was incorrect, with a score rate of 30%, which was below the 60% passing threshold, so it was determined to be an incorrect question; In question 9, the omission of the "tax included" condition in the question stem led to an incorrect equation, with a score rate of 20%, so it was determined to be an incorrect question.

[0062] Ultimately, four incorrect questions (questions 2, 5, 8, and 9) were identified. A 2px wide red rectangle marking box was generated at the corresponding question position in the preprocessed image, and the homework image with the incorrect question markings was output to the flat panel display.

[0063] The question numbers, question stems, student answers, and standard answers of the four incorrect questions are concatenated into a pre-defined structured template as input instructions, which are then passed to a lightweight large-scale language model for parsing. The large model performs semantic matching based on a built-in junior high school mathematics knowledge system map (four levels: grade level, chapter, knowledge point, and sub-knowledge point). Simultaneously, it compares the differences between the answers and the standard answers, matches them against a database of typical error causes, and outputs the parsing results. Question 2: Corresponding knowledge point (definition of a linear equation in one variable), reason for error (conceptual confusion); Question 5: Corresponding knowledge point (solving linear equations in one variable - transposing terms), reason for error (calculation error); Question 8: Corresponding knowledge point (solving linear equations in one variable - transposing terms and removing parentheses), reason for error (not mastering the solution method); Question 9: Corresponding knowledge point (practical application of linear equations in one variable - profit problems), reason for error (misreading the question).

[0064] Learning labels are generated based on the analysis results: the number of wrong questions is counted by knowledge point. Among them, (solving a linear equation in one variable - transposing terms) corresponds to 2 wrong questions, accounting for 100% of the total number of questions for this knowledge point; the other 3 knowledge points each correspond to 1 wrong question, accounting for 50%.

[0065] The mastery score is calculated according to the preset scoring rules: 100 points for the basic score, 20 points deducted for each incorrect answer; an additional 15 points are deducted for errors in concept understanding and methodology, an additional 5 points for calculation errors, and an additional 10 points for errors in reading the question carefully. The calculation yields: (Solve a linear equation in one variable - transpose terms): 40 points, corresponding to the "not mastered" level; (Definition of a linear equation in one variable): 65 points, corresponding to the preliminary mastery level; (Practical application of linear equations in one variable - profit problem): Score 70 points, corresponding to the preliminary mastery level.

[0066] At the same time, error distribution labels for each knowledge point are generated, and a complete set of learning information labels is compiled and persistently stored in the student's personal learning information file, serving as the core basis for subsequent test paper compilation.

[0067] Phase Two: Personalized Test Paper Generation and Layout (corresponding to step S102): Input the set of learning tags, the list of knowledge points, and the preset test paper generation rules (total number of 15 questions, question type ratio of 40% selection + 30% fill-in-the-blank + 30% problem solving, difficulty distribution of easy: medium: difficult = 3:5:2) into the large model to trigger the generation of personalized test questions.

[0068] The large model first allocates the number of questions for each knowledge point based on the level of mastery: the lower the level of mastery, the more questions are allocated to the knowledge points; 7 questions are allocated to knowledge points that have not been mastered (redirected); and 4 questions are allocated to each of the two knowledge points that have been initially mastered, for a total of 15 questions. Then, questions are selected from the built-in structured question resource library (categorized by subject, knowledge point, question type, difficulty, and examination angle) according to the principle of (knowledge point matching + error cause adaptation). For the main reason of error in the knowledge point of transposition (lack of mastery of problem-solving methods), 3 basic questions on transposition rules, 2 transposition calculation practice questions, and 2 comprehensive questions on removing parentheses were matched; For the reasons of (concept confusion) in the definition of knowledge points, match 2 concept identification multiple-choice questions and 2 easily confused concept judgment questions; Based on the error of (misreading the question) in the knowledge point of profit application, match 2 fill-in-the-blank questions with hidden conditions and 2 application questions with multiple conditions.

[0069] The large model rewrites the parameters of the selected test questions (changing values ​​and adjusting the application context) to avoid duplication with the original questions; then it performs deduplication verification, excluding one question with highly overlapping test points and adding one extension question of the same difficulty, finally generating 15 test questions, which fully cover the knowledge points corresponding to all the wrong questions.

[0070] After generating the test questions, standardized formatting is implemented: Questions are categorized and sorted by type: 6 multiple-choice questions, 4 fill-in-the-blank questions, and 5 problem-solving questions are arranged sequentially, with each question type sorted from lowest to highest difficulty; a basic layout is set: suitable for A4 paper, with top and bottom margins of 2.5cm and left and right margins of 2cm; a test paper title ("Consolidation Test for Incorrectly Solving Linear Equations in One Variable") is placed at the top, and below the title are fields for name, class, and score; each question type is preceded by a bolded title and its corresponding score (4 points for multiple-choice questions, 5 points for fill-in-the-blank questions, and 10 points for problem-solving questions, for a total of 100 points); answer spaces are provided: a horizontal line is added after the question stem for multiple-choice and fill-in-the-blank questions; 6-10 lines of blank space are reserved below the question stem for problem-solving questions, with more blank lines for higher score questions. Formatting and pagination verification: The question stems are in Song typeface, size 5, and the question numbers and question type titles are in boldface, size 5, with a line spacing of 1.5. After verification, the 15 questions are arranged in two pages with no questions breaking across pages, and a standard electronic test paper file in PDF format is generated.

[0071] Phase 3: Printing Device Adaptation and Printing Control (corresponding to step S103): Start the tablet's device discovery service and scan for connectable printing devices through three channels: Wi-Fi LAN, Bluetooth, and USB interface. A total of 2 available devices are detected: a wireless laser printer on the LAN and a desktop Bluetooth thermal printer.

[0072] The user selects a wireless laser printer as the target device. The tablet establishes a communication connection via Wi-Fi, sends a parameter query command, and reads the printer's format specifications: supports A4 / A5 paper, 1200×1200dpi printing resolution, black and white grayscale mode, compatible with PDF / Word format, and supports automatic duplex printing.

[0073] Adaptive formatting of electronic exam papers based on printer parameters: Keep the page size at A4 and fine-tune the margins to fit the printer's printable area boundaries to avoid content cropping; Adjust the test paper rendering resolution to 1200dpi to ensure that the text and lines are printed clearly without jagged edges. Convert to pure black and white grayscale mode to remove all color elements and adapt to black and white laser printing; Preserves the PDF file format; no additional conversion required.

[0074] After adjustments are made, the final print output file is generated. After previewing and verifying that the content is complete and the layout is correct, the tablet sends print control commands and the print file to the printer, setting it to single-sided printing mode. During the printing process, the tablet receives status information returned by the printer in real time and displays the printing progress synchronously on the screen. After printing is complete, a "(Printing Complete)" prompt pops up, and a two-page paper-based test paper for reviewing and consolidating incorrect answers is finally output.

[0075] Based on this embodiment, a closed-loop update of learning progress can be implemented. After students complete the paper test, the test paper image is captured again via tablet and graded. The overlap of knowledge points between the new incorrect questions and the original incorrect questions is compared. If there are still errors in the knowledge points being moved, the mastery level is further reduced, and the number of questions and the focus are increased when generating subsequent test papers; if all the knowledge points of the original incorrect questions are answered correctly, the mastery level is increased, and the corresponding number of questions is reduced, forming a closed-loop learning mechanism of "identification-consolidation-verification-update".

[0076] If the scan does not detect a directly connectable printing device, the tablet will send an electronic exam paper and a print trigger command to the pre-bound mobile terminal. The mobile terminal will then connect to a nearby printing device to complete the printing and synchronize the printing status back to the learning device, thus achieving cross-terminal printing adaptation.

[0077] In some embodiments, the step of calling the large model to analyze all incorrect questions, extracting the knowledge points corresponding to each incorrect question, and analyzing the reasons for each incorrect question includes: extracting the question stem and answer content of each incorrect question and inputting them into the large model; matching the knowledge system database through the large model to obtain the knowledge point entries corresponding to each incorrect question; and comparing the differences between the answer content and the correct answer through the large model to determine the type of error reason corresponding to each incorrect question.

[0078] This embodiment provides a detailed explanation of the specific implementation method of analyzing incorrect questions, extracting knowledge points, and analyzing the causes of errors in step S101.

[0079] In this embodiment, the process of calling the large model to analyze all incorrect questions, extracting the knowledge points corresponding to each incorrect question, and analyzing the reasons for each incorrect question is implemented through the following process: All incorrect questions, including their question numbers, question stems, student responses, and correct answers, are compiled into structured input text. This text is then assembled into complete input instructions according to a pre-defined prompt template and fed into a pre-trained large language model. The prompt template explicitly specifies the output format requirements, requiring the large model to output the corresponding knowledge points and reasons for errors for each incorrect question, ensuring the output conforms to a structured format for easy data extraction later.

[0080] After receiving input instructions, the large model first performs semantic understanding on the question stem of each incorrect question, extracting core test point keywords from the question stem and semantically matching these keywords with the built-in knowledge system database. The knowledge system database is constructed according to a subject-specific knowledge graph, containing hierarchical relationships between different subjects from grade level to chapter and then to specific knowledge points. Each knowledge point has a corresponding set of keywords and typical question stem features. Through similarity calculation, the knowledge point entry with the highest matching degree to the test point of the incorrect question is selected as the corresponding knowledge point for that incorrect question. For comprehensive questions involving multiple knowledge points, multiple related knowledge points can be matched and output simultaneously.

[0081] After matching the knowledge points, the large model compares the student's answers with the correct answers to pinpoint the specific location of the error: for questions with incorrect results, it traces back the derivation steps in the answer process to identify the point where the error occurred; for questions with incorrect steps but a coincidentally correct result, it identifies the flawed derivation logic. Combining this with a built-in database of typical error causes, the model categorizes the error types and determines the corresponding error reasons. Typical error reasons include six categories: conceptual confusion, incorrect formula memorization, calculation errors, omission of conditions during question reading, incorrect problem-solving approach, and errors in writing style.

[0082] After the large model completes the analysis of all incorrect questions, it outputs the analysis results in a preset structured format, including the question number, knowledge point list, and error reason type for each incorrect question. The corresponding data can be directly extracted from the output results for subsequent learning tag generation and test paper assembly.

[0083] In some embodiments, generating learning labels corresponding to incorrect questions includes: counting the number of incorrect questions for the same knowledge point, combining the error reason type of each incorrect question to generate a mastery level label and error reason distribution label for the corresponding knowledge point, and summarizing all the labels to form a learning label set for the corresponding student.

[0084] This embodiment provides a detailed explanation of the specific implementation method for generating learning tags in step S101.

[0085] In this embodiment, the generation of learning tags corresponding to incorrect questions is specifically achieved through the following process: First, extract the knowledge points and reasons for all incorrect questions from the analysis results output by the large model. Then, perform statistical summaries based on knowledge points, and count the total number of incorrect questions under each knowledge point, as well as the frequency of each type of error under each knowledge point.

[0086] For each knowledge point, the mastery score is calculated based on the proportion of incorrect answers to the total number of assignments for that knowledge point, combined with the type of error. The calculation rules are pre-set as follows: the higher the percentage of incorrect answers, the lower the mastery score; if the error is a fundamental error related to conceptual understanding, more points will be deducted; if it is an accidental error related to calculation, fewer points will be deducted.

[0087] Based on the calculated mastery scores, the mastery of knowledge points is divided into four levels: proficient, basic, preliminary, and no mastery. Each level corresponds to a score range, and a mastery label for the corresponding knowledge point is generated based on the score range.

[0088] At the same time, for each knowledge point, the frequency of various error reasons is counted, and sorted from high to low frequency to generate error reason distribution labels for the corresponding knowledge point. Error reason distribution labels record the most important and secondary error reasons under the corresponding knowledge point, which can be used for subsequent targeted question creation.

[0089] The mastery level labels and error distribution labels for all knowledge points are summarized and combined with student identification information to form a set of student learning labels for this assignment. This set of learning labels can be persistently stored and merged with students' historical learning data to create a continuously updated student learning profile, providing data support for long-term personalized learning planning.

[0090] In some embodiments, the step of formatting multiple test questions into an electronic test paper according to a preset layout includes: classifying and sorting all test questions according to question type, setting a blank answer area for each question, adjusting the spacing between questions and font size to match the size parameters of standard paper, and generating an electronic test paper file that conforms to printing specifications.

[0091] This embodiment provides a detailed explanation of the specific implementation method of electronic test paper layout in step S102.

[0092] In this embodiment, the step of formatting multiple test questions into an electronic test paper according to a preset format is achieved through the following process: First, all generated test questions are categorized by question type into five categories: multiple choice, fill-in-the-blank, calculation, application, and problem-solving. Following the standard test paper layout, multiple choice, fill-in-the-blank, calculation, application, and problem-solving questions are arranged in that order. Within each question type, questions are sorted from easiest to hardest, creating an ordered question list.

[0093] Next, set the basic layout parameters of the exam paper, including page size, margins, header and footer content, and title style. The exam title is placed at the top of the paper, and below it are three fields for filling in the name, class, and score. Each question type is preceded by a question type title and corresponding score description, and each question is preceded by a consecutive question number, arranged sequentially according to the question type.

[0094] For each question, a corresponding size of blank answer area is set according to the question type: for multiple-choice and fill-in-the-blank questions, a horizontal line of corresponding length is set after the question stem; for calculation and problem-solving questions, multiple blank areas are set below the question stem to meet the needs of writing the solution process. The size of the blank area is dynamically adjusted according to the question score and the estimated writing amount, with larger blank areas corresponding to higher score questions.

[0095] Adjust the font formatting within the exam paper. Use standard font size for the question stems and bold for question numbers and question type titles to ensure clear hierarchy. Adjust line spacing and character spacing to ensure the exam content is clear and easy to read. After completing the initial layout, perform pagination verification. If a single page cannot accommodate all the content, automatically paginate to ensure that questions do not break across pages and to guarantee the standardization of the layout.

[0096] The final output is a standard format electronic exam paper file containing complete exam content and layout, which can be directly used for display preview and printing.

[0097] In some embodiments, obtaining the connection type and format specifications of the connectable printing devices, and adjusting the output format of the electronic exam paper according to the connection type and format specifications of the printing devices, includes: scanning all currently connectable printing devices, reading the connection method and supported paper size and resolution parameters of each printing device, adjusting the page size, resolution and layout of the electronic exam paper according to the parameters of the selected printing device, and generating an output file adapted to the selected printing device.

[0098] This embodiment provides a detailed explanation of the specific implementation method of printing device adaptation and format adjustment in step S103.

[0099] In this embodiment, obtaining the connection type and format specifications of the connectable printing device, and adjusting the output format of the electronic exam paper according to the connection type and format specifications of the printing device, is specifically achieved through the following process: The device discovery service is activated, and the system scans for nearby connectable printing devices through multiple communication channels. This includes detecting locally connected printing devices via USB, wireless printing devices within the same network via LAN communication protocols, and nearby Bluetooth printing devices via Bluetooth protocols. Once the scan is complete, all connectable printing devices are aggregated into a list containing information such as device name, connection type, and connection status for each device.

[0100] Receive the target printing device selected by the user and establish the corresponding communication connection according to the connection type of the target printing device: local direct connection devices establish connection through USB communication protocol, local area network devices establish connection through network protocol, and Bluetooth devices establish connection through Bluetooth pairing.

[0101] After the communication connection is successfully established, a parameter query command is sent to the target printing device to read the printing device's capability parameters, including five categories of format specification parameters: supported paper sizes, supported printing resolutions, supported color modes, supported file formats, and whether duplex printing is supported.

[0102] Based on the retrieved format specification parameters, the electronic exam paper undergoes adaptive adjustments: adjusting the page width, height, and margins according to supported paper sizes to ensure the exam content perfectly matches the paper size; adjusting the rendering resolution of text and lines within the exam paper according to supported print resolutions to ensure clear and unblurred printing; adjusting the exam paper to either black-and-white or color mode according to supported color modes; and converting the electronic exam paper to the corresponding format according to supported file formats to ensure the printing device can correctly parse the file.

[0103] After completing all formatting adjustments, generate the final print output file and preview and verify it. Once the content is complete and the layout is correct, save it as a file to be printed and wait for the print command to be triggered.

[0104] In some embodiments, the method further includes: acquiring images of paper test papers completed by students, identifying and correcting the test paper images to obtain new incorrect questions, comparing the overlap of knowledge points between the new incorrect questions and the original incorrect questions, and updating the corresponding student's learning tag set.

[0105] This embodiment is an optimized scheme for closed-loop learning information updates, which is added to the core method.

[0106] In this embodiment, the method further includes a secondary grading and learning progress update step after the paper test papers are completed. The specific implementation process is as follows: After students complete the printed test paper, the AI ​​learning device captures the image of the completed test paper again through its image acquisition module. Following the same image processing, text recognition, and answer comparison process as step S101, the test paper is graded to obtain the new incorrect questions.

[0107] Extract the knowledge points and reasons for the errors corresponding to the new incorrect answers. Compare the overlap of knowledge points between the new incorrect answer data and the original incorrect answer data generated from this assignment, and count the number of times the same knowledge point is repeatedly answered incorrectly in two attempts. If the same knowledge point appears incorrectly in both the original assignment and the consolidation test, the corresponding knowledge point is determined to be a key weak point for the student, and the mastery level of the corresponding knowledge point is lowered accordingly. If all the knowledge points corresponding to the original incorrect answers are answered correctly in the consolidation test, the mastery level of the corresponding knowledge point is raised accordingly.

[0108] Based on the comparison results, the set of learning labels for the corresponding students is updated, and the mastery level labels and error distribution labels for each knowledge point are adjusted. Simultaneously, the updated learning data is stored in the students' historical learning files. When generating subsequent reinforcement tests, the tests will be compiled based on the updated learning labels. For weak knowledge points with repeated errors, the number and relevance of test questions will be further increased, forming a closed-loop learning mechanism of error identification, reinforcement training, effect verification, and learning update, continuously improving the accuracy and effectiveness of error reinforcement.

[0109] In some embodiments, the method further includes: when no directly connectable printing device is detected, sending an electronic test paper file and a print trigger command to a pre-bound mobile terminal, connecting to the corresponding printing device through the mobile terminal, and performing the test paper printing operation.

[0110] This embodiment is an extended cross-terminal printing solution added to the core method.

[0111] In this embodiment, the method further includes a control method for mobile terminal-linked printing, the specific implementation process of which is as follows: If no printing device that can directly connect to the AI ​​learning device is detected after the printing device scanning step S103, the cross-terminal printing process is triggered.

[0112] First, the system retrieves information about the mobile terminal device that is pre-linked to the corresponding AI learning device. The mobile terminal and the AI ​​learning device are linked in advance through account binding. A print request notification is sent to the linked mobile terminal via wireless network, and the generated electronic test paper file is simultaneously transmitted to the mobile terminal.

[0113] After receiving the print request and electronic exam paper file, the mobile terminal automatically invokes its own printing service, scans for connectable printing devices in the vicinity of the mobile terminal, selects the target printing device, performs format adaptation and print triggering operations, and controls the corresponding printing device to output the paper exam paper.

[0114] During the printing process, the mobile terminal synchronizes the printing status information back to the AI ​​learning device in real time. The AI ​​learning device then displays the current printing progress and status on its display module. This method overcomes the limitations of the AI ​​learning device's own connectivity, enabling it to connect to various printing devices via the mobile terminal, significantly expanding the range of compatible printing devices and meeting printing needs in different scenarios.

[0115] In some embodiments, the method further includes: after identifying and marking all incorrect questions, calling a large model to generate explanations of the knowledge points corresponding to each incorrect question, displaying the explanations through the display module of the learning device, receiving feedback on the students' level of mastery, and adjusting the number and difficulty of subsequent test questions based on the feedback on the level of mastery.

[0116] This embodiment is an optimization scheme for incorrect question explanation and feedback-based test paper generation, which is added to the core method.

[0117] In this embodiment, the method also includes a function that links the explanation of incorrect knowledge points with mastery feedback to create test papers. The specific implementation process is as follows: After all incorrect questions are identified and marked in step S101, the large model is invoked to generate explanation content for the knowledge points corresponding to each incorrect question. The explanation content includes four categories: explanation of the core concepts of the knowledge points, analysis of the reasons for the incorrect questions, explanation of the correct problem-solving ideas and steps, and summary of problem-solving techniques for similar questions. The explanation content is organized logically from simple to complex, and is suitable for the comprehension ability of students at the corresponding grade level.

[0118] After generating the explanation content, the AI ​​learning device displays the explanations for each incorrect question through its display module. Students can view the explanations for each question through the interactive module. After completing the explanation for a single incorrect question, the display module shows feedback options on the student's level of understanding, allowing students to submit feedback on their level of mastery of the corresponding knowledge point. Feedback options include "completely understand," "basically understand," and "still have questions."

[0119] After receiving feedback from students on their level of mastery, the feedback results are combined with the learning information tags generated in step S101 to adjust the parameters of the subsequent test questions: for knowledge points that students still have questions about, the number of corresponding test questions is increased, while the difficulty of the questions is reduced and the proportion of basic questions is increased; for knowledge points that students fully understand, the number of test questions is appropriately reduced, the difficulty of the questions is increased, and the focus is on extension-type questions.

[0120] By combining students' subjective feedback on their mastery with objective data on incorrect answers, the number and difficulty of personalized test questions are adjusted to make the generated test papers more closely match students' actual mastery levels, thereby further improving the targeted nature and effectiveness of consolidation of incorrect answers.

[0121] Please see Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a personalized test paper generation and printing control system 200 based on an AI learning device provided in this application embodiment. The personalized test paper generation and printing control system 200 based on an AI learning device is used to execute the steps of the personalized test paper generation and printing control method based on an AI learning device shown in the above embodiments. The personalized test paper generation and printing control system 200 based on an AI learning device can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0122] like Figure 3 As shown, the personalized test paper generation and printing control system 200 based on AI learning devices includes: The tag generation unit 201 is used to acquire student homework images through the image acquisition module of the learning device, perform text recognition on the homework images and compare them with standard answers, identify and mark all wrong questions in the homework, including: starting the image acquisition module to acquire a complete image of the homework page, performing tilt correction and noise removal on the complete image, performing text recognition on the processed image to obtain the homework answers, comparing the homework answers with the pre-stored standard answers question by question, determining the wrong answers, marking the wrong question area in the corresponding position in the homework image; calling the large model to analyze all wrong questions, extracting the knowledge points corresponding to each wrong question, analyzing the reasons for each wrong question, and generating learning tags corresponding to the wrong questions; The test paper formatting unit 202 is used to generate multiple test questions associated with the knowledge points of the wrong questions by calling a large model based on the extracted knowledge points and the reasons for the errors analyzed. This includes: inputting the extracted knowledge points and the reasons for the errors analyzed into the large model; matching test question materials that are consistent with the test points of the wrong questions through the large model; generating multiple test questions according to different examination angles and question types; controlling the examination scope of the test questions to cover all knowledge points corresponding to the wrong questions; and formatting the multiple test questions into an electronic test paper according to a preset format. The test paper output unit 203 is used to obtain the connection type and format specifications of the connectable printing device, adjust the output format of the electronic test paper according to the connection type and format specifications of the printing device, and send printing control instructions to the printing device to output the paper test paper.

[0123] In some embodiments, the step of calling the large model to analyze all incorrect questions, extracting the knowledge points corresponding to each incorrect question, and analyzing the reasons for each incorrect question includes: extracting the question stem and answer content of each incorrect question and inputting them into the large model; matching the knowledge system database through the large model to obtain the knowledge point entries corresponding to each incorrect question; and comparing the differences between the answer content and the correct answer through the large model to determine the type of error reason corresponding to each incorrect question.

[0124] In some embodiments, generating learning labels corresponding to incorrect questions includes: counting the number of incorrect questions for the same knowledge point, combining the error reason type of each incorrect question to generate a mastery level label and error reason distribution label for the corresponding knowledge point, and summarizing all the labels to form a learning label set for the corresponding student.

[0125] In some embodiments, the step of formatting multiple test questions into an electronic test paper according to a preset layout includes: classifying and sorting all test questions according to question type, setting a blank answer area for each question, adjusting the spacing between questions and font size to match the size parameters of standard paper, and generating an electronic test paper file that conforms to printing specifications.

[0126] In some embodiments, obtaining the connection type and format specifications of the connectable printing devices, and adjusting the output format of the electronic exam paper according to the connection type and format specifications of the printing devices, includes: scanning all currently connectable printing devices, reading the connection method and supported paper size and resolution parameters of each printing device, adjusting the page size, resolution and layout of the electronic exam paper according to the parameters of the selected printing device, and generating an output file adapted to the selected printing device.

[0127] In some embodiments, the method further includes: acquiring images of paper test papers completed by students, identifying and correcting the test paper images to obtain new incorrect questions, comparing the overlap of knowledge points between the new incorrect questions and the original incorrect questions, and updating the corresponding student's learning tag set.

[0128] In some embodiments, the method further includes: when no directly connectable printing device is detected, sending an electronic test paper file and a print trigger command to a pre-bound mobile terminal, connecting to the corresponding printing device through the mobile terminal, and performing the test paper printing operation.

[0129] In some embodiments, the method further includes: after identifying and marking all incorrect questions, calling a large model to generate explanation content for the knowledge points corresponding to each incorrect question. The explanation content sequentially includes the explanation of the core concepts of the knowledge point, the location of the error cause of the corresponding incorrect question, the breakdown of the standard problem-solving ideas and steps, and the summary of general problem-solving techniques for similar question types; displaying the explanation content of each incorrect question through the display module of the learning device, and displaying the mastery level feedback option after the explanation of each incorrect question is completed, receiving the mastery level feedback submitted by the student, which is divided into levels of complete understanding, basic understanding, and still having questions; combining the student's mastery level feedback with the generated learning labels to jointly adjust the number and difficulty of the subsequently generated test questions: for knowledge points where the feedback is "still having questions", increasing the number of corresponding test questions generated, while reducing the difficulty of the questions and increasing the proportion of basic questions; for knowledge points where the feedback is "completely understood", reducing the number of corresponding test questions generated, while increasing the difficulty of the questions and increasing the proportion of extended questions.

[0130] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the personalized test paper generation and printing control system and its modules based on the AI ​​learning device described above can be found in the corresponding contents of the various embodiments of the personalized test paper generation and printing control method based on the AI ​​learning device, and will not be repeated here.

[0131] The aforementioned method for personalized test paper generation and printing control based on AI learning devices can be implemented as a computer program, which can, for example... Figure 3 It runs on the system shown.

[0132] Please see Figure 4 , Figure 4 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0133] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any personalized test paper generation and printing control method based on AI learning devices.

[0134] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0135] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any personalized test paper generation and printing control method based on AI learning devices.

[0136] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0137] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0138] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: The learning device's image acquisition module captures images of student assignments. Text recognition is performed on these images, comparing them with standard answers to identify and mark all incorrect answers. This process includes: activating the image acquisition module to capture a complete image of the assignment page; performing tilt correction and noise removal on the complete image; performing text recognition on the processed image to obtain the assignment answers; comparing the assignment answers with pre-stored standard answers question by question to identify incorrect answers and marking the corresponding areas in the assignment image; and then using a large model to analyze all incorrect answers, extracting the knowledge points corresponding to each incorrect answer, analyzing the reasons for each incorrect answer, and generating learning tags corresponding to the incorrect answers. Based on the extracted knowledge points and the analyzed reasons for errors, a large model is invoked to generate multiple test questions associated with the knowledge points of the incorrect questions. This includes: inputting the extracted knowledge points and the analyzed reasons for errors into the large model; matching test materials consistent with the test points of the incorrect questions through the large model; generating multiple test questions according to different examination angles and question types; controlling the examination scope of the test questions to cover all knowledge points corresponding to the incorrect questions; and formatting the multiple test questions into an electronic test paper according to a preset format. Obtain the connection type and format specifications of the connectable printing devices, adjust the output format of the electronic exam paper according to the connection type and format specifications of the printing devices, and send print control commands to the printing devices to output the paper exam paper.

[0139] In some embodiments, the step of calling the large model to analyze all incorrect questions, extracting the knowledge points corresponding to each incorrect question, and analyzing the reasons for each incorrect question includes: extracting the question stem and answer content of each incorrect question and inputting them into the large model; matching the knowledge system database through the large model to obtain the knowledge point entries corresponding to each incorrect question; and comparing the differences between the answer content and the correct answer through the large model to determine the type of error reason corresponding to each incorrect question.

[0140] In some embodiments, generating learning labels corresponding to incorrect questions includes: counting the number of incorrect questions for the same knowledge point, combining the error reason type of each incorrect question to generate a mastery level label and error reason distribution label for the corresponding knowledge point, and summarizing all the labels to form a learning label set for the corresponding student.

[0141] In some embodiments, the step of formatting multiple test questions into an electronic test paper according to a preset layout includes: classifying and sorting all test questions according to question type, setting a blank answer area for each question, adjusting the spacing between questions and font size to match the size parameters of standard paper, and generating an electronic test paper file that conforms to printing specifications.

[0142] In some embodiments, obtaining the connection type and format specifications of the connectable printing devices, and adjusting the output format of the electronic exam paper according to the connection type and format specifications of the printing devices, includes: scanning all currently connectable printing devices, reading the connection method and supported paper size and resolution parameters of each printing device, adjusting the page size, resolution and layout of the electronic exam paper according to the parameters of the selected printing device, and generating an output file adapted to the selected printing device.

[0143] In some embodiments, the method further includes: acquiring images of paper test papers completed by students, identifying and correcting the test paper images to obtain new incorrect questions, comparing the overlap of knowledge points between the new incorrect questions and the original incorrect questions, and updating the corresponding student's learning tag set.

[0144] In some embodiments, the method further includes: when no directly connectable printing device is detected, sending an electronic test paper file and a print trigger command to a pre-bound mobile terminal, connecting to the corresponding printing device through the mobile terminal, and performing the test paper printing operation.

[0145] In some embodiments, the method further includes: after identifying and marking all incorrect questions, calling a large model to generate explanation content for the knowledge points corresponding to each incorrect question. The explanation content sequentially includes the explanation of the core concepts of the knowledge point, the location of the error cause of the corresponding incorrect question, the breakdown of the standard problem-solving ideas and steps, and the summary of general problem-solving techniques for similar question types; displaying the explanation content of each incorrect question through the display module of the learning device, and displaying the mastery level feedback option after the explanation of each incorrect question is completed, receiving the mastery level feedback submitted by the student, which is divided into levels of complete understanding, basic understanding, and still having questions; combining the student's mastery level feedback with the generated learning labels to jointly adjust the number and difficulty of the subsequently generated test questions: for knowledge points where the feedback is "still having questions", increasing the number of corresponding test questions generated, while reducing the difficulty of the questions and increasing the proportion of basic questions; for knowledge points where the feedback is "completely understood", reducing the number of corresponding test questions generated, while increasing the difficulty of the questions and increasing the proportion of extended questions.

[0146] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the personalized test paper generation and printing control method based on an AI learning device provided in any embodiment of this application.

[0147] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0148] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for personalized test paper generation and printing control based on AI learning devices, characterized in that, include: The learning device's image acquisition module captures images of student assignments. Text recognition is performed on these images, comparing them with standard answers to identify and mark all incorrect answers. This process includes: activating the image acquisition module to capture a complete image of the assignment page; performing tilt correction and noise removal on the complete image; performing text recognition on the processed image to obtain the assignment answers; comparing the assignment answers with pre-stored standard answers question by question to identify incorrect answers and marking the corresponding areas in the assignment image; and then using a large model to analyze all incorrect answers, extracting the knowledge points corresponding to each incorrect answer, analyzing the reasons for each incorrect answer, and generating learning tags corresponding to the incorrect answers. Based on the extracted knowledge points and the analyzed reasons for errors, a large model is invoked to generate multiple test questions associated with the knowledge points of the incorrect questions. This includes: inputting the extracted knowledge points and the analyzed reasons for errors into the large model; matching test materials consistent with the test points of the incorrect questions through the large model; generating multiple test questions according to different examination angles and question types; controlling the examination scope of the test questions to cover all knowledge points corresponding to the incorrect questions; and formatting the multiple test questions into an electronic test paper according to a preset format. Obtain the connection type and format specifications of the connectable printing devices, adjust the output format of the electronic exam paper according to the connection type and format specifications of the printing devices, and send print control commands to the printing devices to output the paper exam paper.

2. The method according to claim 1, characterized in that, The large model is invoked to analyze all incorrect questions, extract the knowledge points corresponding to each incorrect question, and analyze the reasons for the errors in each question, including: The question stem and answer content of each incorrect question are extracted and input into the large model. The large model is then matched with the knowledge system database to obtain the knowledge point entries corresponding to each incorrect question. The large model compares the differences between the answer content and the correct answer to determine the type of error cause for each incorrect question.

3. The method according to claim 2, characterized in that, The generation of learning tags corresponding to incorrect questions includes: Count the number of incorrect questions for the same knowledge point, combine the error reason type of each incorrect question to generate the mastery level label and error reason distribution label for the corresponding knowledge point, and summarize all the labels to form the corresponding student's learning status label set.

4. The method according to claim 1, characterized in that, The process of formatting multiple test questions into an electronic exam paper according to a preset format includes: All test questions are categorized and sorted according to question type. A blank answer area is set for each question. The spacing between questions and the font size are adjusted to match the size parameters of standard paper, and an electronic test paper file that conforms to printing specifications is generated.

5. The method according to claim 1, characterized in that, The process of obtaining the connection type and format specifications of connectable printing devices, and adjusting the output format of the electronic exam paper according to the connection type and format specifications of the printing devices, includes: Scan all currently connectable printing devices, read the connection method and supported paper sizes and resolution parameters of each printing device, adjust the page size, resolution and layout of the electronic exam paper according to the parameters of the selected printing device, and generate an output file adapted to the selected printing device.

6. The method according to claim 1, characterized in that, The method further includes: Images of completed paper test papers by students are collected, and the test paper images are identified and graded to obtain new incorrect questions. The overlap of knowledge points between the new incorrect questions and the original incorrect questions is compared, and the corresponding student's learning status tag set is updated.

7. The method according to claim 1, characterized in that, The method further includes: When no directly connectable printing device is detected, the electronic test paper file and print trigger command are sent to the pre-bound mobile terminal, which then connects to the corresponding printing device and executes the test paper printing operation.

8. The method according to claim 1, characterized in that, The method further includes: After identifying and marking all incorrect questions, the large model is invoked to generate explanation content for the knowledge points corresponding to each incorrect question. The explanation content includes, in turn, the core concept of the knowledge point, the location of the error cause of the corresponding incorrect question, the standard problem-solving ideas and steps, and the summary of general problem-solving techniques for similar question types. The learning device displays the explanation of each incorrect question through its display module. After each incorrect question is explained, a mastery level feedback option is displayed to receive mastery level feedback submitted by the student. The mastery level feedback is divided into three levels: complete understanding, basic understanding, and still have questions. By combining student mastery feedback with existing learning tags, the number and difficulty of subsequent test questions can be adjusted accordingly. For knowledge points that are still unclear based on feedback, the number of corresponding test questions will be increased, while the difficulty of the questions will be reduced and the proportion of basic questions will be increased. For knowledge points that are fully understood based on feedback, the number of corresponding test questions will be reduced, while the difficulty of the questions will be increased and the proportion of extended question types will be increased.

Images