Double-writing thinking training level evaluation method and system based on Chinese character recognition

Abstract

The invention discloses a double-writing thinking training level evaluation method and system based on Chinese character recognition, and the method comprises the steps: carrying out the scanning and analysis of a handwritten general standard Chinese regular script or a regular script composition through a computer vision technology, and extracting regular script, regular script and regular script font information and text content information through a point location recognition technology of square, round and clock writing grids; according to the method, five double-writing thinking ability portraits are generated, the double-writing thinking level is quantitatively evaluated through analysis and evaluation of the portraits, and according to basic information and evaluation results of students, personalized general regular script or regular script specification Chinese character pattern specification five thinking training suggestions and practice five thinking training and cultivating suggestions are provided. According to the method, automatic evaluation of general standard Chinese character handwriting regular script or running script fonts is achieved, five thinking parameters are introduced, the writing ability can be effectively improved, the method is suitable for the field of intelligent education, the writing and thinking ability of students can be improved, and the method has remarkable education application value.

Description

Technical Field

[0001] The invention relates to a method and system for evaluating the level of double writing thinking training based on Chinese character recognition, which is used for evaluating and training students' double writing thinking ability and belongs to the cross-technical field of education and artificial intelligence. Background Art

[0002] The concept of "double writing" in Chinese education, namely "writing" and "writing", is an important part of students' written expression ability. The two complement each other, and it is particularly important to pay attention to the integration of "double writing". It is the dream of every student to write beautiful characters and brilliant writing, especially in the written examinations of the middle school entrance examination and the college entrance examination, the position of "double writing" ability is irreplaceable. However, the traditional double writing thinking training method mainly relies on the teacher's experience and subjective judgment, lacks objectivity and systematicness, and the existing evaluation method cannot provide personalized training suggestions, which is not conducive to the effective improvement of students' double writing thinking ability. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a method and system for evaluating the level of double writing thinking training based on Chinese character recognition, so as to solve the problems of inaccurate evaluation and lack of personalized guidance in the prior art. The system can help students improve their writing and composition abilities simultaneously through an objective and systematic evaluation system combined with personalized training suggestions, thereby achieving twice the result with half the effort in Chinese learning and realizing the comprehensive development of students' written expression ability.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A method for evaluating the level of double writing thinking training based on Chinese character recognition comprises the following steps:

[0006] S101: Collect and organize the commonly used regular script and semi-cursive Chinese characters in the national unified edition of primary school Chinese textbooks, including 3,500 first-level characters, 3,000 second-level characters and 1,605 third-level characters in the "General Standard Chinese Character Table", a total of 8,105 common standard Chinese characters in regular script and semi-cursive fonts;

[0007] S102: Establish and maintain a square and round font standard five-thinking database containing the five-thinking weights and point data of common regular script and regular script standard Chinese characters;

[0008] S103: Regularly update the five thinking databases of the square and round grid font specifications to keep pace with the latest textbook specifications and thinking weights;

[0009] S104: Use reinforcement learning technology to train the intelligent agent to optimize the five thinking weights;

[0010] S105: Construct a database of five thinking language models for writing exercises;

[0011] S106: Upload photos of students' homework and related basic information;

[0012] S107: Extract the Chinese handwritten font information and text content information from students’ homework, generate five major thinking ability portraits, and analyze and evaluate the thinking level;

[0013] S108: Provide personalized thinking training suggestions based on the analysis results.

[0014] Furthermore, the step S101 specifically includes:

[0015] (1) 3,500 first-level characters, 3,000 second-level characters, and 1,605 third-level characters from the "General Standard Chinese Character List" were selected from the national unified edition of primary school Chinese textbooks to ensure that the representativeness of the selected Chinese characters covers the range of commonly used Chinese characters.

[0016] (2) Classify the selected Chinese characters and sort out the writing samples of regular script and running script respectively to ensure that the samples of each font are standardized and accurate.

[0017] (3) Establish a database containing 8105 commonly used standard Chinese characters, which stores writing samples of regular script and semi-cursive script fonts respectively to facilitate subsequent recognition and analysis.

[0018] Furthermore, the step S102 specifically includes:

[0019] (1) Initialize the database framework, including the national standard ASCII code of standardized Chinese characters, the five thinking weights of standardized Chinese characters, the point data of standardized Chinese characters on the circle line outside the square circle, and the international standard ASCII code of standardized Chinese characters, denoted as Hi;

[0020] (2) Use a structured database management system to store and manage data;

[0021] (3) Define the database table structure, including the standard Chinese character table T H T H , Five thinking weight table T w T W , Standard Chinese characters on the circle outside the square grid data table T p T P ;

[0022] (4) Storing detailed information of the standardized Chinese characters in the standardized Chinese character table, including the national standard ASCII code of the standardized Chinese characters and the source of the standardized Chinese characters in the primary school Chinese textbooks; H T HThe detailed information of the standardized Chinese characters is stored in the Chinese character, including the Chinese character national standard ASCII code Hi and the position index of the Chinese character in the textbook;

[0023] (5) The five major thinking weight arrays mainly trained by the character are stored in the five major thinking weight table, that is, the proportion of the character in intuitive thinking, image thinking, logical thinking, dialectical thinking, and innovative thinking training. The standard Chinese characters appearing in the textbook are divided into several categories, and each category is assigned a weight according to the same label; it is recorded as the weight vector W i for:

[0024] ·W i =[W i1 ,W i2 ,W i3 ,W i4 ,W i5 ]

[0025] Among them, W ij Represents the weight of the i-th Chinese character in the j-th type of thinking training.

[0026] Use a mathematical formula to assign weights. The calculation formula is as follows:

[0027]

[0028] Among them, f ij is the frequency of the ith Chinese character in the jth type of thinking training.

[0029] (6) The coordinates of the starting and ending points of the strokes near the outer circle line of the square circle grid are stored in the point position data table of the standardized Chinese characters on the outer circle line of the square circle grid. The point position data of the standardized Chinese characters on the outer circle line of the square circle grid are recorded as the coordinate set Pi:

[0030] P i ={(x ik ,,y ik ))|k=1,2...,n}

[0031] (7) The standard Chinese character table, the five thinking weight table, and the standard Chinese character point data table on the circle line outside the square circle are linked together, and T is converted into H , T W and T P are interrelated to form a complete database architecture; assuming that T H The primary key of the table is H i d (unique identifier of standardized Chinese characters), T W Table by H i d and T H Table association, where H i d in T W As a foreign key in T P The table is also passed through H id and T H Table association, H i d in T P It is also a foreign key;

[0032] (8) Maintain and update the database to reflect the latest teaching material standards and thinking weights;

[0033] (9) Output the constructed five-thinking database framework of square and round grid font specifications.

[0034] Furthermore, the step S103 specifically includes:

[0035] (1) Regularly review the Chinese character samples and the five thinking weights in the database to ensure consistency with the latest textbook specifications and teaching requirements;

[0036] (2) Update the database structure, if necessary, to accommodate new data requirements and analysis needs;

[0037] (3) Record database update operations to track change history and ensure data consistency.

[0038] Furthermore, the step S104 uses reinforcement learning technology to train the intelligent agent to optimize the five thinking weights. The specific calculation method and steps include:

[0039] (1) Define the objective function:

[0040] The objective function J(θ) is expressed as:

[0041]

[0042] Where N is the number of samples in the training data, s i and a i is the state and action of the i-th sample, R(s i ,a i ) is the reward of the sample, Q(s i ,a i ,θ) is the Q value predicted by the agent, and θ is the agent parameter;

[0043] (2) Constructing a Q-network: Use a neural network to approximate the Q function, which takes the state and action as input and outputs the Q value;

[0044] (3) Establishing an experience replay buffer: During training, the observed state-action-reward-next-state pairs are stored in the experience replay buffer;

[0045] (4) Strategy selection: In the actual environment, the agent uses a strategy π(a|s) to select actions. This strategy can be greedy or ε-greedy.

[0046] (5) Training the Q network: Use stochastic gradient descent (SGD) to update the parameters θ of the Q network. In each iteration, a batch of samples are randomly selected and the target Q value Q′(s, a, θ′) is calculated, where θ′ is the parameter of the target network. The target network parameters are periodically copied from the Q network parameters. The target Q value is calculated as:

[0047]

[0048] Where γ is the discount factor, s′ is the next state, and the gradient for updating the Q network parameters is:

[0049]

[0050] (6) Weight update: Based on the output of the Q network, the agent adjusts the weights of the five major thoughts to minimize the objective function. The specific steps of weight update are:

[0051]

[0052] Among them, α w is the learning rate for weight updates.

[0053] Through this training process, the agent is able to learn how to adjust the five major thinking weights to minimize the objective function, thereby more accurately evaluating students' bilingual thinking ability.

[0054] Furthermore, the step S105 specifically includes:

[0055] (1) Collection of theoretical content and requirements: Collect the theoretical content and requirements of composition in the unified edition of primary school Chinese textbooks, including guiding principles in composition structure, content, expression, etc., to form a complete theoretical library as a benchmark for model training and evaluation.

[0056] (2) Student composition collection: Collect a certain amount of student composition examples. These compositions should cover students of different grades (grades 1-6) and levels to ensure the generalization ability of the model. The compositions can be actual classroom assignments, composition competition works, etc. to ensure the diversity and representativeness of the compositions.

[0057] (3) Collecting correction comments: Collect corresponding correction comments, which come from the teacher’s detailed comments and modification suggestions. Correction comments should cover the identification, judgment and modification suggestions of the five major thinking methods to facilitate subsequent model training and evaluation.

[0058] (4) Corpus construction: The collected writing examples and correction comments are digitized and converted into a corpus that can be used for machine learning model training.

[0059] (5) Model design: A neural network-based five-thinking language model is used, which can identify and analyze the five thinking features in the composition. The model structure includes an encoder (Transformer) and a decoder (Attention mechanism) to process the composition text and extract thinking features. Specifically, the model adopts a deep neural network structure, which contains multiple layers of neurons.

[0060] Input layer: Receives composition text data, converts it into a digital sequence, and converts the composition text into a fixed-length vector sequence through word embedding.

[0061] Encoding layer: Use the encoder (Transformer) to encode the input sequence and extract the semantic features of the composition.

[0062] Decoding layer: Use the decoder (Attention mechanism) to decode the features output by the encoding layer and further extract the five major thinking features.

[0063] Output layer: Maps the features output by the decoding layer into scores of the five major thinking features (a rating scale of 1-5, where 1 represents the worst performance and 5 represents the best performance), indicating the performance of the composition in the five major thinking aspects.

[0064] (6) Model training: The model is trained using the collected corpus and the model parameters are updated through the back propagation algorithm. During the training process, the review comments are used as the supervision signal of the model to improve the accuracy of the model in identifying and analyzing the five major thinking. During the training process, the cross entropy loss function method is adopted. During the training process, the calculation formula of the cross entropy loss function is:

[0065]

[0066] Where L is the loss function, y i is the true label, p i is the probability predicted by the model.

[0067] (7) Database construction and maintenance: Store the trained model parameters, scoring rules, and related theoretical content to build a database of the five major thinking and language models of writing. Update the database to incorporate new teaching theories and student writing data to maintain the timeliness and accuracy of the database.

[0068] Furthermore, the step S106 specifically includes:

[0069] (1) Students first write Chinese characters in the square and round hour grid according to the standard points of regular script and running regular script. After repeated practice, they develop intuitive thinking and remember the corresponding points. Finally, they break away from the square and round hour grid and can find the points of the characters on the grid, thus achieving standard writing of points and shapes with "no grid in the eyes but a grid in the heart";

[0070] (2) Students should learn to scientifically classify the “radicals” of Chinese characters in the standard regular script and running script. They only need to remember the positions of the 201 main radicals on the circular lines outside the square and circle clock grids, and learn the 38 conventional strokes, 16 stroke orders, and 52 structural writing rules. They can then apply what they have learned to other situations and assemble and combine the components of Chinese characters to write the standard regular script and running script characters.

[0071] (3) Students log into the system and fill in their personal information and writing experience, which will be used for subsequent analysis of the five thinking training levels and weak links of their writing;

[0072] (4) Students upload photos of their essays to the system. The photos should clearly show the full text of the essay to ensure the system's recognition accuracy and evaluation quality.

[0073] Furthermore, the step S107 specifically includes:

[0074] (1) Image preprocessing: Receive the homework photos uploaded by students and perform a series of preprocessing operations on them, including grayscale conversion, binarization, denoising, etc., to improve the effect of subsequent processing.

[0075] It includes the following sub-steps:

[0076] Image grayscale conversion: convert the color images uploaded by students into grayscale images, and use the following formula to represent the grayscale value of each pixel:

[0077] I gray (x,y)=0.2989R(x,y)+0.5870G(x,y)+0.1140B(x,y)

[0078] Among them I gray is the pixel value of the grayscale image at position (x, y), R, G, B are the pixel values ​​of the red, green, and blue channels of the original color image respectively;

[0079] Image binarization: Convert the grayscale image into a binary image containing only black and white colors, using the threshold segmentation method. The formula is as follows:

[0080]

[0081] Among them, B(x,y) is the pixel value after binary, G(x,y) is the pixel value of the grayscale image, and T is the set threshold;

[0082] Image denoising: Gaussian filtering algorithm is used to reduce image noise. The formula is as follows:

[0083]

[0084] Where I′(x,y) is the pixel value after denoising, I(x,y) is the pixel value of the original image, w(i,j) is the weight coefficient of the Gaussian kernel, and k is the size of the filter kernel;

[0085] Use the Canny edge detection algorithm to identify edges in the image: Use the Canny algorithm to identify edges in the image. The algorithm includes steps such as Gaussian filtering, gradient calculation, non-maximum suppression, and double threshold edge connection.

[0086] Use Hough Transform to detect straight lines in the image and identify the boundaries of the grid. The basic formula is:

[0087] r=xcos(θ)+ysin(θ)

[0088] Where (r,θ) are the parameters in Hough space, and (x,y) are the coordinates of the edge point in image space.

[0089] By detecting the intersection of the straight lines, the vertex coordinates of each grid are determined, and the image is divided into individual grids according to the vertex coordinates for subsequent Chinese character recognition and analysis. Assume that the equations of the two straight lines are:

[0090] Line L1: Ax+By+C1=0

[0091] Straight line L2: A′x+B′y+C2=0

[0092] Then their intersection point (x, y) can be obtained by solving the following system of equations:

[0093]

[0094] According to these vertex coordinates, the image is segmented into individual grids for subsequent Chinese character recognition and analysis.

[0095] (2) Chinese character recognition: Use optical character recognition (OCR) technology to perform Chinese character recognition on the preprocessed image and extract the handwritten font information of Chinese characters. The recognized Chinese characters are represented as strings for subsequent analysis. The OCR recognition process can be expressed as a function f OCR , which will preprocess the image I pre Mapped to the recognized Chinese character string S, the calculation formula is as follows: S = f OCR (I pre ).

[0096] (3) Text content analysis: Segment and tag the recognized Chinese characters to obtain information about the text content. Use natural language processing (NLP) technology to perform semantic analysis on the composition content and extract key information. Among them: Segmentation and part-of-speech tagging can be performed using function f seg and f tag Indicates that the string S is operated respectively:

[0097] T=f seg (S)

[0098] P = f tag (T)

[0099] Where T is the word sequence after word segmentation, and P is the corresponding part-of-speech tag sequence.

[0100] (4) Extraction of the five major thinking features: Based on the definition of the five major thinking features, design corresponding feature extraction algorithms, including logical thinking features, innovative thinking features, etc. These features can be extracted based on the results of text content analysis, where feature extraction is represented by a series of functions f feat,i , each function corresponds to a thinking characteristic F i :

[0101] F i =f feat,i (P)

[0102] Where i represents the i-th thinking characteristic.

[0103] (5) Thinking ability portrait generation: The thinking features extracted from the text and image are integrated to generate the student's thinking ability portrait, which is represented by a score. The generation of the thinking ability portrait can be represented by the function f profile , which combines all the features F i :

[0104] profile=f profile (F1,F2,…,F n )

[0105] Where n is the total number of thinking characteristics.

[0106] (6) Analysis and evaluation of thinking level: Based on the thinking ability portrait, the scoring model is used to analyze the students’ thinking level. The evaluation results give the students’ overall performance in the five major thinking aspects, as well as the specific scores for each thinking dimension. The scoring model for thinking level is expressed as function f score , generate a score S based on the thinking ability profile total :

[0107] S total =f score (profile)

[0108] The score for each thinking dimension is expressed as:

[0109] S i =f score,i (F i )

[0110] Where S i is the score of the i-th thinking dimension. Further, the step S108 specifically includes:

[0111] (1) Based on the basic information and historical evaluation results of the original portraits of students' double writing of the five major thinking training, give personalized suggestions for the five major thinking training of the standard Chinese character shapes of general regular script and running regular script, and generate corresponding improvement plans;

[0112] (2) Provide suggestions for training the five major thinking skills in writing and generate corresponding improvement plans;

[0113] (3) Students click the output button to print the double writing five thinking training plan and download the relevant training questions.

[0114] The present invention also provides a double writing thinking training level assessment system based on Chinese character recognition, comprising:

[0115] (1) A writing format recognition module for analyzing writing ability information;

[0116] (2) Writing thinking analysis module, used to analyze the five major thinking abilities;

[0117] (3) a writing content recognition module, used to obtain writing ability information;

[0118] (4) Writing thinking analysis module, used to analyze the five major thinking abilities;

[0119] (5) Dual writing thinking evaluation module, used to comprehensively analyze the level of dual writing thinking;

[0120] (6) Thinking evaluation management module, used to store and maintain scoring criteria;

[0121] (7) an information recognition module, used to convert image information;

[0122] (8) Information analysis module, used to collect and analyze student information;

[0123] (9) Evaluation module, used to collect and analyze evaluation results;

[0124] (10) System management module, used to manage system users, permission settings, data backup and recovery.

[0125] Beneficial effects: Compared with the prior art, the advantages of the present invention are:

[0126] (1) Objectivity and accuracy: The present invention can objectively and accurately reflect the students' double writing thinking training level. By using the point recognition technology of the square and round hour writing grid, the students' regular script and semi-cursive character shape information and text content information are extracted, solving the problem of handwriting font recognition and error correction, thereby ensuring the objectivity and accuracy of the evaluation results.

[0127] (2) Personalized training suggestions: According to the actual situation and needs of students, the present invention provides personalized training suggestions to help students improve their double writing ability in a targeted manner. By generating portraits of students' five major thinking abilities in double writing and analyzing and evaluating their thinking levels, the system can provide students with personalized improvement plans to make training more effective.

[0128] (3) Systematic evaluation process: Compared with the prior art, the present invention focuses on the educational goal of "beautiful writing and writing" in writing, and adopts a systematic evaluation process for the writing evaluation questions in the unified primary school Chinese textbook. This not only improves the comprehensiveness and meticulousness of the evaluation, but also helps to improve students' overall Chinese literacy.

[0129] (4) Scientificity and efficiency of teaching: The present invention enhances the scientificity and efficiency of teaching by providing detailed thinking portraits and personalized training programs. Students can better understand their own level of dual-writing thinking, which provides direction and training content for the next step of learning, and improves the pertinence and effectiveness of teaching. BRIEF DESCRIPTION OF THE DRAWINGS

[0130] Figure 1 It is a schematic diagram of the process of the present invention;

[0131] Figure 2 It is the regular script and semi-regular script square and circle grid bitmap of the present invention;

[0132] Figure 3 It is a schematic diagram of the square and circle hour grid structure of the present invention;

[0133] Figure 4 It is a schematic diagram of the square and circle hour grid point structure of the present invention;

[0134] Figure 5 It is the square and circle grid calligraphy chart of the present invention;

[0135] Figure 6 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0136] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0137] The specific technical details and schemes of each step of this method are as follows:

[0138] 1: According to the textbook teaching standards and the square and circle grid system, five thinking databases of square and circle clock grid point analysis method, general regular script and running regular script standard Chinese characters are established.

[0139] (I) First, collect and sort out the commonly used regular script and semi-cursive Chinese characters in the national unified edition of primary school Chinese textbooks by grade, mainly the 3,500 characters in the first-level character table of the commonly used character set of the "General Standard Chinese Character Table", then enter the 3,000 characters in the second-level character table and the 1,605 characters in the third-level character table, a total of 8,105 common regular script and semi-cursive Chinese characters.

[0140] (II) Establish the framework of the five thinking databases of the general standard Chinese characters in regular script and running script for the square and circle hour grid point analysis method. Including: the point data of the general regular script and running script standard Chinese characters on the outer circle of the square and circle grid, and the five thinking weights of the general regular script and running script standard Chinese characters.

[0141] Specifically, the method for constructing the five thinking databases of the square and round grid font standard includes the following steps:

[0142] 1. Initialize the database framework, including the standard ASCII code of Chinese characters, the five thinking weights of standard Chinese characters, the point data of standard Chinese characters on the circle line outside the square and circle grid, and other basic information.

[0143] 2. Use structured database management systems (such as MySQL, Oracle, SQL Server) to store and manage data;

[0144] 3. Define the database table structure, including the standard Chinese character table T H , Five thinking weight table T W , Standard Chinese characters on the circle outside the square grid data table T P ;

[0145] 4. Detailed information of the standardized Chinese characters is stored in the standardized Chinese character table, including the standardized Chinese character national standard ASCII code Hi and the source of the standardized Chinese characters in the primary school Chinese textbooks; H The detailed information of the standard Chinese characters is stored in the text book, including Hi and the position index of the Chinese characters in the textbook;

[0146] 5. In the five thinking weight table, store the five thinking weight arrays that the character mainly trains, that is, the weight of the character for intuitive thinking, image thinking, logical thinking, dialectical thinking, and innovative thinking training. For example, pictographic characters mainly train students' image thinking, so the image thinking weight is the largest (for example: 0.1, 0.6, 0.1, 0.1, 0.1, the weight is adjusted according to the specific situation). In this way, the standard Chinese characters appearing in the textbook are divided into several categories, and each category is assigned a weight according to the same annotation;

[0147] Denoted as weight vector W i ,but

[0148] ·W i =[W i1 ,W i2 ,W i3 ,W i4 ,W i5 ]

[0149] Where W ij represents the weight of the i-th Chinese character in the j-th type of thinking training;

[0150] The weights are assigned using a mathematical formula, which is calculated as follows:

[0151]

[0152] Among them, f ij is the frequency of the ith Chinese character in the jth type of thinking training.

[0153] 6. In the data table of the points of the standard Chinese characters on the outer circle of the square and circle grid, the coordinates of the starting and ending points of the strokes near the outer circle of the square and circle grid are stored. Among them, the data of the points of the standard Chinese characters on the outer circle of the square and circle grid are recorded as the coordinate set Pi ,

[0154] P i ={(x ik ,,y ik ))|k=1,2...,n}

[0155] a. The ratio of the circle to the square in the square grid satisfies the golden ratio. Most standard Chinese character structures can be determined by the positions of the points on the hour grid.

[0156] b. Establish a rectangular coordinate system with the center of the grid as the origin, and take one pixel of resolution as the unit. Then, the coordinate arc range function corresponding to the twelve points and twelve half-points of the circle line outside the square grid can be determined. By setting the range value with the arc-shaped point as the center, the twenty-four square grid points and half-point areas can be determined.

[0157] c. Extract the starting point, ending point, turning point, and hook-foot coordinate point regions of the standard Chinese characters within the outer circle line of the square grid through computer vision techniques such as edge detection and contour extraction methods, and record them. For example, the character "划" has 4 points close to the outer circle line, which are 10 o'clock up, 8 o'clock, 1 o'clock, and 5 o'clock respectively.

[0158] 7. Correlate the standard Chinese character table T H , the five major thinking weight tables T W , and the data table T of the position points of the standard Chinese characters on the outer circle line of the square grid P to establish associations. Through mathematical relationships, T H , T W , and T P are mutually associated to form a complete database architecture; assume that the primary key of the T H table is H i d, and the T W table is associated with the T i d table. Among them, H H d is used as a foreign key in the T i table; the T W table is also associated with the T P table through H i d, and H H d is also a foreign key in the T i table.

[0159] 8. After the database is constructed, database maintenance and updates can be carried out to reflect the latest textbook specifications and updates of the five major thinking weights;

[0160] 9. Output the constructed five major thinking database framework of the square grid font specification.

[0161] (3) Continuously update the database, establish a database update mechanism, regularly check and update the database content to reflect the latest textbook specifications and updates of the five major thinking weights.

[0162] (4) Based on reinforcement learning technology, by training an agent to learn, evaluate, and optimize the five major thinking weights of standard regular script and semi-cursive script Chinese characters, the agent optimizes the five major thinking weights of standard regular script and semi-cursive script Chinese characters according to the basic information and historical evaluation data of the students collected. Specifically, it includes the following steps:

[0163] (1) Define the objective function:

[0164] The objective function J(θ) is expressed as:

[0165]

[0166] Among them, N is the number of samples in the training data, s i ​​and a i is the state and action of the i-th sample, R(s i ,a i ) is the reward of the sample, Q(s i ,a i ,θ) is the Q value predicted by the agent, and θ is the agent parameter;

[0167] (2) Constructing a Q-network: Use a neural network to approximate the Q function, which takes the state and action as input and outputs the Q value;

[0168] (3) Establishing an experience replay buffer: During training, the observed state-action-reward-next-state pairs are stored in the experience replay buffer;

[0169] (4) Strategy selection: In the actual environment, the agent uses a strategy π(a|s) to select actions. This strategy can be greedy or ε-greedy.

[0170] (5) Training the Q network: Use stochastic gradient descent (SGD) to update the parameters θ of the Q network. In each iteration, a batch of samples are randomly selected and the target Q value Q′(s, a, θ′) is calculated, where θ′ is the parameter of the target network. The target network parameters are periodically copied from the Q network parameters. The target Q value is calculated as:

[0171]

[0172] Where γ is the discount factor, s′ is the next state, and the gradient for updating the Q network parameters is:

[0173]

[0174] (6) Weight update: Based on the output of the Q network, the agent adjusts the weights of the five major thoughts to minimize the objective function. The weight update is:

[0175]

[0176] Among them, α w is the learning rate for weight updates.

[0177] 2: Based on the writing exercises in the unified primary school Chinese textbooks and the five major thinking related theories, a writing exercise database of the five major thinking language models is established.

[0178] (1) First, collect and organize the contents and requirements of the unified Chinese textbook writing and the five major thinking related theories, including writing sections, genre distribution, teaching requirements, and basic concepts of the five major thinking.

[0179] 1. Five major types of thinking: intuitive thinking, figurative thinking, logical thinking, dialectical thinking and creative thinking.

[0180] 2. Collect a large number of student essays from the unified primary school Chinese textbooks and the corresponding identification, judgment, and correction opinions for their five major thinking, and digitize them and convert them into a corpus for machine learning model training. These essays should cover students of different grades and levels from grades 1 to 6 to ensure the generalization ability of the model; the essays can be actual classroom assignments or essay competition works to ensure the diversity and representativeness of the essays; the correction opinions come from the teacher's detailed comments and modification suggestions, and the correction opinions should cover the identification, judgment, and modification suggestions of the five major thinking to facilitate subsequent model training and evaluation. This corpus needs to cover the essays of each unit of the unified primary school Chinese textbook, and the identification, judgment, and correction opinions need to be detailed and consistent.

[0181] 3. Select a pre-trained language model and use machine learning (ML) technology to fine-tune the pre-trained language model on the practice thinking corpus.

[0182] (1) Use text feature extraction, classification models and other methods to identify the thinking that appears in the exercises.

[0183] (2) Standard scoring of thinking level is performed using two parameters: frequency of thinking and level of thinking application (divided into four levels: A, B, C, and D according to the data set);

[0184] (3) Use methods such as cross-validation and confusion matrix to evaluate model performance and continuously adjust parameters.

[0185] 4. Afterwards, according to the classification of exercises in the unified textbooks, and according to the five major thinking training requirements and difficulty requirements corresponding to the teaching objectives, formulate specific scoring criteria.

[0186] (1) First, experts weight the five major thinking skills based on the training requirements corresponding to the teaching objectives. The scores of the key training skills are multiplied by the key factor (>1) and then by the grade influence factor (which decreases with increasing grade).

[0187] (2) Select a large number of essay samples from this unit and input them into the writing thinking language model for scoring. Divide the normal distribution mean obtained after key weighting by the standard normal distribution mean of the score to obtain the difficulty weighting.

[0188] (3) Use methods such as cross-validation and confusion matrix to evaluate model performance and continuously adjust parameters.

[0189] (4) Import emphasis weighting and difficulty weighting into the writing unit evaluation information database.

[0190] 5. Continuously update the model, establish a model update mechanism, regularly input new corpus training models and regularly evaluate and adjust parameters to further optimize the model.

[0191] 3. Students write their compositions on paper and upload pictures and information according to the dual-writing training requirements.

[0192] (I) Training and cultivation of double writing thinking

[0193] Based on the development of students' core Chinese literacy, taking the cultivation of "thinking ability", one of the four core literacy in the "Compulsory Education Chinese Curriculum Standards (2022 Edition)" as a breakthrough, and based on the psychological development laws of primary school students, we will construct a five-level thinking evaluation system for "writing and writing" (ie "double writing") in primary schools, and carry out relevant training with this goal.

[0194] In order to solve the technical problem of poor writing effect of standardized Chinese characters in writing, the present invention has developed a square and circle clock grid calligraphy method. Students use the square and circle clock grid calligraphy method to carry out the "radical" classification learning of common regular script and regular script standardized Chinese characters. Students only need to remember the positions of 201 main radicals and 99 attached radicals on the outer circle of the square and circle clock grid, and learn 38 strokes, 16 stroke orders, and 52 structural writing rules at the same time, so as to apply what they have learned to other cases and assemble and combine Chinese character components, and then they can write common regular script and regular script standardized Chinese characters.

[0195] (1) Combination of intuitive thinking and square and round grid writing

[0196] Intuitive thinking refers to the ability to quickly judge, guess or imagine the answer to a problem based on the perception of internal factors without step-by-step analysis. When writing in a square and round grid, students need to accurately judge the structure of the font, the direction of the strokes and the overall layout in a short period of time, which requires a high level of intuitive thinking ability. The intuitive thinking training of writing in a square and round grid aims to improve writing skills and creativity by cultivating intuitive perception and quick judgment ability. Specifically, students first write in accordance with the general standard Chinese character regular script and running script points on the square and round clock grid, repeatedly practice writing to form hand muscle memory and writing habits, remember the corresponding points, and eventually be able to break away from the square and round clock grid and find the correct character shape points on the square grid, from the square and round clock cross grid to the square and round clock grid and then to the square grid, to achieve the intuitive thinking of character shape standard writing with "no grid in the eyes, but grid in the heart".

[0197] The square and circle clock grid designed by the present invention according to the law of students' thinking development is a scientific writing grid that is conducive to students writing Chinese characters well. The grid is a "circle" added in the middle of the square grid, and the "circle" is evenly divided into 12 blank points according to the clock. The "circle" can be placed in the mouth grid, field grid and rice grid, or even in the horizontal grid. Various auxiliary lines in the grid conduct intuitive analysis of the strokes or structure of Chinese characters to improve students' writing comprehension. The length of the square and circle grid and the diameter of the circle are strictly set according to the ratio of the new character writing grid and the new character in the unified primary school Chinese textbook, which conforms to the basic golden ratio principle. When the square is 14 cm long, the diameter of the circle is 10 cm; when the square is 12 cm long, the diameter of the circle is 9 cm. The square and circle grid improves the theory of "tight center palace and extended outer palace" of Chinese character writing, and combines the theory of "extension point" of Chinese character writing. The square and circle grid has five functions: stroke positioning, pen shape reference, outline (controlling the size of the character), structural boundary (controlling the proportion of the character) and center of gravity boundary (controlling the center of gravity of the character).

[0198] (2) Combination of image thinking and square and round grid writing

[0199] When writing Chinese characters in the square and circle grid, students need to carefully observe the structure, proportion and relationship between the strokes of each character. This kind of meticulous observation helps to cultivate students' observation skills and improve their ability of visual thinking. The size of the square and circle in the square and circle grid is strictly set according to the ratio of the Chinese character writing grid and the new characters, which conforms to the basic golden ratio principle. Using the square and circle clock grid as a positioning reference, the stroke starting point, end point, hook foot or inflection point of the strokes of the standardized Chinese characters in the outer contour mostly fall on the outer circle line. The grid adds a "circle" in the middle of the square grid, and the "circle" is evenly divided into 12 points according to the clock. The outer circle line of the "circle" and the "cross" auxiliary line in the grid analyze the strokes or structural position of the standardized Chinese characters. There are two types of points in the square and circle cross grid. One is the whole point, such as "1 o'clock"; the second is the non-whole point in the clockwise direction, with 12 o'clock and 6 o'clock as the boundary. The points between 1 o'clock and 6 o'clock are called "down", such as "1 o'clock down"; the non-whole points between 6 o'clock and 12 o'clock are called "up", such as "7 o'clock up" and "10 o'clock up". As shown in Example 1: "Hua" is a left-right structured character, wide on the left and narrow on the right. It has 4 points connected to the outer circle, namely 10 o'clock up, 8 o'clock, 1 o'clock and 5 o'clock. As long as the longer strokes are written on the outer circle line, it can be ensured that the character "Hua" is of appropriate size in the square grid and written in the center.

[0200] Fangyuan Hour-grid Calligraphy has three major characteristics in terms of image thinking. First, it controls the center of gravity and positions it in the middle: When students do the "double-writing" assignment in the Fangyuan grid, the "circle" always reminds students to write the characters exactly in the middle of the grid, without leaning left or right, up or down. The coordinate positioning of the center point of the Fangyuan grid is highly accurate, with "eight directions converging at the center", providing a reference area for the center of gravity. When writing, the main strokes of the characters should converge towards the center of gravity area, and the probability of the main strokes passing through the center of gravity area is extremely high. The 12 o'clock, 3 o'clock, 6 o'clock, and 9 o'clock positions are the marking points of the middle horizontal line and the middle vertical line, with particularly high usage rates. For example, the long horizontal stroke of the single-structure character "ping" is at the 9 o'clock and 3 o'clock positions, the hanging-needle vertical stroke is at the 12 o'clock and 6 o'clock positions, and the intersection of the horizontal and vertical strokes is at the center point. Another example is the upper-lower structure character "kong", where the upper dot and the vertical stroke are aligned with the 12 o'clock and 6 o'clock positions. For the semi-surrounding structure character "yi", the starting points of the two left-falling strokes are both aligned with the 12 o'clock and 6 o'clock positions, the fourth horizontal stroke starts at the 9 o'clock position and ends at the 3 o'clock position, and the intersection of the horizontal and left-falling strokes is also at the center point. For the left-right structure character "pa", the fifth vertical stroke is aligned with the 12 o'clock and 6 o'clock positions; with the prompt of the position points, the above characters are written in the middle and are not easy to shift. Second, it is hour-grid positioning: The 12 positions of the circle can prompt students about the positions of the strokes and structures of Chinese characters. Since students are very familiar with the 12 positions of the clock, it is extremely intuitive and vivid. For example, the starting or ending points of the three strokes of the dot, horizontal, and vertical hook of the character "jing" are exactly at the 12 o'clock, 10 o'clock, 2 o'clock, and 6 o'clock positions; the starting or ending points of the four strokes of the horizontal, vertical hook, horizontal fold and oblique stroke, and right-falling stroke of the character "guo" are exactly at the 1 o'clock, 2 o'clock, 9 o'clock, 8 o'clock, and 5 o'clock positions. Third, it controls the outer contour. The outer circle of the Fangyuan grid fits the characteristics of Chinese character formation. It is found through research that most of the starting points, ending points, hook feet, or turning points of the strokes on the outer contour of Chinese characters will fall on the outer circle line. For example, the starting point, ending point, and hook foot of the horizontal stroke of the character "ding" exactly fall on the outer circle line. For the character "qing", the starting points, ending points, or hook feet of the five strokes of the dot, lift, vertical, horizontal, and hook fall on the outer circle line; for the character "hui", the starting points, ending points, or turning points of the four strokes of the left-falling stroke, right-falling stroke, left-falling and right-falling stroke, and dot fall on the outer circle line, etc. The Fangyuan grid can thus help students control the outer contour of Chinese characters when writing.

[0201] (3) The combination of logical thinking and Fangyuan-grid calligraphy

[0202] As a tool to assist writing, the original intention of the design of the Fangyuan grid is often to help students write Chinese characters or other scripts more standardly and beautifully. Through clear boundaries and divisions, the structure of the font becomes clearer and the strokes become more even. However, expanding the application of the Fangyuan grid to the training of logical thinking ability, although not its direct purpose, can indeed indirectly promote the improvement of the flexibility of students' thinking quality.

[0203] Spatial layout and structural thinking: The "square" in the square and circle grid represents rules and boundaries, while the "circle" symbolizes tolerance and change. When students use the square and circle grid to write, they need to constantly think about how to reasonably arrange the direction and position of the strokes in a limited space, which helps to cultivate their spatial layout ability and structural thinking ability when solving problems. This ability is also important in logical thinking, especially when dealing with complex problems, it is necessary to clearly divide the problem structure and arrange the solution steps in an orderly manner.

[0204] Following rules and being flexible: The use of the square and circle grid requires students to follow certain rules (such as stroke order, font structure, etc.), but at the same time encourages students to seek innovation and change within the rules. This ability to follow rules and be flexible is exactly what logical thinking pursues. In logical thinking, students need to understand and follow logical rules, and at the same time have the ability to flexibly adjust strategies and solve new problems within the framework of rules.

[0205] Observation and analysis: When writing in a square grid, students need to carefully observe the position, size, and proportion of each word in the grid, which helps to cultivate their observation and analysis skills. In logical thinking training, observation and analysis skills are indispensable. By observing the appearance of the problem and analyzing the internal logic, students can more accurately grasp the essence of the problem and find effective solutions.

[0206] Creative stimulation and imagination: Although the design of the square grid itself is fixed, students can try different layouts, font styles, etc. during the writing process, which helps stimulate their creativity and imagination. In logical thinking training, encouraging students to jump out of traditional thinking patterns and boldly imagine new solutions is also an important way to improve the flexibility of thinking quality.

[0207] (4) Combination of dialectical thinking and square and round grid writing

[0208] Dialectical thinking emphasizes analyzing problems from a comprehensive, connected, and developmental perspective, and seeking reasonable solutions. Writing is not just a simple combination of strokes, but also requires consideration of the overall layout and coordination. When writing, we must have systematic thinking and a global concept, pay attention to the spacing, size, height, and other relationships between characters, so that the overall look is harmonious and beautiful. In addition to being able to determine the position and direction of the starting and ending strokes, the length and twists of the strokes, and the size and straightness of the line angles, the square and round grid also has a very profound cultural connotation of "the sky is round and the earth is square", which is a concentrated embodiment of Chinese dialectical thought. The differences in squareness and roundness of Chinese characters are caused by different strokes, such as cutting or lifting the pen, and stopping and stopping to produce square corners, and turning the pen, connecting the pen, and pressing the pen to produce round turns. The first fold of the horizontal stroke is a square fold, while the second fold is a round turn. One square, one round, one dynamic, and one static fully show the balance of yin and yang. Because the squareness of the square frame of the character is in opposition and complement to the arc of the internal stroke and the roundness of the dot, the squareness and roundness of the overall structure constitute the dialectical balance of philosophical thought. By analyzing the layout of Chinese characters in the square and round grid, students can understand the role and position of each stroke in the overall structure. This analysis process not only helps improve students' writing skills, but also cultivates their analytical skills and learns to look at problems from multiple perspectives.

[0209] Observation and analysis: Character structure analysis. When writing Chinese characters in the square and round grid, first observe the overall structure of the character and analyze which parts need to be rounded and which parts need to be squared. This process of observation and analysis helps to cultivate students' observation and analytical skills, enabling them to look at problems from multiple angles. Analysis of stroke relationships. Chinese characters are composed of multiple strokes, and there is a certain relationship between each stroke. In the square and round grid, by comparing the length, angle and position of different strokes, students can more clearly understand the mutual influence and restriction between strokes, thereby cultivating their dialectical thinking ability and learning to consider the interaction between various factors when dealing with problems.

[0210] (5) Combination of creative thinking and square and round grid writing

[0211] Creative thinking requires deep processing of existing sensory materials, and the use of imagination to diverge and aggregate. It is the result of a high degree of unity of multiple thinking. As a writing tool or practice method, square and circle grid writing, through creative application and guidance, tries free writing, creative composition, font design and other activities in the square and circle grid, and combines writing with graphics, colors and other elements. It can become an effective way to cultivate students' creative thinking ability and enhance the originality of thinking quality. The following are some strategies on how to train students' creative thinking and enhance their originality through square and circle grid writing:

[0212] Imitation and innovation: imitate the style of excellent calligraphy works by copying calligraphy, and try to add your own understanding and innovation on the basis of imitation to make the font more personalized and creative. It helps to cultivate innovative thinking and aesthetic ability.

[0213] Deformation Exercise: Deform the common standard Chinese characters according to Yan Zhenqing's brushwork, such as changing the stroke order, shape or structure, to create calligraphy works with a unique style. This exercise can stimulate students' imagination and creativity and cultivate their innovative thinking.

[0214] Dynamic adjustment: During the writing process, you need to constantly adjust the pen tip, strength and speed to adapt to different strokes and font styles. This dynamic adjustment ability helps to cultivate the ability to be flexible and solve problems. Font design: Guide students to try different font styles, such as regular script, running script, cursive script, etc., or combine the characteristics of multiple fonts for innovative design to form their own unique writing style.

[0215] Diversified expression: Encourage students to associate and imagine based on the content of writing, and transform abstract thinking into concrete written expression. For example, when writing the word "mountain", you can associate it with majestic peaks and magnificent natural scenery; when writing the word "dream", you can imagine yourself striving forward on the road to chasing your dreams. Through association and imagination, students can better understand and feel the cultural connotation and emotional value contained in Chinese characters, thereby improving their cultural literacy and aesthetic ability.

[0216] Innovative layout design: Encourage students to break the boundaries of the traditional square and circle grid and try to make creative layout designs within the grid. For example, students can freely plan the position, size, direction and even color of the font to create a unique writing effect. This process of challenging tradition and giving free play can stimulate students' creativity and imagination.

[0217] Integrate multiple elements: Guide students to combine the square circle grid with other art forms or elements, such as painting, pattern, color, etc., to create writing works with personal characteristics. By integrating different elements, students can explore new ways of expression and visual effects, thereby cultivating their cross-border thinking and innovation ability.

[0218] Theme creation challenge: set a specific theme or situation, and students create around these themes in the square circle. For example, students can express their imagination and creativity through writing and graphics with themes such as "future city" and "dream world". This theme creation challenge can stimulate students' desire to create, prompting them to think about problems from different angles and find unique solutions.

[0219] Personalized font design: Encourage students to design personalized fonts according to their own preferences and styles, and practice in the square and circle grid. Through repeated attempts and adjustments, students can gradually form their own unique writing style. This process not only exercises their creativity and aesthetic ability, but also enhances their self-confidence and desire to express themselves.

[0220] Critical reflection and iteration: After completing their work, guide students to reflect critically, analyze the strengths and weaknesses of their work, and think about how to improve and iterate. This reflection process helps students recognize their creative potential, learn to learn from failures, and continuously improve their creative thinking ability and originality.

[0221] The square and round grid calligraphy training guides students to use their imagination, improve their perception, and be able to draw inferences from one example and apply it creatively. For example, the eight basic strokes of Chinese characters are extremely varied in terms of inversion, inclination, flexion and extension. Take the "dot" for example, there are many different strokes. For example, the three dots at the bottom of the heart character echo each other, and the strokes are different, which can guide students to observe carefully, think and summarize. All the three dots at the bottom of the heart character are like the trajectory of the sun's movement in a day. The dot on the left is like the rising sun quietly poking its head out from the top of the mountain, the dot in the middle is just like the scorching sun at noon, and the dot on the right is just like the setting sun beside the mountains. The variability and expressiveness of the dots and strokes are the beauty of the Chinese character shape. Students are still unable to understand things and grasp the concepts of things from the essential attributes of things. In the thinking activities of solving problems, it is necessary to rely on intuition to understand abstract concepts, and even rely on concrete images to make logical inferences. According to the thinking characteristics of students at this stage, the square and round grid "double writing" homework adopts more methods such as "real object intuition", "language intuition" and "illustration and text". Chinese characters originated from pictograms and are based on pictograms. When lower grade students learn to write Chinese characters, they are supplemented with the origins and explanations of pictographic characters, indicative characters, ideographic characters, and phono-semantic characters. Students can not only practice writing, but also write the characters they have practiced into short Chinese character stories. This is undoubtedly a better way to train "double writing" thinking.

[0222] Comprehensive training of double writing thinking ability

[0223] The comprehensive training of double writing (handwriting and writing) thinking ability is a comprehensive process that aims to exercise and improve individual thinking ability through both handwriting and writing. Writing exercises are interspersed in the writing process, such as asking students to practice calligraphy or copy classic passages before writing to calm their minds; in writing, pay attention to neat handwriting and beautiful layout to improve the overall quality of the article. Integrate writing skills into writing teaching: such as teaching students how to use different strokes and font styles to enhance the expressiveness of the article. Encourage students to apply the knowledge and skills they have learned in daily life, such as recording life details and insights through writing diaries and blogs; and apply the knowledge they have learned to actual situations through participating in community activities or social practice. The comprehensive training of double writing thinking is a comprehensive and systematic process that combines the advantages and characteristics of both handwriting and writing activities, and improves students' thinking ability, comprehensive literacy and innovation ability through diversified practice strategies and practical application scenarios. Through systematic training and practice, individual thinking ability can be gradually improved and reflected in handwriting and writing.

[0224] (II) Writing and uploading

[0225] After double-writing thinking training, students practice writing Chinese characters in general regular script and semi-cursive script on exercise paper, and log in to upload photos.

[0226] 1. Student login, information collection and processing.

[0227] a. By providing a user interface and form, students are allowed to fill in basic information online and verify the integrity of the information, such as date of birth, date of enrollment, etc. A basic profile of students is established, including personal information and writing experience, which will be used for subsequent analysis of the five thinking training levels and weak links of students' writing.

[0228] b. Clean and organize the students’ historical assessment data to form a complete data set. The decision tree model is used to learn the distribution of the data and identify points that are significantly different from the data distribution. Specifically, the anomaly detection methods in Python’s Scikit-Learn library, such as IsolationForest or LocalOutlierFactor, are used to identify outliers and remove or resample them to ensure the quality of the data and the accuracy of the model.

[0229] c. For numerical data, the mean or median of the data can be used to fill missing values. For categorical data, the mode (the most common value) can be used to fill missing values.

[0230] 2.Upload the photos.

[0231] 4. Based on the pictures uploaded by students, extract the general regular script and running script standard Chinese handwritten font information and text content information, generate a portrait of the students' five major double writing thinking abilities, and analyze and evaluate their double writing thinking levels.

[0232] (I) Based on the pictures uploaded by students, collect and process the information of students' commonly used regular script and semi-cursive Chinese character writing exercises, and extract the regular script and semi-cursive character shape information and text content information.

[0233] This system uses the square and round hour grid to identify the structural points of the writer's handwritten standard Chinese characters in order to analyze the writing ability information of the target standard Chinese character writer.

[0234] This system analyzes the strokes, structure, and composition of standard Chinese characters in the user's handwritten composition. It not only obtains the writing ability information of the target standard Chinese character writer, but also forms complete writing practice reference information, providing users with simple and convenient training templates.

[0235] This system inputs the representative points of standardized Chinese characters on the outer circle line, and inputs the writing rules of strokes, radicals, and composition. It analyzes the positional relationship, distance relationship, and writing length of each part of the Chinese character, and extracts the corresponding length, height, spacing, angle, and other data from the stroke length, spacing, angle, and symmetry relationship in the radical structure based on the starting point, outer circle point, and stroke structure in the radical structure as the unit of measurement to locate the writing trajectory of all strokes of other radicals (including writing direction, writing angle, writing length, writing height, writing spacing, and various aesthetic symmetry relationships, etc.), and then completes the final writing of the whole word according to the writing rules of the Chinese character itself. It mainly relies on the representative points of standardized Chinese characters in the square and circle hour grid, and thoroughly and comprehensively analyzes the positioning of Chinese characters and the structure and composition, saving a lot of time cost.

[0236] 1. Perform a series of preprocessing on the image, including image grayscale conversion, image binarization, image denoising, normalization and other steps to ensure the consistent quality of the input image.

[0237] The image grayscale conversion is to convert the color image of the homework uploaded by the students into a grayscale image, and use the following formula to represent the grayscale value of each pixel:

[0238] I gray (x,y)=0.2989R(x,y)+0.5870G(x,y)+0.1140B(x,y)

[0239] Among them I gray is the pixel value of the grayscale image at position (x, y), and R, G, and B are the pixel values ​​of the red, green, and blue channels of the original color image, respectively.

[0240] Image binarization is to convert a grayscale image into a binary image containing only black and white colors, using the threshold segmentation method. The formula is as follows:

[0241]

[0242] Among them, B(x,y) is the pixel value after binary, G(x,y) is the pixel value of the grayscale image, and T is the set threshold;

[0243] Image denoising uses Gaussian filtering algorithm to reduce image noise. The formula is as follows:

[0244]

[0245] Where I′(x,y) is the pixel value after denoising, I(x,y) is the pixel value of the original image, w(i,j) is the weight coefficient of the Gaussian kernel, and k is the size of the filter kernel;

[0246] 2. Use the Canny edge detection algorithm to identify the edges in the image. Use the Hough Transform to detect the straight lines in the image, thereby identifying the boundaries of the grid. Use the Hough Transform to detect the straight lines in the image, thereby identifying the boundaries of the grid. The calculation formula is:

[0247] r=xcos(θ)+ysin(θ)

[0248] Among them, (r,θ) are the parameters in Hough space, and (x,y) are the coordinates of the edge point in image space.

[0249] 3. Determine the vertex coordinates of each grid by detecting the intersection of the straight lines. Divide the image into individual grids according to the vertex coordinates for subsequent Chinese character recognition and analysis. Assume that the equations of the two straight lines are:

[0250] Line L1: Ax+By+C1=0

[0251] Straight line L2: A′x+B′y+C2=0

[0252] Then their intersection point (x, y) can be obtained by solving the following system of equations:

[0253]

[0254] According to the vertex coordinates, the image is segmented into individual grids for subsequent Chinese character recognition and analysis.

[0255] 3. Use text recognition methods to identify the text in each grid, and call up five thinking databases including common regular script, semi-regular script, standard Chinese character shapes, and standard square and round grid points.

[0256] Chinese character recognition: Use optical character recognition (OCR) technology to perform Chinese character recognition on the preprocessed image, extract the handwritten font information of Chinese characters, and represent the recognized Chinese characters as strings for subsequent analysis. The OCR recognition process can be represented as a function f OCR , which will preprocess the image I pre Mapped to the recognized Chinese character string S, the calculation formula is:

[0257] S=f OCR (I pre )

[0258] Text content analysis: segment and tag the recognized Chinese characters to obtain information about the text content. Use natural language processing technology to perform semantic analysis on the composition content and extract key information. The segmentation and part-of-speech tagging functions are used. seg and f tag Indicates that the string S is operated respectively:

[0259] T=f seg (S)

[0260] P = f tag (T)

[0261] Where T is the word sequence after word segmentation, and P is the corresponding part-of-speech tag sequence.

[0262] 5. In each grid, extract the starting, ending, folding and hooking strokes within a certain range inside and outside the circle line of the square grid through morphological operations (such as corrosion and expansion). Analyze the geometric characteristics of the strokes, determine the areas where the four situations fall, and score them.

[0263] a. Establish a rectangular coordinate system with the center of the grid as the origin, and take one pixel of resolution as the unit. Then, the coordinate arc range function corresponding to the twelve points and twelve half-points of the circle line outside the square grid can be determined. Set the range value with the arc-shaped point as the center, and the twenty-four square grid points and half-point areas can be determined.

[0264] b. Check the areas where the monitored starting, ending, folding and hooking strokes fall, and compare them with the points of the standardized Chinese characters in the database.

[0265] c. Determine the number of standard Chinese character points, and evenly distribute the scores according to the number (score for each point = total points / number of points). For every deviation of half a point, 25% of the score will be deducted. If the deviation exceeds two points, no points will be awarded.

[0266] d. Assign the basic score of the character shape to the weight of the character shape thinking training, obtain the sub-score of each thinking training, and record it in the database.

[0267] e. After scoring the entire text, perform a weighted summation of the scores of each thinking training sub-score to obtain the training frequency and training level. The training frequency is the frequency of occurrence of the words belonging to this type of thinking training, and the training level is equal to the total score divided by the training frequency and then multiplied by an adjustment factor.

[0268] f. The training frequency and training level scores of the five major thinking are stored in the historical records.

[0269] (ii) Collect historical assessment results for analysis and evaluation of dual writing thinking levels and establish a database.

[0270] (III) Analyze the five thinking training levels and weak links of students’ writing and composition.

[0271] 1. Extract the assessment data obtained from the model and database evaluation, and use statistical analysis methods to calculate students' average scores of the five major thinking skills, teaching achievement level and other indicators.

[0272] 2. Perform feature engineering to extract features related to the five major thinking in the writing, such as the five major thinking scores, the training frequency and training level scores of the five major thinking, the frequency of thinking application, the thinking application level indicators, the weak thinking types, etc. Specifically, use the data preprocessing tool Pandas library to clean and transform the data. Use the information gain feature selection method to select the most representative features, and use feature conversion methods such as normalization and standardization to process the features for model training.

[0273] 3. Use machine learning algorithms to model features and train prediction models. Specifically, use the Scikit-Learn library, a model training tool, to select the decision tree algorithm in the machine learning algorithm, train the model, and adjust model parameters such as tree depth and learning rate to improve model performance.

[0274] 4. Use cross-validation and other methods to evaluate the performance of the model, and optimize the model based on the evaluation results. Use K-fold cross-validation technology to evaluate the performance of the model. Use Python's Matplotlib library to visualize and analyze the model performance, and optimize the model based on the evaluation results, such as adjusting model parameters and selecting better features.

[0275] 5. Use the trained model to predict the five thinking training levels and weak links of students' exercises. Generate a prediction report, including prediction results, weak links, etc. Visually display the prediction results in the form of bar charts, pie charts, etc. to facilitate students and teachers to understand.

[0276] (IV) Generate students’ personalized double-writing original portraits of the five major thinking training.

[0277] 1. The five thinking training portraits of students' personalized exercises include:

[0278] a. Basic information: students’ age, gender, grade and other basic information.

[0279] b. Five thinking training levels: students’ training levels in intuitive thinking, figurative thinking, logical thinking, dialectical thinking and creative thinking.

[0280] c. Common error types: The types of errors that students often make in history assessments and the impact of these errors on their writing.

[0281] d. Comprehensive expression ability: students’ comprehensive expression ability in their essays, including the richness and organization of the content.

[0282] e. Learning needs: students’ needs and expectations regarding the five major thinking exercises in writing exercises, as well as the learning resources and guidance they hope to receive.

[0283] 2. The method for generating the five thinking training portraits of students’ personalized exercises is as follows:

[0284] Input: basic information of students, historical assessment data

[0285] Output: Five thinking training portraits and training suggestions for students' personalized exercises

[0286] The main steps of the algorithm are:

[0287] a. Initialize the five thinking training portraits of students' personalized exercises, including students' basic information (such as age, gender, grade, etc.) and historical assessment data (such as assessment scores, five thinking training levels, etc.).

[0288] b. Use statistical analysis methods and machine learning algorithms to analyze students’ historical assessment data to identify the five thinking training levels and weak links in students’ exercises.

[0289] c. Carry out feature engineering, including extracting features related to the five major thinking aspects of exercises, such as the five major thinking scores, the training frequency and training level scores of the five major thinking aspects, the frequency of thinking application, the thinking application level indicators, the weak thinking types, etc.

[0290] d. Use machine learning algorithms to model the features and evaluate the performance of the model through cross-validation methods.

[0291] e. Based on the prediction results of the model, generate personalized portraits of the five major thinking training in writing for students, including the students' five major thinking training levels in writing, common error types and learning needs.

[0292] 3. Output five thinking training portraits of students’ personalized exercises.

[0293] Extraction of the five major thinking features: Based on the definition of the five major thinking features, design the corresponding feature extraction algorithm, including logical thinking features and innovative thinking features. These features can be extracted based on the results of text content analysis; among them, feature extraction is represented by a series of functions f feat,i , each function corresponds to a thinking characteristic F i :

[0294] F i =f feat,i (P)

[0295] Where i represents the i-th thinking characteristic;

[0296] Thinking ability portrait generation: The thinking features extracted from text and images are integrated to generate a student's thinking ability portrait, which is represented by a score;

[0297] The generation of thinking ability portrait is represented by function f profile, which combines all the features F i :

[0298] profile=f profile (F1,F2,…,F n )

[0299] Where n is the total number of thinking characteristics;

[0300] Analysis and evaluation of thinking level: Based on the thinking ability portrait, the scoring model is used to analyze the students' thinking level. The evaluation results give the students' overall performance in the five major thinking aspects and the specific scores of each thinking dimension. Among them, the scoring model of thinking level is expressed as function f score , generate a score S based on the thinking ability profile total :

[0301] S total =f score (profile)

[0302] The score for each thinking dimension is expressed as:

[0303] S total =f score,i (F i )

[0304] S i is the score of the i-th thinking dimension.

[0305] (V) Establish a database to collect historical evaluation results.

[0306] Establish a historical assessment result database to store students’ assessment data. The data includes students’ scores of the five major thinking skills in each assessment, the training frequency and training level scores of the five major thinking skills, the frequency of thinking application, and the level of thinking application indicators, etc. The accuracy and completeness of the data are ensured for subsequent statistical analysis and machine learning modeling.

[0307] 5. Based on the basic information and historical evaluation results of students' double-writing original portraits of the five major thinking training, give personalized suggestions for the five major thinking training of general regular script, standard Chinese character shapes, and the five major thinking training cultivation suggestions for writing.

[0308] (I) Based on the basic information and evaluation results of students' double-writing original portraits of the five major thinking training, give personalized suggestions for the five major thinking training of general regular script and running regular script standard Chinese character shapes, and generate corresponding improvement plans.

[0309] According to the students' double-writing thinking ability level and weak links, the next step of training questions suitable for the student is intelligently screened and generated from the question bank and combined into training plan suggestions.

[0310] (ii) Provide suggestions for developing the five major thinking skills in writing exercises and generate corresponding improvement plans.

[0311] 1. Suggestions on cultivating intuitive thinking in writing exercises

[0312] First, with regard to the "target results" of intuitive thinking in composition, students' composition writing is not standardized enough. We use the intuitive thinking square and circle grid training book to guide students to write neatly and promote expression in composition.

[0313] Second, focusing on the “object content” of intuitive thinking in practice, we should strengthen the teaching of empathy practice in teaching, and appropriately add teaching content such as emotional empathy, cognitive empathy, moral empathy, and cultural empathy.

[0314] The third is to focus on the "method ability" of intuitive thinking in practice, and conduct multi-faceted training on the thinking habits that students lack, and cultivate thinking habits such as "image, concept, proposition, reasoning, comprehensiveness, and connection". Thinking habits include single training of various types of thinking. In a sense, thinking habits also include other types of thinking, such as image thinking, logical thinking, dialectical thinking, creative thinking, etc., and training is carried out in multiple rounds of actions until relevant thinking intuition is formed.

[0315] Fourth, we develop training content for intuitive thinking in writing that is suitable for students of different intelligence levels, and implement differentiated teaching in class to suit each student. We develop a question bank that is suitable for different students, ranging from easy to difficult, combining various types of thinking.

[0316] 2. Suggestions on cultivating image thinking in writing

[0317] First, focusing on the "target results" of figurative thinking in composition exercises, when students first come into contact with new words in written language (such as nouns, verbs, adjectives) that represent the images of things, guide students to use their imagination to interpret the connotation and extension of the words, write the fresh words they have learned in their compositions and diaries, and keep a diary regularly.

[0318] The second is to focus on the "object content" of figurative thinking in composition, combine composition teaching with life in teaching, guide and awaken students' real life, be attentive to life, and write diaries diligently.

[0319] Third, with regard to the "methodological ability" of figurative thinking in composition exercises, teachers should improve their literary accomplishment, delve deeply into the teaching materials, have the awareness and ability of "imagery" teaching, and guide students to transfer their emotions to all things in the world and to what they write.

[0320] Fourth, in terms of the "strategic quality" of figurative thinking in essays, students should be organized to read more, think more, and combine their own experiences to write profound essays that are relevant to society and the country.

[0321] 3. Suggestions on cultivating logical thinking in writing exercises

[0322] First, based on the "target results" of logical thinking in composition, we sort out the logical thinking resources in primary school Chinese textbooks and carry out composition logic teaching.

[0323] The second is to use mind mapping or outlining methods to construct the framework of the writing context for the "object content" of the writing logical thinking, and express the content information clearly in a hierarchical and organized manner.

[0324] The third is to use concepts, propositions and reasoning methods to carry out relevant writing thinking training for the "method ability" of logical thinking in writing. For example, guide students to use causal complex sentences or sentence groups to express their views. Including comparison, analysis, generalization, deduction, induction, analogy, backtracking and other reasoning methods.

[0325] Fourth, with regard to the "strategic quality" of logical thinking in writing, students should be organized to conduct more training, such as learning from the methods of idea selection and composition layout learned in reading classes; drawing a mind map for each writing exercise to sort out ideas, etc.

[0326] 4. Suggestions on cultivating dialectical thinking in writing

[0327] First, based on the "target results" of dialectical thinking in composition, we sort out the dialectical thinking resources in primary school Chinese textbooks and carry out dialectical teaching in composition.

[0328] The second is to focus on the "object content" of dialectical thinking in composition, dig deep into the dialectical thinking in classic articles, and transfer it to composition teaching.

[0329] The third is the "method ability" of dialectical thinking in writing. When formulating the idea of ​​writing, we should strengthen the view of everything in the world with a "comprehensive, connected, and developmental" perspective, and write out the depth, height, and breadth of thinking.

[0330] Fourth, based on the "strategic quality" of dialectical thinking in writing, we organize the use of dialectical thinking for critical expression and cultivate the habit of dialectical thinking in writing.

[0331] 5. Suggestions on cultivating creative thinking in writing

[0332] First, based on the "target results" of creative thinking in composition, we sort out the creative thinking resources in primary school Chinese textbooks and carry out creative thinking teaching in composition.

[0333] The second is to focus on the "object content" of creative thinking in composition, dig deep into the creative thinking in classic articles, and transfer it to composition teaching.

[0334] The third is the "method ability" of creative thinking in writing, which is to connect what you have learned with real life and apply it, look at life from multiple perspectives, give full play to imagination to solve life problems, and express them creatively in words.

[0335] Fourth, focusing on the "strategic quality" of creative thinking in writing, we organize students to use creative thinking for original expression and cultivate the habit of creative thinking in writing.

[0336] (III) Students click the output button to print out the double-writing five-thinking training plan and download related training questions.

[0337] (IV) The teacher clicks the output button to print the double-writing five major thinking training lesson plan and download the relevant teaching design.

[0338] The present invention also provides a thinking training and evaluation system based on double writing training, comprising:

[0339] Writing format recognition module: used to obtain the strokes, structure, and format of the target general standard Chinese character writer in the grid, and use the square circle clock method to analyze the target standard Chinese character to obtain its writing ability information;

[0340] Writing thinking analysis module: It is used to analyze the writing ability of the target general standard Chinese character writer based on the five major thinking theories and from the perspective of digital intelligence vision, including intuitive thinking, image thinking, logical thinking, dialectical thinking and creative thinking;

[0341] Writing content recognition module: used to obtain information about the writing theme, keywords, layout, word choice, rhetoric, and references of the target general standard Chinese character writer, and to perform abnormal monitoring, including checking for duplicates, typos, grammatical errors, random smudges, etc., to obtain information about his or her writing ability;

[0342] The writing thinking analysis module is used to analyze the writing ability of the target general standard Chinese character writers based on the five major thinking theories and from the perspective of digital intelligence vision, including intuitive thinking, image thinking, logical thinking, dialectical thinking and creative thinking;

[0343] Double writing thinking evaluation module: It is used to comprehensively analyze the cognitive and non-cognitive influencing factors of the double writing thinking of the target general standard Chinese character writers based on the information of the first four modules, judge their internal and external control levels of thinking, and give improvement suggestions to promote the further improvement of the writers' double writing thinking ability and the level of "double beauty of characters and texts". For example, correct writing points, skillful strokes, and standardized format and layout indicate that the writer has high intuitive thinking ability; standard stroke shapes and coordinated overall format layout indicate that the writer has strong image thinking ability; orderly writing, clear strokes, neat paragraphs, neat paper, and no smudges indicate that the writer has strong logical thinking ability.

[0344] Enter the five thinking evaluation forms for primary school writing, and the computer can give writing evaluation for the composition. The specific steps are as follows:

[0345] (1) Select the texts for students’ writing exercises;

[0346] (2) The system first imports the essay to be evaluated into the screening module for data analysis.

[0347] (3) The system first determines whether the writer writes within the 1.0 cm square grid, whether he uses the square grid to find the correct position of the Chinese characters, and writes the strokes and structure correctly, thereby judging the writer's intuitive thinking level.

[0348] (4) The system begins to collect keywords from the title of the essay and the key content at the beginning, end, and middle of the essay to find the main point of the essay.

[0349] (5) Systematically analyze from the perspective of image, and judge from the content related to the six senses (sight, hearing, smell, touch, taste and movement) in the writing to see whether the writer has conducted in-depth analysis and thinking on common problems in life. Whether he has integrated a variety of information to transform intuitive phenomena into systematic problems. From the above dimensions, train and improve students' image thinking.

[0350] (6) From the perspective of categorical propositions, if the writer uses sentences with conjunctions to show the necessity of the content of the proposal, making the proposal more rational, it shows that the writer has a certain level of logical thinking. In addition, if the writer uses inductive and deductive reasoning methods, uses causal complex sentences or sentence groups to reveal the essential characteristics and regularity of "general" and "individual" things, making the expression more layered and clear, it shows that the writer has a high level of logical thinking.

[0351] (7) From the perspective of connection, the system determines whether the writer's ideas are derived from problems in life and whether they can solve actual problems. Secondly, from the perspective of contradiction, the system determines whether the writer pays attention to the difference between the writing format of the proposal and other styles, and whether the title and content are clear, concise and clear. The system can determine the writer's dialectical thinking from the above dimensions.

[0352] (8) The system analyzes whether the writer can boldly but reasonably associate the possible serious consequences during the writing process, highlighting the urgency and necessity of the initiative. The system analyzes whether the writer can use multidisciplinary knowledge, think about the problem from multiple angles, and propose reasonable and effective solutions. From the above dimensions, the system determines whether the writer's creative thinking is active, and proposes improvement methods.

[0353] (9) The essays that have undergone thinking weight analysis will be intelligently evaluated according to exclusive evaluation standards.

[0354] (10) Teachers or AI intelligent reviewers can make corrections through online review or sharing.

[0355] Thinking evaluation management module: responsible for storing and maintaining standardized Chinese character font information and scoring standards, exercise thinking language model and scoring standards.

[0356] Information recognition module: responsible for preprocessing uploaded images and converting image information into text and data information to facilitate subsequent evaluation and scoring.

[0357] Information analysis module: collects, stores and analyzes students' basic information and historical assessment results, generates personalized portraits of the five major thinking training levels of candidates, and identifies candidates' dual writing skill levels and weak links.

[0358] Assessment and Evaluation Module: Collect assessment results and evaluate and provide feedback on students' double writing five major thinking skills, and analyze the assessment results of candidates using machine learning methods such as random forest algorithm.

[0359] The system management module is used to manage system users, permission settings, data backup and recovery.

[0360] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.

Claims

1. A method for evaluating the level of double writing thinking training based on Chinese character recognition, characterized in that: The following steps are involved: S101: Collect and organize the commonly used regular script and semi-cursive Chinese characters in the national unified edition of primary school Chinese textbooks, including 3,500 first-level characters, 3,000 second-level characters and 1,605 third-level characters in the "General Standard Chinese Character Table", a total of 8,105 common standard Chinese characters in regular script and semi-cursive fonts; S102: Establish and maintain a square and round font standard five-thinking database containing the five-thinking weights and point data of common regular script and regular script standard Chinese characters; S103: Regularly update the five thinking databases of the square and round grid font specifications to keep pace with the latest textbook specifications and thinking weights; S104: Using reinforcement learning techniques, train the agent to optimize the weights of the five major thoughts; S105: Construct a database of five thinking language models for writing exercises; S106: Upload photos of students' homework and related basic information; S107: Extract the Chinese handwritten font information and text content information from students’ homework, generate five major thinking ability portraits, and analyze and evaluate the thinking level; S108: Provide personalized thinking training suggestions based on the analysis results.

2. The method for evaluating the level of double writing thinking training based on Chinese character recognition according to claim 1 is characterized in that: The step S101 includes: (1) 3,500 first-level characters, 3,000 second-level characters, and 1,605 third-level characters from the "General Standard Chinese Character List" were selected from the national unified edition of primary school Chinese textbooks to ensure that the representativeness of the selected Chinese characters covers the range of commonly used Chinese characters; (2) Classify the selected Chinese characters and sort out writing samples of regular script and running script respectively to ensure that the samples of each font are standardized and accurate; (3) Establish a database containing 8105 commonly used standard Chinese characters, which stores writing samples of regular script and semi-cursive script fonts respectively to facilitate subsequent recognition and analysis.

3. The method for evaluating the level of double writing thinking training based on Chinese character recognition according to claim 1 is characterized in that: The step S102 includes: (1) Initialize the database framework, including the standard Chinese character ASCII code recorded as Hi, the five major thinking weights, and point data; (2) Use a structured database management system to store and manage data; (3) Define the database table structure, including the standard Chinese character table T H , Five thinking weight table T W , Standard Chinese characters on the circle outside the square grid data table T P ; (4) Storing detailed information of the standardized Chinese characters in the standardized Chinese character table, including the standardized Chinese character national standard ASCII code Hi and the source of the standardized Chinese characters in the primary school Chinese textbooks; storing detailed information of the standardized Chinese characters in the standardized Chinese character table T H The detailed information of the standard Chinese characters is stored in the text book, including Hi and the position index of the Chinese characters in the textbook; (5) The five major thinking weight arrays mainly trained by the character are stored in the five major thinking weight table, that is, the proportion of the character in intuitive thinking, image thinking, logical thinking, dialectical thinking, and innovative thinking training. The standard Chinese characters appearing in the textbook are divided into several categories, and each category is assigned a weight according to the same label; it is recorded as the weight vector W i ; W i =[W i1 ,W i2 ,W i3 ,W i4 ,W i5 ], where W ij represents the weight of the i-th Chinese character in the j-th type of thinking training; The weights are assigned using a mathematical formula, which is calculated as follows: Among them, f ij is the frequency of the ith Chinese character in the jth type of thinking training; (6) The coordinates of the starting and ending points of the strokes near the outer circle line of the square circle grid are stored in the point position data table of the standard Chinese characters on the outer circle line of the square circle grid, where the point position data of the standard Chinese characters on the outer circle line of the square circle grid is recorded as the coordinate set P i, P i {(x ik ,,y ik ))k=1,2...,n}: (7) Standardize Chinese Character Table T H , Five thinking weight table T H , Standard Chinese characters on the circle outside the square grid data table T P To associate, through mathematical relationship, T H , T W and T P are interrelated to form a complete database architecture; assuming that T H The primary key of the table is H i d, T H Table by H i d and T H Table association, where H i d in T W As a foreign key in T P The table is also passed through H i d and T H Table association, H i d in T P It is also a foreign key; (8) Maintain and update the database to reflect the latest teaching material standards and thinking weights; (9) Output the constructed five-thinking database framework of square and round grid font specifications.

4. The method for evaluating the level of double writing thinking training based on Chinese character recognition according to claim 1 is characterized in that: The step S103 includes: (1) Regularly review the Chinese character samples and the five thinking weights in the database to ensure consistency with the latest textbook specifications and teaching requirements; (2) Update the database structure to adapt to new data requirements and analysis needs; (3) Record database update operations to track change history and ensure data consistency.

5. The method for evaluating the level of double writing thinking training based on Chinese character recognition according to claim 1 is characterized in that: The step S104 includes: (1) Define the objective function: The objective function J(θ) is expressed as: Where N is the number of samples in the training data, s i and a i is the state and action of the i-th sample, R(s i ,a i ) is the reward of the sample, Q(s i ,a i ,θ) is the Q value predicted by the agent, and θ is the agent parameter; (2) Constructing a Q-network: Use a neural network to approximate the Q function, which takes the state and action as input and outputs the Q value; (3) Establishing an experience replay buffer: During training, the observed state-action-reward-next-state pairs are stored in the experience replay buffer; (4) Strategy selection: In the actual environment, the agent uses a strategy π(a|s) to select actions. This strategy can be greedy or ε-greedy. (5) Training the Q network: Use stochastic gradient descent (SGD) to update the parameters θ of the Q network. In each iteration, a batch of samples are randomly selected and the target Q value Q′(s, a, θ′) is calculated, where θ′ is the parameter of the target network. The target network parameters are periodically copied from the Q network parameters. The target Q value is calculated as: Where γ is the discount factor, s′ is the next state, and the gradient for updating the Q network parameters is: (6) Weight update: Based on the output of the Q network, the agent adjusts the weights of the five major thoughts to minimize the objective function. The specific steps of weight update are: Among them, α w is the learning rate for weight updates.

6. The method for evaluating the level of double writing thinking training based on Chinese character recognition according to claim 1 is characterized in that: The step S105 includes: (1) Collection of theoretical content and requirements: Collect the theoretical content and requirements of composition in the unified edition of primary school Chinese textbooks, including the guiding principles of composition structure, content, and expression, to form a complete theoretical library as a benchmark for model training and evaluation; (2) Student composition collection: Collect a certain amount of student composition examples. These compositions should cover students of different grades and levels from grades 1 to 6 to ensure the generalization ability of the model. The compositions can be actual classroom assignments or composition competition works to ensure the diversity and representativeness of the compositions. (3) Collecting correction comments: Collecting corresponding correction comments, which come from teachers’ detailed comments and modification suggestions. Correction comments should cover the identification, judgment and modification suggestions of the five major thinking methods to facilitate subsequent model training and evaluation; (4) Corpus construction: The collected essay examples and comments are digitized and converted into a corpus that can be used for machine learning model training; (5) Model design: A neural network-based five-thinking language model is used. This model can identify and analyze the five thinking features in the composition. The model structure includes an encoder and a decoder for processing the composition text and extracting thinking features. The model uses a deep neural network structure, which contains multiple layers of neurons: Input layer: receives composition text data, converts it into a digital sequence, and converts the composition text into a fixed-length vector sequence through word embedding; Encoding layer: Use the Transformer encoder to encode the input sequence and extract the semantic features of the composition; Decoding layer: Use the Attention mechanism decoder to decode the features output by the encoding layer and further extract the five major thinking features; Output layer: Map the features output by the decoding layer into scores of the five major thinking features, with a rating of 1-5 points, where 1 represents the worst performance and 5 represents the best performance, indicating the performance of the essay in the five major thinking aspects; (6) Model training: The model is trained using the collected corpus, and the model parameters are updated through the back propagation algorithm. During the training process, the review comments are used as the supervision signal of the model to improve the accuracy of the model in identifying and analyzing the five major thinking. The cross entropy loss function method is used in the training process. During the training process, the calculation formula of the cross entropy loss function is: Among them, L is the loss function, y i is the true label, p i is the probability predicted by the model; (7) Database construction and maintenance: Store the trained model parameters, scoring criteria, and related theoretical content to build a database of the five major thinking and language models of writing exercises; update the database to incorporate new teaching theories and student writing data to maintain the timeliness and accuracy of the database.

7. The method for evaluating the level of double writing thinking training based on Chinese character recognition according to claim 1 is characterized in that: The step S106 includes: (1) Write Chinese characters in the square and round hour grid according to the standard points of regular script and running regular script. Practice writing repeatedly to form intuitive thinking, remember the corresponding points, and eventually break away from the square and round hour grid. You can find the points of the characters on the square grid and achieve the standard writing of the points and shapes of characters with "no grid in the eyes but a grid in the heart"; (2) According to the standard Chinese characters of regular script and running script, we can carry out scientific "radical" classification learning. We only need to remember the positions of 201 main radicals on the circle outside the square and circle clock grid, and learn 38 strokes, 16 stroke orders, and 52 structural writing rules. Then we can apply what we have learned to other cases and assemble and combine the components of Chinese characters to write the standard Chinese characters of regular script and running script. (3) Log in to the system and fill in personal information and writing experience for subsequent analysis of students’ writing skills and weaknesses in the five major thinking training areas; (4) Upload photos of your essays to the system. The photos should clearly show the full text of the essay to ensure the system’s recognition accuracy and evaluation quality.

8. The method for evaluating the level of double writing thinking training based on Chinese character recognition according to claim 1 is characterized in that: The step S107 includes: (1) Image preprocessing: receiving the homework photos uploaded by students and preprocessing the photos, including image grayscale conversion, image binarization, image denoising, using the Canny edge detection algorithm to identify the edges in the image, using the Hough transform to detect the straight lines in the image and thus identify the grid boundaries, determining the vertex coordinates of each grid by detecting the intersection of the straight lines, and dividing the image into individual grids according to the vertex coordinates for subsequent Chinese character recognition and analysis, so as to improve the effect of subsequent processing; The image grayscale conversion is to convert the color image of the homework uploaded by the students into a grayscale image, and use the following formula to calculate the grayscale value of each pixel: I gray (x,,y)=0.2989R(x,y)+0.5870G(x,y)+0.1140B(x,y) Among them I gray is the pixel value of the grayscale image at position (x, y), R, G, B are the pixel values ​​of the red, green, and blue channels of the original color image respectively; Image binarization is to convert a grayscale image into a binary image containing only black and white colors. The threshold segmentation method is used. The calculation formula is as follows: Among them, B(x,y) is the pixel value after binary, G(x,y) is the pixel value of the grayscale image, and T is the set threshold; Image denoising uses Gaussian filtering algorithm to reduce image noise. The calculation formula is as follows: Where I′(x,y) is the pixel value after denoising, I(x,y) is the pixel value of the original image, w(i,j) is the weight coefficient of the Gaussian kernel, and k is the size of the filter kernel; Use Hough transform to detect straight lines in the image and identify the boundaries of the grid. The calculation formula is: r=xcos(θ)+ysin(θ) Where (r,θ) is the parameter in Hough space, (x,y) is the coordinate of the edge point in image space; By detecting the intersection of the straight lines, the vertex coordinates of each grid are determined, and the image is divided into individual grids according to the vertex coordinates for subsequent Chinese character recognition and analysis. Assume that the equations of the two straight lines are: Line L1: Ax+By+C1=0 Straight line L2: A′x+B′y+C2=0 Then their intersection point (x, y) can be obtained by solving the following system of equations: According to these vertex coordinates, the image is segmented into individual grids for subsequent Chinese character recognition and analysis; (2) Chinese character recognition: Use optical character recognition (OCR) technology to perform Chinese character recognition on the preprocessed image, extract the handwritten font information of the Chinese characters, and represent the recognized Chinese characters as strings for subsequent analysis. The OCR recognition process can be represented as a function f OCR , which will preprocess the image I pre Mapped to the recognized Chinese character string S, the calculation formula is as follows: S=f OCR (I pre )S=f OCR (I pre ) (3) Text content analysis: The recognized Chinese characters are segmented and POS tagged to obtain information about the text content. Natural language processing technology is used to perform semantic analysis on the composition content and extract key information. The segmentation and POS tagging functions are used to seg and f tag Indicates that the string S is operated separately: T=f seg (S) P=f tag (T) Where T is the word sequence after word segmentation, and P is the corresponding part-of-speech tag sequence; (4) Extraction of the five major thinking features: Based on the definition of the five major thinking features, design the corresponding feature extraction algorithm, including logical thinking features and innovative thinking features. These features can be extracted based on the results of text content analysis; the feature extraction is represented by a series of functions f feat,i , each function corresponds to a thinking characteristic F i : F i =f feat,i (P) Where i represents the i-th thinking characteristic; (5) Thinking ability profile generation: The thinking features extracted from the text and image are integrated to generate a student's thinking ability profile, which is represented by a score; The generation of thinking ability portrait is represented by function f profile , which combines all the features F i : profile=f profile (F1,F2,…,F n ) Where n is the total number of thinking characteristics; (6) Analysis and evaluation of thinking level: Based on the thinking ability portrait, the scoring model is used to analyze the students' thinking level. The evaluation results give the students' overall performance in the five major thinking aspects and the specific scores of each thinking dimension. The scoring model of thinking level is represented by the function f score , generate a score S based on the thinking ability profile total ,but S total =f score (profile) The score for each thinking dimension is expressed as: St otal =f score,i (F i ) S i is the score of the i-th thinking dimension.

9. The method for evaluating the level of double writing thinking training based on Chinese character recognition according to claim 1 is characterized in that: The step S108 includes: (1) Based on the basic information and historical evaluation results of the original portraits of students' double writing of the five major thinking training, give personalized suggestions for the five major thinking training of the standard Chinese character shapes of general regular script and running regular script, and generate corresponding improvement plans; (2) Provide suggestions for training the five major thinking skills in writing and generate corresponding improvement plans; (3) Students click the output button to print the double writing five thinking training plan and download the relevant training questions.

10. A system for evaluating the level of double writing thinking training based on Chinese character recognition, characterized in that: Includes the following modules: A writing format recognition module is used to parse writing ability information; Writing thinking analysis module, used to analyze the five major thinking abilities; A writing content recognition module is used to obtain writing ability information; Writing thinking analysis module, used to analyze the five major thinking abilities; The dual writing thinking evaluation module is used to comprehensively analyze the dual writing thinking level; Thinking evaluation management module, used to store and maintain scoring criteria; An information recognition module, used for converting image information; Information analysis module, used to collect and analyze student information; Evaluation module, used to collect and analyze evaluation results; The system management module is used to manage system users, permission settings, data backup and recovery.

Images