An automatic homework grading system

Abstract

The application discloses a kind of automatic correction system of job;Including handwriting collection module, OCR identification module, wrong character correction module and intelligent scoring module;The handwriting collection module carries out collection to handwriting;The handwriting is identified to the OCR identification module;The wrong character correction module corrects wrong character;The intelligent scoring module is used for comprehensive consideration;The application is by collecting test paper handwriting, quickly it is converted into text data, CR identification module can accurately identify the writing content of test paper, even in the face of different writing style and handwriting characteristics, it can accurately be converted into text, wrong character correction module can ensure the text accuracy of examinee answer;Intelligent scoring module combines question requirement, standard problem solving process and examinee answering process, in-depth analysis is carried out by LLM large language model, can give objective, accurate score and targeted comment, improve the accuracy and reliability of examination review.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of educational technology, more specifically, it relates to an automatic homework grading system. BACKGROUND

[0002] In the modern educational environment, with the advancement of digital teaching, the way of homework grading for students needs to be reformed. Currently, there are many problems in the process of handwriting recognition and text conversion. Taking some existing homework grading software as an example, in the aspect of handwriting collection, due to the complexity of handwriting (such as messy writing, connected writing, etc.), the recognition accuracy is seriously affected. The recognition system cannot accurately grasp the shape, direction and connection method of strokes and other key information. For some special writing tools (such as colored pens, highlighters, etc.), the recognition effect is even worse. The color and fluorescent effect of these special handwriting can interfere with the normal work of the recognition system, making it unable to accurately extract the features of the handwriting.

[0003] In the aspect of text correction, although some existing software can recognize some wrong characters, the correction effect of some complex wrong characters (such as similar characters, homophonic characters, etc.) is not good. For some easily confused characters such as "yi" and "ji", "de" and "di", it may not be able to accurately correct them. Moreover, for some professional terms in specific fields, it may not be able to correctly identify and correct errors due to the lack of corresponding vocabulary support.

[0004] The existing intelligent scoring system often has the problem of inaccurate scoring. Some systems may only score according to the correctness of the answers without considering the student's problem-solving approach, steps and methods used. For example, in a math homework, a student may have the correct answer, but the problem-solving process is not standardized or the wrong method is used, and the existing system may still give full marks, which is not conducive to the student's learning and growth.

[0005] The present application mainly aims at the deficiencies in the existing homework grading system and solves the following key problems:

[0006] Inaccurate handwriting recognition: whether it is complex handwriting written by students (such as messy writing, connected writing, etc.) or handwriting left by special writing tools (such as colored pens, highlighters, etc.), the current system has low accuracy in recognizing these handwriting and converting them into text content.

[0007] Weak text correction ability: the existing software has defects in text correction, and the correction effect of some complex wrong characters (such as similar characters, homophonic characters, etc.) and professional terms in specific fields is not good, which cannot guarantee the accuracy of the text content.

[0008] Intelligent scoring is unreasonable: the existing intelligent scoring system often only scores according to the correctness of the answer, ignores the problem solving ideas, steps and methods of students, and the unreasonable scoring method is not conducive to the learning and growth of students, so an automatic homework correction system is proposed. SUMMARY

[0009] In view of the problems existing in the prior art, the purpose of the present application is to provide an automatic homework correction system for improving the recognition accuracy of handwriting, enhancing the text correction capability, comprehensively considering the requirements of the questions, the standard problem solving process and the problem solving process of students, and giving more reasonable and accurate scoring and comments.

[0010] To achieve the above purpose, the present application provides the following technical scheme: an automatic homework correction system, comprising a handwriting collection module, an OCR recognition module, a wrong character correction module and an intelligent scoring module.

[0011] The handwriting collection module is used to collect the handwriting in the writing process, and the handwriting collection module records the pen tip movement path through the optical tracking mode of the miniature camera and records the pen tip pressure change through the strain gauge sensor; various characteristics of the handwriting are accurately recorded, including the position, force, and writing speed information of the strokes;

[0012] The OCR recognition module recognizes the handwriting collected by the handwriting collection module and outputs computer-processed and recognized text, and performs multi-dimensional analysis on the handwriting through recognition algorithms and technologies; first, the morphology of the strokes is analyzed, including the length, thickness, and bending degree of the strokes; second, the direction of the strokes is determined; and finally, the connection mode of the strokes is judged, whether it is connected, intersected or disconnected;

[0013] The wrong character correction module is used to correct the wrong characters in the recognized text output by the OCR recognition module; a rich word bank resource is used, including a general word bank and a specific field professional term word bank, for complex wrong characters, a combination of multiple recognition methods is used for correction, and the accuracy of text correction is improved through shape comparison and pronunciation comparison methods;

[0014] The intelligent scoring module comprehensively considers the text information output by the OCR recognition module in combination with the requirements of the questions, the standard problem solving process and the problem solving process of students, realizes the scoring and comments of the homework, establishes a comprehensive intelligent scoring model, considers the multiple factors of the requirements of the questions, the standard problem solving process and the problem solving process of students, analyzes the problem solving ideas, steps and methods of students through the correctness of the answers in the scoring process, gives more accurate evaluation to the students, and marks the error contents.

[0015] Specifically, the handwriting collection module captures the pen tip pressure, motion trajectory and writing speed of the handwriting in real time through the graphic code, and converts the features into digital signals to form handwriting data.

[0016] The graphic code is designed as a specific graphic code mode through optical marking, and is composed of micro patterns in different shapes and directions. The patterns are arranged on the paper surface in a certain rule to form a two-dimensional coordinate grid, and each pattern has unique code information for determining the pen tip pressure, motion trajectory and writing speed.

[0017] Specifically, the micro camera is arranged near the pen tip and captures the paper surface at a fixed frame rate of 30 frames per second, and the obtained image contains the paper area around the pen tip and the corresponding code pattern.

[0018] The collected image is preprocessed. The collected image is first subjected to grayscale processing to convert the color image into a grayscale image, and then subjected to noise reduction processing. The median filtering method is used to remove noise points in the image and improve the image quality.

[0019] Specifically, the optical tracking mode is used for marker recognition and coordinate calculation. A template matching algorithm is used to identify the code pattern in the image. For each collected image, a pre-designed code pattern template is matched with each region in the image. When the matching degree reaches 80%, the code pattern position is determined. According to the pre-designed coordinate rule of the code pattern on the paper, the coordinate position of the pen tip relative to the paper is calculated.

[0020] Specifically, the OCR recognition module processes the handwriting picture collected by the handwriting collection module, including removing noise, enhancing contrast and adjusting brightness.

[0021] The handwriting stroke form analysis algorithm, the handwriting stroke direction analysis algorithm and the handwriting stroke connection mode analysis algorithm are used to analyze the form of the strokes.

[0022] The handwriting stroke form analysis algorithm uses the Canny edge detection algorithm. First, the gradient amplitude and direction of the calculated image are determined by finding the local maximum value of the gradient amplitude to determine the edge points. Then, the non-maximum suppression is performed on the edge points to remove points that are not edges, and the edges are further refined. Finally, through double threshold processing, the real edge contour is determined.

[0023] The stroke direction analysis algorithm adopts the method of directional derivative. Firstly, gradient calculation is performed on the handwriting image to obtain the gradient direction and gradient value of each point in the image. Then, the stroke direction is determined according to the gradient direction of each point on the stroke edge contour. If the gradient direction points to the left, the stroke direction is to the left. If the gradient direction points upwards, the stroke direction is upwards. Alternatively, a method based on skeleton extraction is adopted. Firstly, skeleton extraction is performed on the handwriting image to obtain the skeleton structure of the stroke. Then, the stroke direction is determined according to the direction of the skeleton structure. If the skeleton structure extends from left to right, the stroke direction is to the right.

[0024] The stroke connection mode analysis algorithm adopts the connected domain analysis method. Firstly, binaryzation processing is performed on the handwriting image to distinguish the stroke from the background. Then, the number and shape of the connected domains are calculated to determine the connection mode of the stroke. If the number of connected domains is 1 and the shape is a continuous curve, the stroke is in the connected mode. If the number of connected domains is 2 and the shape is two intersecting curves, the stroke is in the intersecting mode. If the number of connected domains is 3 and the shape is three disconnected curves, the stroke is in the disconnected mode. Alternatively, a method based on graph theory is adopted. The stroke is regarded as a node in the graph, and the connection relationship between the strokes is regarded as an edge in the graph. By analyzing the connectivity and structure of the graph, the connection mode of the stroke is determined. If the graph is connected and all nodes are connected through edges, the stroke is in the connected mode. If there are two nodes intersecting through edges in the graph, the stroke is in the intersecting mode. If there are three nodes disconnected through edges in the graph, the stroke is in the disconnected mode.

[0025] Specifically, the wrong word correction module is used to comprehensively check the recognized text content, compare common words with a general word library, compare specific field professional terms with a corresponding professional term library, correct complex wrong words such as homophonic words by combining multiple recognition methods, and determine whether a homophonic word is a wrong word by comparing the shape and stroke order.

[0026] Specifically, the intelligent scoring module corrects the homework through the LLM correction module.

[0027] Firstly, the key elements and correct step determination algorithm: deep semantic analysis is performed on the original question text, and a pre-trained language model in natural language processing technology is used to identify key concepts, conditions and requirements in the question; for mathematical questions, the mathematical concepts, known conditions and solving requirements involved are accurately identified;

[0028] Constructing a knowledge graph based on question types and subject fields: the knowledge graph contains common question type structures, key knowledge points, problem solving steps and logical relationships;

[0029] Match the standard problem-solving process with the knowledge graph to determine the key elements and correct steps for answering questions; ensure that key information and steps can be accurately extracted in different question types and subject areas.

[0030] Specifically, the intelligent scoring module analyzes the student's problem-solving process, cleans the text of the student's problem-solving process, removes irrelevant punctuation, spaces and special characters, and makes the text more standardized; then, a word vector model is used to map the vocabulary in the text to a low-dimensional vector space to obtain the feature representation of the text.

[0031] Specifically, the intelligent scoring module implements scoring and commenting on assignments:

[0032] Scoring rules: Different weights are assigned to key steps and ideas based on the question type and subject area. For questions with a strong theoretical focus, the rationality of the solution is emphasized, with a 60% weight given to the rationality of the idea and a 40% weight given to the key steps. For questions with a strong computational focus, the correctness of the key steps is emphasized, with a 70% weight given to the key steps and a 30% weight given to the rationality of the idea.

[0033] Scoring is based on how closely the student's problem-solving process matches the standard problem-solving process: if all key steps are correct, but there are some minor flaws in the problem-solving process, a certain percentage of points will be deducted, 10%-20% of the total score;

[0034] If the student's answer is correct but the solution is not standardized, points will be deducted according to the degree of non-standardization and the degree of impact on the result; if steps are missing or the order is wrong during the solution, a corresponding proportion of points will be deducted according to the severity of the omission or error. Missing an important step will result in a deduction of 30% of the total score, and incorrect order of steps will result in a deduction of 15% of the total score.

[0035] Comment generation: Generate targeted comment content based on the strengths and weaknesses of students in the problem-solving process.

[0036] Specifically, the steps of error marking are as follows:

[0037] Rule-based recognition: Divide the questions into regions, obtain question region information and incorrect solution process information from the LLM correction module, determine whether the text region or image region corresponds to the incorrect solution process, and if the solution steps in a certain question region do not conform to the standard solution process, it will be identified as an incorrect region;

[0038] For some specific error types, such as formula errors and logic errors, corresponding rules are set for identification. If a mathematical formula contains symbols or structures that do not conform to mathematical operation rules, the area containing the formula is identified as an error area.

[0039] Annotation algorithm: annotation on the corresponding text display area or original picture; for text area, use specific text annotation format;

[0040] For the original picture, use a graphics drawing algorithm to draw a mark in the error area; use a common graphics drawing library to draw a circle on the picture to mark the error area, ensuring that the position and size of the mark accurately correspond to the actual position and range of the error area.

[0041] Technical effects and advantages of the present application:

[0042] The present application can greatly improve the efficiency and accuracy of examination review, by collecting the handwriting of the test paper and quickly converting it into text data, avoiding the cumbersome process of collecting and organizing traditional paper test papers, the OCR recognition module can accurately recognize the written content of the test paper, even if it faces different writing styles and handwriting characteristics, it can accurately convert it into text, reducing the misjudgment caused by unclear handwriting; the wrong word correction module can ensure the text accuracy of the examinee's answers, avoiding the influence of the fairness of the score caused by wrong words; the intelligent scoring module combines the requirements of the questions, the standard problem solving process and the examinee's answering process, and analyzes in depth through the LLM large language model, which can give objective and accurate scores and targeted comments, improving the accuracy and reliability of the examination review;

[0043] The comment content of the intelligent scoring module can help the examinee understand his performance in the examination and clarify his strengths and weaknesses; for the wrong problem solving process, the red circle annotation function enables the examinee to intuitively see his mistakes, which is convenient for targeted review and improvement; at the same time, teachers can also use data analysis function to develop personalized learning plan and guidance scheme for each examinee according to the examinee's examination situation, improve the learning effect and examination result of the examinee;

[0044] The present application can effectively guarantee the fairness of the examination, and the accurate handwriting recognition and OCR recognition technology can ensure that the answers of all examinees can be accurately recognized and reviewed, avoiding the influence of human factors on the review result; the standardized processing flow of the wrong word correction and intelligent scoring system makes the scoring more objective and fair, reducing the interference of subjective factors; at the same time, the examination data can also be encrypted and safely managed to ensure the confidentiality and security of the examination result;

[0045] The application of the present application provides strong support for the development of examination informatization; through the intelligent examination review system, digital management and analysis of examination data can be realized, more comprehensive and accurate examination information and data analysis reports are provided for the education department and the school; the education department and the school are helped to understand the learning situation and the teaching effect of students, and scientific basis is provided for teaching reform and education decision; meanwhile, more convenient and efficient examination experience is provided for the examinee, and the development trend of modern education and examination is adapted.

[0046] Other features and advantages of the present application will become apparent from the following detailed description of exemplary embodiments thereof, taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0047] Fig. 1 It is the system structure schematic diagram provided by the present application.

[0048] Fig. 2 It is the working principle diagram of the automatic correction of homework provided by the present application.

[0049] Fig. 3 It is the flow chart of the automatic correction of homework provided by the present application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0051] As shown in Figs. 1-3 , the automatic homework correction system provided by the present application comprises a handwriting collection module, an OCR recognition module, a wrong character correction module and an intelligent scoring module;

[0052] The handwriting collection module is used to collect handwriting in the writing process, the handwriting collection module records the pen tip moving path through the optical tracking mode of a miniature camera, and records the pen tip pressure change through a strain gauge sensor; various characteristics of handwriting are accurately recorded, including the position, force, writing speed information of strokes;

[0053] The OCR recognition module recognizes the handwriting collected by the handwriting collection module, outputs computer capable of processing and recognizing text, and performs multi-dimensional analysis on handwriting through recognition algorithms and technologies; first, the shape of strokes is analyzed, including the length, thickness and bending degree of strokes; second, the direction of strokes is determined; finally, the connection mode of strokes is judged, whether connected, intersected or disconnected;

[0054] The misspelling correction module is used for correcting misspelling in the recognized text output by the OCR recognition module; a rich vocabulary resource is utilized, including a general vocabulary and a specific field professional term vocabulary, for complex misspelling, a combination of multiple recognition methods is adopted for correction, and the accuracy of text correction is improved through shape comparison and pronunciation comparison methods;

[0055] The intelligent scoring module comprehensively considers the text information output by the OCR recognition module, the question requirements, the standard problem solving process and the student problem solving process to realize scoring and commenting on the homework; a comprehensive intelligent scoring model is established, considering various factors of the question requirements, the standard problem solving process and the student problem solving process, in the scoring process, the student's problem solving thought, step and method are analyzed through the correctness of the answer, to give the student a more accurate evaluation, and the error content is marked with error.

[0056] In this embodiment, preferably, the handwriting collection module captures the pen tip pressure, motion trajectory and writing speed of the handwriting in real time through the graphic code, and converts the features into digital signals to form handwriting data;

[0057] The graphic code is designed as a specific graphic code mode through optical marking, a micro pattern composed of lines with different shapes and directions, and the patterns are arranged on the paper surface in a certain rule to form a two-dimensional coordinate grid, each pattern has unique code information for determining the pen tip pressure, motion trajectory and writing speed;

[0058] It should be noted that through the formed two-dimensional coordinate grid, each pattern has unique code information, which can be used to determine its position on the paper, so as to facilitate the determination of the pen tip pressure, motion trajectory and writing speed.

[0059] In this embodiment, preferably, the miniature camera is arranged near the pen tip, and the paper surface is photographed through a fixed frame rate of 30 frames per second, and the obtained image contains the paper area around the pen tip and the corresponding code pattern;

[0060] The collected image is preprocessed, the collected image is first grayed, the color image is converted into a gray image, and then noise reduction processing is performed, the median filtering method is adopted to remove noise points in the image, and the image quality is improved;

[0061] It should be noted that the image obtained by the miniature camera contains the paper area around the pen tip and the corresponding code pattern, and the image is filtered to improve the image quality;

[0062] The steps of median filtering are as follows:

[0063] Selecting neighborhood size: select a window of odd size (e.g. 3x3, 5x5, etc.); this window will be used to determine the new value of each pixel;

[0064] Defining neighborhood: for a pixel p(x, y) in the image, its neighborhood is a region around the pixel, and the set of pixels in the window is denoted as N(x, y);

[0065] Sorting neighborhood pixels: sort all the pixel values in the neighborhood region, and get a sorted pixel value sequence S = [v1, v2,..., vn], where n is the number of pixels in the neighborhood; n

[0066] Selecting median: the median is the middle value in the sorted sequence, denoted as v median , if the neighborhood size is odd, the median is the middle element in the sequence; the formula is:

[0067]

[0068] where n is the number of pixels in the window, usually n = k x k, where k is the size of the window (usually odd).

[0069] Replacing pixel value: replace the value of the original pixel p(x, y) with the median v median .

[0070] In this embodiment, preferably, the optical tracking method is used for marker recognition and coordinate calculation, a template matching algorithm is used to recognize the coded pattern in the image, for each captured image, a pre-designed coded pattern template is matched with each region in the image, when the matching degree reaches 80%, it is determined that the coded pattern is located; according to the pre-designed coordinate rule of the coded pattern on the paper, the coordinate position of the pen tip relative to the paper is calculated;

[0071] It should be noted that according to the pre-designed coordinate rule of the coded pattern on the paper, the coordinate position of the pen tip relative to the paper is calculated, for example, if the coordinate of a coded pattern on the paper is (x0, y0), and the position of the coded pattern in the image is determined through the template matching algorithm;

[0072] For example, the position in the image is (x1, y1), then the coordinate (x, y) of the pen tip can be calculated by the relative position relationship, assuming that the distance between the camera and the paper is d, and the focal length of the camera is f, then the coordinate (x, y) of the pen tip can be calculated by the similar triangle principle: x = x1*(d / f) + x0, y = y1*(d / f) + y0. By continuously recording the coordinate position of the pen tip, the moving path of the pen tip on the paper can be obtained.

[0073] ​In the embodiment, the OCR recognition module processes the handwriting picture collected by the handwriting collection module, including removing noise, enhancing contrast and adjusting brightness;

[0074] The OCR recognition module analyzes the shape of the stroke, including stroke shape analysis algorithm, stroke direction analysis algorithm and stroke connection mode analysis algorithm;

[0075] The stroke shape analysis algorithm adopts Canny edge detection algorithm, first calculates the gradient amplitude and direction of the image, determines the edge point by finding the local maximum value of the gradient amplitude; then performs non-maximum suppression on the edge point to remove points that are not edges, and further refines the edge; finally, through double threshold processing, the real edge contour is determined;

[0076] By analyzing the curvature, length, thickness and other characteristics of the edge contour, the shape of the stroke can be determined. Specifically, by calculating the pitch radius of each point on the edge contour, the degree of curvature of the stroke can be determined; by counting the number of pixels of the edge contour, the thickness of the stroke can be obtained; by measuring the length of the edge contour, the shape of the stroke can be determined. On this basis, methods based on shape descriptors can also be used, such as Hu moments, which can quantitatively describe the shape of the stroke. By calculating the value of Hu moments, the shape characteristics of the stroke can be obtained, such as symmetry and flatness, thereby assisting in determining the shape of the stroke;

[0077] The stroke direction analysis algorithm adopts the method of directional derivative, first calculates the gradient of the handwriting image to obtain the gradient direction and gradient value of each point in the image; then, according to the gradient direction of each point on the stroke edge contour, the direction of the stroke is determined; if the gradient direction points to the left, the stroke direction is to the left; if the gradient direction points upwards, the stroke direction is upwards; or a method based on skeleton extraction is used, first the skeleton of the handwriting image is extracted to obtain the skeleton structure of the stroke; then, according to the direction of the skeleton structure, the direction of the stroke is determined; if the skeleton structure extends from left to right, the stroke direction is to the right;

[0078] The stroke connection mode analysis algorithm adopts a connected domain analysis method, first performs binaryzation processing on the handwriting image to distinguish the strokes from the background; then, the number and shape of the connected domains are calculated to determine the stroke connection mode; if the number of connected domains is 1 and the shape is a continuous curve, the stroke is in the connected mode; if the number of connected domains is 2 and the shape is two intersecting curves, the stroke is in the intersecting mode; if the number of connected domains is 3 and the shape is three disconnected curves, the stroke is in the disconnected mode; or a graph theory-based method is adopted, the strokes are regarded as nodes in the graph, and the connection relationship between the strokes is regarded as edges in the graph, the connectedness and structure of the graph are analyzed to determine the stroke connection mode; if the graph is connected and all nodes are connected through edges, the stroke is in the connected mode; if there are two nodes intersecting through edges in the graph, the stroke is in the intersecting mode; if there are three nodes disconnected through edges in the graph, the stroke is in the disconnected mode;

[0079] It should be noted that the above algorithm is used to calculate the stroke direction and connection mode, so as to analyze and recognize the written characters through handwriting.

[0080] The calculation formula for removing noise is as follows:

[0081] ,

[0082] Among them, represents the original data corresponding to the time point t in the data information, represents the smooth value corresponding to the time point in the data information, is a smoothing factor, which is used to allocate the weights of the new data point and the historical data point, and represents the transformation structure of the data information in scale and position , is the conjugate of wavelet transform, is a scale parameter, is a translation parameter.

[0083] In this embodiment, preferably, the wrong word correction module is used to comprehensively check the recognized text content, compare common words with a general word library, compare specific field professional terms with a corresponding professional term library, correct complex wrong words such as homophonic words by combining multiple recognition methods, and judge whether the homophonic words are wrong words by comparing the shapes, i.e., comparing the external structure and stroke order; and judge whether the homophonic words are wrong words by comparing the pronunciation, i.e., comparing the reading.

[0084] It should be noted that the analysis and identification of similar characters and similar characters facilitate the determination and correction of errors in the text, and the errors in the text are not identified.

[0085] In this embodiment, preferably, the intelligent scoring module corrects the homework through the LLM correction module.

[0086] First, the key elements and correct steps determination algorithm: deep semantic analysis of the original question text, using the pre-trained language model in natural language processing technology to identify key concepts, conditions and requirements in the question; for mathematical problems, accurately identify the mathematical concepts involved, known conditions and solving requirements;

[0087] Construct a knowledge graph based on the type of question and the field of study: the knowledge graph contains common question type structures, key knowledge points, problem solving steps and logical relationships;

[0088] Match the standard problem solving process with the knowledge graph to determine the key elements and correct steps of the answer; ensure that the key information and steps can be accurately extracted under different question types and subject areas;

[0089] It should be noted that the original question text is analyzed by the underlying technology principle of BERT and GPT series to identify key concepts, conditions and requirements in the question. For mathematical problems, accurately identify the mathematical concepts involved, such as functions and geometric figures; known conditions, such as numerical values, side lengths of figures; and solving requirements, such as area, proof of equality, etc.

[0090] The knowledge graph contains common question type structures, key knowledge points, problem solving steps and their logical relationships; for the solution of the quadratic equation, the knowledge graph contains the definition of the equation, the calculation of the discriminant, the application of the root formula, and other key knowledge points and their order;

[0091] Match the standard problem solving process with the knowledge graph to facilitate the extraction of key elements and correct steps of the answer, and facilitate the analysis and processing of the problem.

[0092] In this embodiment, preferably, the intelligent scoring module analyzes the student's problem solving process, cleans the text of the student's problem solving process, removes irrelevant punctuation marks, spaces and special characters, and makes the text more standardized; then, the word vector model is used to map the words in the text to a low-dimensional vector space to obtain the feature representation of the text.

[0093] It should be noted that the word vector model of Word2Vec and GloVe is used to map the words in the text to a low-dimensional vector space and obtain the feature representation of the text; the feature representation can capture the semantic relationship between words and provide a basis for subsequent analysis.

[0094] In this embodiment, preferably, the intelligent scoring module realizes scoring and commenting on the homework:

[0095] Scoring rules are established: different weights are set for key steps and ideas according to the type of the question and the subject field; for questions with strong theoretical nature, the rationality of the problem solving idea is emphasized, and the rationality of the idea is given a weight of 60%, while the key step is given a weight of 40%; for questions with strong calculation, the correctness of the key step is emphasized, and the key step is given a weight of 70%, while the rationality of the idea is given a weight of 30%;

[0096] The matching degree of the student's problem solving process and the standard problem solving process is considered for scoring: if the key steps are all correct, but there are some small flaws in the problem solving idea, a certain percentage of the score is deducted, and the total score is deducted by 10-20%;

[0097] If the student's answer is correct but the problem solving process is not standardized, the score is deducted according to the degree of non-standardization and the degree of influence on the result; if the steps are missing or the order is wrong in the problem solving process, the score is deducted according to the severity of the missing or wrong steps, and the total score is deducted by 30% for missing an important step and by 15% for wrong order of steps;

[0098] Comment generation: according to the advantages and disadvantages of the student's problem solving process, targeted comment content is generated;

[0099] It should be noted that through the scoring standard, the homework of the student is precisely scored and processed, which facilitates the automatic correction of the test paper or the homework;

[0100] According to the advantages and disadvantages of the student's problem solving process, targeted comment content is generated; if the student uses a novel method or idea in a certain step, give affirmation and encouragement; for example, "you used [specific method] in this question, which is very creative and worth learning"; if the student has shortcomings, point out the problem and give suggestions for improvement; for example, "you made [specific error description] in the problem solving process, please [improve the suggestion content]"; in this way, targeted feedback is provided for students to promote their learning and growth.

[0101] In this embodiment, preferably, the error marking steps are as follows:

[0102] Rule-based recognition: divide the question into regions, obtain the question region information, and obtain the error solving process information from the LLM correction module, determine the text region or image region corresponding to the error solving process, if the solving steps in a question region do not conform to the standard solving process, it is identified as an error region;

[0103] For some specific error types, such as formula error and logic error, set corresponding rules for recognition; if a mathematical formula contains symbols or structures that do not conform to mathematical operation rules, the region where the formula is located is identified as an error region;

[0104] Labeling algorithm: label on the corresponding text display region or original picture; for text regions, use a specific text labeling format;

[0105] For original pictures, use a graphics drawing algorithm to draw markers in the error region; use a common graphics drawing library to draw circles on the picture to mark the error region, ensuring that the position and size of the label accurately correspond to the actual position and range of the error region;

[0106] It should be noted that by standardizing the error content, the examinee can intuitively see where his or her mistakes are, and can review and improve accordingly. At the same time, teachers can also analyze students' test results;

[0107] For text regions, use a specific text labeling format, add specific symbols before and after the error region, [error start] and [error end]) or color marking (change the text color of the error region to red);

[0108] For original pictures, use a common graphics drawing library (OpenCV) to draw circles (red circles) on the picture to mark the error region.

[0109] Principle of question region division:

[0110] Overall question region determination: in the generation of test papers, for each question, check the starting coordinates (x, y) of the question elements (including question text, graphics, and all elements that constitute the question) on the paper. At the same time, record the width and height of the question elements. Determine the region range of the entire question, with the top-left corner coordinates as (x, y) and the bottom-right corner coordinates as (x+w, y+h);

[0111] Different question type answer region division: multiple choice question / judgment question answer region division:

[0112] Option element positioning:

[0113] For multiple-choice and true-or-false questions (using the format of the subject network's questions as an example, assuming the options are included in the question), first find the option elements in the question; options can be identified through text feature matching or specific markers;

[0114] Answer area annotation: Insert a square box before each option as an answer area marker. The size of this box can be determined according to the pre-set standard size or the proportional relationship with the question text; at the same time, record the coordinate information of each box for subsequent identification and processing;

[0115] Blank question answer area division:

[0116] Specific element positioning: Find the underline and bracket elements in the question, which are common answer hints for fill-in-the-blank questions; if these elements exist in the question, use them as the basis for determining the answer area;

[0117] Answer area style addition and coordinate acquisition: Add specific styles (such as changing colors, bolding, etc.) and markers (such as adding a small icon to indicate the answer position) to the found underline and bracket elements, and then get the coordinates of these elements as the answer area. If there are no underlines and brackets in the question, create a certain number of answer areas (determined according to the number of answers or the requirements of the question) at the bottom of the question. Each answer area can be a rectangular area, and its size and position also need to be reasonably set and the coordinate information recorded.

[0118] Question and answer question answer area division:

[0119] Leave a large blank at the bottom of the question as the answer area. The size of the blank area can be determined comprehensively according to the question type, the estimated number of words, and the page layout. At the same time, record the coordinate information of the blank area to accurately identify and process the correspondence between the answer content and the question area.

[0120] Pen trace collection

[0121] Structure: Specific smart pen cooperates with signal processing unit. Smart pen has built-in pressure sensor, position sensor and speed sensor, etc.

[0122] Function: Accurately collect the physical characteristics of users' writing traces, providing raw data for subsequent processing.

[0123] Principle: When students write, the smart pen uses internal sensors to capture the physical characteristics of the pen trace in real time, such as pen tip pressure, movement trajectory, and writing speed. These characteristics are transmitted to the signal processing unit to be converted into digital signals, forming pen trace data that is transmitted to the OCR recognition. When writing a stroke, the pressure sensor senses the writing force, the position sensor records the pen tip movement path, and the speed sensor monitors the writing speed, which is processed to form the pen trace data of the stroke.

[0124] OCR recognition

[0125] Structure: includes image preprocessing module, multi-dimensional analysis engine, character matching module and coordinate labeling unit.

[0126] Function: Recognize handwriting data and output recognized text with coordinate labels.

[0127] Principle: After receiving handwriting data from handwriting collection, the image preprocessing module first removes noise and enhances contrast to improve handwriting image clarity. The multi-dimensional analysis engine analyzes handwriting form (such as length, thickness, bending degree, etc.), direction (left, right, up or down) and connection mode (connected, intersected or disconnected). The character matching module matches the analysis results with predefined character models, recognizes characters and combines them into text. The coordinate labeling unit labels the coordinates of each line of text to distinguish the topic area. The recognized text and coordinate information are transmitted to the wrong word correction.

[0128] Wrong word correction

[0129] Structure: composed of general word library, professional term library, character shape comparison unit and character sound comparison unit.

[0130] Function: Check and correct wrong words in OCR recognition results.

[0131] Principle: After receiving text from OCR recognition, common words are compared with the general word library, and specific domain professional terms are compared with the professional term library. For complex wrong words such as similar characters and similar sounds, the character shape comparison unit judges whether they are wrong words by comparing the shape structure and stroke order of the characters, and the character sound comparison unit assists the judgment by comparing the pronunciation of the characters. The corrected text is transmitted to intelligent scoring.

[0132] Intelligent scoring:

[0133] Structure: composed of question analysis module, problem solving process evaluation module, scoring point review generation module and interaction interface with LLM (large language model).

[0134] Function: Comprehensive consideration of question requirements, standard problem solving process and student problem solving process to give accurate score and detailed comments.

[0135] Principle: After receiving the corrected answer text from the error correction part, the question analysis module determines the key elements and correct steps of the answer according to the requirements and standard solving process of the question. The solving process evaluation module deeply analyzes whether the student's solving process conforms to the standard and the method is correct, and whether the solving idea is reasonable. Then, the question and the corrected answer text are transmitted into the LLM large language model for deep analysis. The LLM large language model uses its powerful language understanding and processing ability to comprehensively compare the question and the student's answer text, and further excavates the advantages and disadvantages of the student's solving process. The scoring and comment generation module gives more accurate and targeted scoring and comments according to the analysis results of the LLM large language model and the question score. If an incorrect solving process is found, it is marked with a red circle in the original picture or the corresponding text display area, so that the student can clearly see the error and better understand their own answer situation, promoting learning and improvement.

[0136] Finally, it should be noted that the above-described is only the preferred embodiment of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent replacements to some technical features, as long as they are within the spirit and principles of the present application. Any modification, equivalent replacement, improvement, etc. made shall be included in the protection scope of the present application.

Claims

1. An automatic homework grading system, characterized in that: It includes handwriting collection module, OCR recognition module, typo correction module and intelligent scoring module; The handwriting acquisition module is used to collect handwriting during the writing process. The handwriting acquisition module records the movement path of the pen tip through optical tracking using a micro camera and the pressure change of the pen tip using a strain gauge sensor. It accurately records various handwriting characteristics, including stroke position, force, and writing speed information. The OCR recognition module recognizes the handwriting collected by the handwriting collection module and outputs text that can be processed and recognized by the computer. Through recognition algorithms and technologies, the handwriting is analyzed in multiple dimensions. First, the stroke shape is analyzed, including the length, thickness, and curvature of the strokes. Second, the direction of the strokes is determined. Finally, the connection mode of the strokes is determined, whether they are connected, intersecting, or disconnected. The typo correction module is used to correct typos in the recognized text output by the OCR recognition module; it utilizes rich vocabulary resources, including general vocabulary and domain-specific professional terminology vocabulary, and uses a combination of multiple recognition methods to correct complex typos, and improves the accuracy of text correction through glyph comparison and pronunciation comparison methods; The intelligent scoring module comprehensively considers the text information output by the OCR recognition module in combination with the question requirements, the standard problem-solving process, and the student's problem-solving process to achieve scoring and commenting on the homework; establishes a comprehensive intelligent scoring model, taking into account multiple factors such as the question requirements, the standard problem-solving process, and the student's problem-solving process. During the scoring process, the module analyzes the student's problem-solving ideas, steps, and methods used based on the correctness of the answer, giving the student a more accurate evaluation and marking errors. The OCR recognition module processes the handwriting image collected by the handwriting collection module, including removing noise, enhancing contrast and adjusting brightness; The OCR recognition module analyzes the stroke shape including a stroke shape analysis algorithm, a stroke direction analysis algorithm and a stroke connection mode analysis algorithm; The stroke morphology analysis algorithm uses the Canny edge detection algorithm. First, the gradient amplitude and direction of the image are calculated, and the edge points are determined by finding the local maximum of the gradient amplitude. Then, non-maximum suppression is performed on the edge points to remove non-edge points and further refine the edge. Finally, double threshold processing is performed to determine the true edge contour. The stroke direction analysis algorithm adopts the method of directional derivatives. First, the gradient of the handwriting image is calculated to obtain the gradient direction and gradient value of each point in the image. Then, the direction of the stroke is determined based on the gradient direction of each point on the stroke edge contour. If the gradient direction points to the left, the stroke direction is leftward; if the gradient direction points upward, the stroke direction is upward. Alternatively, a skeleton extraction-based method is adopted. First, the skeleton of the handwriting image is extracted to obtain the skeleton structure of the stroke. Then, the direction of the stroke is determined based on the direction of the skeleton structure. If the skeleton structure extends from left to right, the stroke direction is rightward. The stroke connection mode analysis algorithm adopts a connected domain analysis method. First, the handwriting image is binarized to distinguish the strokes from the background. Then, the connection mode of the strokes is determined by calculating the number and shape of connected date domains. If the number of connected domains is 1 and the shape is a continuous curve, the stroke is in a connected mode. If the number of connected domains is 2 and the shape is two intersecting curves, the stroke is in an intersecting mode. If the number of connected domains is 3 and the shape is three disconnected curves, the stroke is in a disconnected mode. Alternatively, a graph theory-based method is adopted, in which the strokes are regarded as nodes in a graph and the connection relationships between the strokes are regarded as edges in the graph. The connection mode of the strokes is determined by analyzing the connectivity and structure of the graph. If the graph is connected and all nodes are connected by edges, the stroke is in a connected mode. If two nodes in the graph intersect by edges, the stroke is in an intersecting mode. If three nodes in the graph are disconnected by edges, the stroke is in a disconnected mode.

2. The automatic homework grading system according to claim 1, characterized in that: The handwriting acquisition module captures the pen tip pressure, motion trajectory and writing speed of the handwriting in real time through graphic coding, and converts the features into digital signals to form handwriting data; The graphic code is designed as a specific graphic coding method through optical marking. It is composed of tiny graphics composed of lines of different shapes and directions. The graphics are arranged on the paper surface according to a certain pattern to form a two-dimensional coordinate grid. Each graphic has unique coding information for determining the pen tip pressure, movement trajectory and writing speed.

3. The automatic homework grading system according to claim 1, characterized in that: The micro camera is placed near the pen tip and captures the paper surface at a fixed frame rate of 30 frames per second. The resulting image includes the paper area around the pen tip and the corresponding coded pattern. The collected images are preprocessed by gray-scaling the collected images first, converting the color images into gray-scale images, and then performing noise reduction processing. The median filtering method is used to remove noise points in the image and improve the image quality.

4. The automatic homework grading system according to claim 1, characterized in that: The optical tracking method is used for mark recognition and coordinate calculation. A template matching algorithm is used to identify the coded pattern in the image. For each captured image, a pre-designed coded pattern template is matched with each area in the image. When the matching degree reaches 80%, the location of the coded pattern is determined; based on the pre-designed coordinate pattern of the coded pattern on the paper, the coordinate position of the pen tip relative to the paper is calculated.

5. The automatic homework grading system according to claim 1, characterized in that: The typo correction module is used to conduct a comprehensive check on the recognized text content, comparing common words with a general vocabulary; comparing professional terms in specific fields with the corresponding professional terminology vocabulary; correcting complex typos, such as similar characters, by combining multiple recognition methods; and determining whether similar characters are typos by comparing the shape of the characters, that is, comparing the appearance structure and stroke order.

6. The automatic homework grading system according to claim 1, characterized in that: The intelligent scoring module corrects the homework through the LLM correction module; First, the algorithm determines the key elements and correct steps: Deep semantic analysis is performed on the original question text, using pre-trained language models in natural language processing technology to identify the key concepts, conditions, and requirements in the question. For math questions, the involved mathematical concepts, known conditions, and solution requirements are accurately identified. Build a knowledge graph based on question types and subject areas: The knowledge graph includes common question structures, key knowledge points, problem-solving steps, and logical relationships; Match the standard problem-solving process with the knowledge graph to determine the key elements and correct steps for answering questions; ensure that key information and steps can be accurately extracted in different question types and subject areas.

7. The automatic homework grading system according to claim 1, characterized in that: The intelligent scoring module analyzes the student's problem-solving process, cleans the text of the student's problem-solving process, removes irrelevant punctuation, spaces and special characters, and makes the text more standardized; then, a word vector model is used to map the words in the text to a low-dimensional vector space to obtain the feature representation of the text.

8. The automatic homework grading system according to claim 1, characterized in that: The intelligent scoring module enables scoring and commenting on assignments: Scoring rules: Different weights are assigned to key steps and ideas based on the question type and subject area. For questions with a strong theoretical focus, the rationality of the solution is emphasized, with a 60% weight given to the rationality of the idea and a 40% weight given to the key steps. For questions with a strong computational focus, the correctness of the key steps is emphasized, with a 70% weight given to the key steps and a 30% weight given to the rationality of the idea. Scoring is based on how closely the student's problem-solving process matches the standard problem-solving process: if all key steps are correct, but there are some minor flaws in the problem-solving process, a certain percentage of points will be deducted, 10%-20% of the total score; If the student's answer is correct but the solution is not standardized, points will be deducted according to the degree of non-standardization and the degree of impact on the result; if steps are missing or the order is wrong during the solution, a corresponding proportion of points will be deducted according to the severity of the omission or error. Missing an important step will result in a deduction of 30% of the total score, and incorrect order of steps will result in a deduction of 15% of the total score. Comment generation: Generate targeted comment content based on the strengths and weaknesses of students in the problem-solving process.

9. The automatic homework grading system according to claim 6, characterized in that: The steps for error marking are as follows: Rule-based recognition: Divide the questions into regions, obtain question region information and incorrect solution process information from the LLM correction module, determine whether the text region or image region corresponds to the incorrect solution process, and if the solution steps in a certain question region do not conform to the standard solution process, it will be identified as an incorrect region; For some specific error types, such as formula errors and logic errors, corresponding rules are set for identification. If a mathematical formula contains symbols or structures that do not conform to mathematical operation rules, the area containing the formula is identified as an error area. Annotation algorithm: Annotate the corresponding text display area or the original image; for the text area, use a specific text annotation format; For the original image, a graphics drawing algorithm is used to draw marks on the error area. A common graphics drawing library is used to draw a circle on the image to mark the error area, ensuring that the position and size of the mark accurately correspond to the actual position and range of the error area.

Images