A method and system for intelligent evaluation of handwriting ability based on handwriting input

Abstract

The application discloses a handwriting input-oriented writing ability intelligent evaluation method and system. The method comprises the following steps: obtaining a handwritten character picture input by a user; performing text recognition on the handwritten character picture based on a preset text recognition model, so as to convert the handwritten character picture into a candidate text; performing text correction on the candidate text based on a preset text correction model, so as to correct the candidate text into a formal text; performing text analysis on the formal text based on a preset text analysis model, so as to output a text analysis result for the user; and performing intelligent evaluation on the writing ability of the user according to the text analysis result. Thus, the method can realize a complete evaluation process of handwriting character recognition and text semantic analysis, is compatible with problems such as random writing and illegible handwriting, and greatly improves the robustness and practicability of the whole evaluation method while ensuring accuracy.

Description

Technical Field

[0001] This invention relates to an intelligent evaluation method for handwriting input, and also to a corresponding intelligent evaluation system for handwriting input, belonging to the field of data recognition technology. Background Technology

[0002] Writing ability assessment is an important component of the Mini-Mental State Examination (MMSE). It involves having a patient suspected of having cognitive impairment write a Chinese sentence that meets the following three criteria: (1) it has a subject; (2) it has a verb; and (3) it is semantically coherent. If the patient can write a sentence that meets these three criteria, they are considered to have basic writing ability, which is used to assess their cognitive level. Therefore, assessing a patient's writing ability is of significant reference value for diagnosing their condition.

[0003] Traditional handwriting ability assessments primarily rely on manual evaluation by professionals, which is time-consuming and susceptible to subjective bias. To address this issue, Chinese patent application CN111651999A discloses an automatic evaluation system for handwriting ability detection based on text semantic analysis. This system mainly includes a corpus input module, a training corpus acquisition module, a corpus preprocessing module, a grammatical integrity judgment module, a semantic fluency analysis module, and a database. This automatic evaluation system combines the grammatical integrity and semantic fluency of the Chinese sentences written by the test taker to determine whether the sentences are comprehensible, thereby assessing whether the test taker possesses basic handwriting ability.

[0004] However, a complete handwriting ability assessment scheme should include both handwriting input recognition and semantic analysis. The technical solutions in the aforementioned patent applications mainly focus on the latter, completely ignoring the handwriting input recognition part, and therefore cannot achieve automated intelligent assessment of users' handwriting abilities. Summary of the Invention

[0005] The primary technical problem to be solved by this invention is to provide an intelligent evaluation method for handwriting ability oriented towards handwriting input.

[0006] Another technical problem to be solved by the present invention is to provide an intelligent evaluation system for handwriting ability oriented towards handwriting input.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] According to a first aspect of the present invention, a method for intelligent evaluation of handwriting ability oriented towards handwriting input is provided, comprising the following steps:

[0009] Get images of handwritten text input by the user;

[0010] The handwritten text image is converted into candidate text based on a preset text recognition model.

[0011] The candidate text is corrected based on a preset text correction model to convert the candidate text into official text.

[0012] The formal text is analyzed based on a preset text analysis model to output text analysis results for the user.

[0013] The user's writing ability is intelligently assessed based on the text analysis results.

[0014] Preferably, the text recognition process includes:

[0015] Based on a preset text detection model, image features are extracted from the handwritten text image to extract the end-to-end text writing area of ​​variable size.

[0016] The text writing area is identified based on a preset text direction recognition model to determine the text direction of the text writing area; wherein the text direction includes at least four directions: up, down, left, and right.

[0017] The system performs character recognition on the text writing area based on a pre-defined character recognition model, and sorts the recognized characters based on the direction of the recognized text to output the correctly sorted candidate text.

[0018] The text detection model, text direction recognition model, and character recognition model together constitute the text recognition model.

[0019] Preferably, image feature extraction is performed on the handwritten text based on a preset text detection model to extract an end-to-end text writing region of variable size, specifically including:

[0020] The handwritten text image is normalized using the first image transformation module;

[0021] The first network feature extraction module extracts features from the normalized handwritten text image based on a convolutional recurrent neural network to extract the feature matrix of the text writing area.

[0022] The first feature enhancement module enhances the feature matrix of the text writing region.

[0023] The first network output module outputs the feature-enhanced text writing area.

[0024] Preferably, the text writing area is identified based on a preset text direction recognition model to determine the text direction of the text writing area, specifically including:

[0025] The second image transformation module performs normalization processing on the segmented image of the text writing area;

[0026] The second network feature extraction module extracts features from the normalized text writing region segmented image based on a convolutional recurrent neural network to extract the text writing direction feature matrix.

[0027] The text writing direction feature matrix is ​​enhanced using a second feature enhancement module.

[0028] The second network output module outputs the text direction of the segmented image of the text writing region after feature enhancement.

[0029] Preferably, the text writing area is recognized based on a preset character recognition model, and the recognized characters are sorted according to the direction of the recognized text to output correctly sorted candidate text, specifically including:

[0030] The third image transformation module performs normalization processing on the cut image after correcting the text direction.

[0031] The third network feature extraction module extracts features from the normalized image cut in the correct direction using a convolutional recurrent neural network to extract a variable-length text feature matrix.

[0032] The extracted variable-length text feature matrix is ​​enhanced using the third feature enhancement module.

[0033] The third network output module outputs the text that is correctly sorted according to the text direction, forming the correctly sorted candidate text.

[0034] Preferably, the third network output module selects the character with the highest probability as the recognition result, and uses the CTC loss function to optimize the text writing area so as to output the text with the correct orientation.

[0035] Preferably, the text correction process specifically includes:

[0036] When a text recognition model encounters illegible characters during the process of recognizing text, it provides multiple candidate characters based on confidence levels, forming a candidate character set.

[0037] Based on a pre-trained text correction model, the illegible characters in the candidate text are hollowed out for prediction, so as to predict multiple predicted characters at the hollowed-out positions and form a set of predicted characters.

[0038] If the candidate character set and the predicted character set have an intersection, then the intersection is taken as the character recognition result;

[0039] If the candidate character set and the predicted character set do not intersect, then the candidate character in the candidate character set that makes the sentence coherent is selected as the text recognition result; if none of the candidate characters in the candidate character set make the sentence coherent, then the predicted character in the predicted character set that makes the sentence coherent is selected as the text recognition result.

[0040] Repeat the above process to correct all illegible characters in the candidate text, thereby correcting the candidate document into official text.

[0041] Preferably, the text analysis process specifically includes:

[0042] The dependency relation tags of the formal text are output according to the syntactic analysis model; wherein, the dependency relation tags include at least subject-verb relation (SBV), verb-object relation (VOB), prepositional relation (POB), core relation (HED), and double object relation (DOB);

[0043] The part-of-speech tagging model outputs the part-of-speech tags of the formal text; wherein the part-of-speech tags include at least noun n, verb v, pronoun r, adjective a, adverb d, and punctuation mark w;

[0044] Based on the dependency relationship tags and the part-of-speech tags, determine whether the formal text is grammatically correct and meets the requirements of the scale items according to preset rules; if the determination result is yes, the user scores 1 point for the corresponding item in the scale; if the determination result is no, the user scores 0 points for the corresponding item in the scale.

[0045] Preferably, the preset rules include:

[0046] If the dependency relationship tags of the formal text include any combination of SBV+VOB, SBV+POB, SBV+DOB, and SBV+HED, and the part-of-speech tag of SBV is n or r, the part-of-speech tags of VOB, POB, and DOB are n or r, and the part-of-speech tag of HED is n, r, or v, then the judgment result is yes; otherwise, the judgment result is no.

[0047] According to a second aspect of the present invention, a handwriting ability intelligent assessment system for handwriting input is provided, comprising a processor and a memory, wherein the processor reads a computer program in the memory for performing the following operations:

[0048] Get images of handwritten text input by the user;

[0049] The handwritten text image is converted into candidate text based on a preset text recognition model.

[0050] The candidate text is corrected based on a preset text correction model to convert the candidate text into official text.

[0051] The formal text is analyzed based on a preset text analysis model to output text analysis results for the user.

[0052] The user's writing ability is intelligently assessed based on the text analysis results.

[0053] Compared with existing technologies, the intelligent evaluation method and system for handwriting ability based on handwriting input provided by this invention not only includes semantic analysis of text but also handwritten character recognition technology. This invention introduces multiple deep learning models for evaluation, solving problems such as careless handwriting and illegible characters, and is compatible with various handwriting situations. While ensuring accuracy, it greatly improves the robustness and practicality of the entire evaluation method. Attached Figure Description

[0054] Figure 1 A flowchart illustrating an intelligent evaluation method for handwriting input based on an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram of the composition structure of a deep learning model in an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram of the ERNIE 3.0 language model in an embodiment of the present invention;

[0057] Figure 4 This is a structural diagram of an intelligent evaluation system for handwriting input provided in an embodiment of the present invention. Detailed Implementation

[0058] The technical content of the present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0059] like Figure 1 As shown, the present invention provides an intelligent evaluation method for handwriting input-oriented writing ability, which specifically includes steps S1 to S5:

[0060] S1: Get an image of the handwritten text entered by the user on the electronic screen.

[0061] Specifically, when users take a handwriting ability assessment, they need to input handwritten text on an electronic screen using a stylus according to the scale requirements. After input, an image of the user's handwritten text will be generated on the electronic screen for subsequent text recognition and analysis.

[0062] For example, in one embodiment of the present invention, if the user inputs the information "I am receiving treatment at the hospital", then the text image "I am receiving treatment at the hospital" will be generated after the input is completed.

[0063] S2: Based on a preset text recognition model, perform text recognition on handwritten text images to convert the handwritten text images into candidate text.

[0064] After a user completes the input of handwritten text, a pre-defined text recognition model is used to recognize the text in the handwritten text image. This text recognition model consists of a text detection model, a text direction recognition model, and a character recognition model. The text detection model extracts image features from the handwritten text image to extract the end-to-end text writing area; the text direction recognition model identifies the direction of the text writing area; and the character recognition model recognizes the characters in the text writing area and sorts the recognized characters based on the identified text direction to output the correctly sorted candidate text.

[0065] It is understood that, in one embodiment of the present invention, the text detection model, the text direction recognition model, and the character recognition model are all constructed using the same deep learning model. For example... Figure 2 As shown, this deep learning model consists of four parts: image transformation, network feature extraction, feature enhancement, and network output. In other words, the text detection model, text direction recognition model, and character recognition model all consist of these four parts, differing only in the specific function of each part.

[0066] Specifically, the text recognition process includes steps S21 to S23:

[0067] S21: Text detection.

[0068] Specifically, steps S211 to S214 are included:

[0069] S211: The handwritten text image is normalized by the first image transformation module to eliminate interference from image resolution, text background color, etc.

[0070] S212: The first network feature extraction module extracts features from the normalized handwritten text image based on a convolutional recurrent neural network to extract the text writing region feature matrix. Specifically, this first network feature extraction module uses a convolutional recurrent neural network. Convolutional networks are effective and fast for image feature extraction, while recurrent neural networks are suitable for handling text-related tasks in natural language processing. The combination of both, the convolutional recurrent neural network, demonstrates good performance in extracting end-to-end variable-length text writing regions.

[0071] S213: Enhance the features of the text writing region feature matrix through the first feature enhancement module;

[0072] S214: Output the feature-enhanced text writing area through the first network output module.

[0073] S22: Text direction detection.

[0074] Specifically, steps S221 to S224 are included:

[0075] S221: Normalize the segmented image of the text writing area through the second image transformation module;

[0076] S222: The second network feature extraction module extracts features from the normalized text writing area segmented image based on the convolutional recurrent neural network to extract the text writing direction feature matrix;

[0077] S223: Enhance the text writing direction feature matrix using the second feature enhancement module;

[0078] S224: The text direction of the segmented image of the text writing region after feature enhancement is output through the second network output module. This second network output module outputs the direction classification, reducing the dimension to 4 dimensions through a fully connected layer, representing four categories: up, down, left, and right.

[0079] S23: Text recognition.

[0080] Specifically, steps S231 to S234 are included:

[0081] S231: The third image transformation module performs normalization processing on the cut image after correcting the text direction;

[0082] S232: The third network feature extraction module extracts features from the normalized image cut in the correct direction based on the convolutional recurrent neural network to extract a variable-length text feature matrix;

[0083] S233: The extracted variable-length text feature matrix is ​​enhanced by the third feature enhancement module;

[0084] S234: The third network output module outputs the correctly sorted text according to the text direction, forming the correctly sorted candidate text. In this step S234, the third network output module uses a fully connected layer to reduce the dimension from n... 512 dimensions increased to n The algorithm has 5529 dimensions, where n represents characters of variable length, and 5529 represents the predicted probability of each Chinese character, punctuation mark, and English character. The character with the highest probability is selected as the recognition result, and the CTC (Connectionist Temporal Classification) loss function is used to optimize the text writing region to output characters correctly ordered in the text direction.

[0085] The CTC loss function optimizes the text writing region as follows:

[0086] When the input is x, the output of LSTM is x. The probability is:

[0087]

[0088] π∈B -1 (l) represents all values ​​after B transformation. The path π.

[0089] For any path π, we have:

[0090]

[0091] For path π1 with T=12:

[0092]

[0093] The CTC loss function adjusts the parameters w of the LSTM through gradients, so that the input sample π∈B -1 p(l|x) is maximum when (l) is present.

[0094] The value of a certain element in the input y matrix of the CTC loss function. tk The probability of:

[0095]

[0096] Where, α t (l k ), β t (l k ) is a constant calculated recursively.

[0097] S3: Based on a preset text correction model, perform text correction on the candidate text to correct it into official text.

[0098] Specifically, this includes steps S31 to S34:

[0099] S31: Generate a set of candidate characters

[0100] When a text recognition model encounters illegible characters during the process of recognizing text, it provides multiple candidate characters based on confidence levels, forming a candidate character set.

[0101] S32: Generate the predicted character set

[0102] Based on a pre-trained text correction model, the model performs cutout prediction on illegible characters in candidate text, predicting multiple characters at the cutout positions to form a set of predicted characters.

[0103] Among them, such as Figure 3 As shown, in one embodiment of the present invention, the pre-trained text correction model is the ERNIE 3.0 language model. In other embodiments, it can be adaptively replaced with other language models as needed.

[0104] S33: Output text recognition results

[0105] Specifically, if there is an intersection between the candidate character set and the predicted character set, the intersection will be used as the character recognition result.

[0106] If there is no intersection between the candidate character set and the predicted character set, the candidate character in the candidate character set that makes the sentence coherent is selected as the text recognition result; if none of the candidate characters in the candidate character set make the sentence coherent, the predicted character in the predicted character set that makes the sentence coherent is selected as the text recognition result.

[0107] S34: Repeat the above process to correct all illegible characters in the candidate text, thereby correcting the candidate document into official text.

[0108] S4: Perform text analysis on the formal text based on the preset text analysis model to output text analysis results tailored to the user.

[0109] Specifically, this includes steps S41 to S43:

[0110] S41: Obtain dependency relationship tags

[0111] Based on the syntactic analysis model, important information such as keywords, subjects, predicates, and objects in the formal text is obtained, thereby outputting dependency relation labels for the formal text; wherein, the dependency relation labels include at least subject-predicate relation (SBV), verb-object relation (VOB), prepositional object relation (POB), core relation (HED), and double object relation (DOB).

[0112] S42: Obtain Part-of-Speech Tags

[0113] The keywords in the formal text are tagged with part-of-speech tags according to the part-of-speech tagging model, thereby outputting the part-of-speech tags of the formal text; among them, the part-of-speech tags include at least noun n, verb v, pronoun r, adjective a, adverb d, and punctuation mark w.

[0114] S43: Output Results

[0115] Based on the dependency relationship tags and part-of-speech tags mentioned above, the system determines whether the formal text is grammatically correct and meets the requirements of the scale items according to preset rules. If the determination result is yes, the user scores 1 point for the corresponding item in the scale; if the determination result is no, the user scores 0 points for the corresponding item in the scale. The preset rules are as follows:

[0116] If the dependency tags of the formal text include any combination of SBV+VOB, SBV+POB, SBV+DOB, and SBV+HED, and the part-of-speech tag of SBV is n or r, the part-of-speech tags of VOB, POB, and DOB are n or r, and the part-of-speech tag of HED is n, r, or v, then the judgment result is yes. Otherwise, the judgment result is no.

[0117] S5: Intelligent assessment of users' writing ability based on text analysis results.

[0118] Based on the user's text analysis results, we can know the user's scale score in writing ability, and then compare it with the norm standard to know the evaluation result of the user's writing ability.

[0119] Based on the aforementioned intelligent evaluation method for handwriting input, this invention further provides an intelligent evaluation system for handwriting input. For example... Figure 4 As shown, the intelligent writing ability assessment system includes one or more processors 21 and a memory 22. The memory 22 is coupled to the processors 21 and is used to store one or more programs. When the one or more programs are executed by the one or more processors 21, the one or more processors 21 implement the intelligent writing ability assessment method for handwriting input as described in the above embodiment.

[0120] The processor 21 controls the overall operation of the intelligent handwriting assessment system to complete all or part of the steps of the aforementioned intelligent handwriting assessment method for handwriting input. The processor 21 can be a central processing unit (CPU), graphics processing unit (GPU), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), digital signal processing (DSP), etc. The memory 22 stores various types of data to support the operation of the intelligent handwriting assessment system. This data may include, for example, instructions for any application or method operating on the intelligent handwriting assessment system, and application-related data. The memory 22 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, etc.

[0121] In one exemplary embodiment, the evaluation system may be implemented by a computer chip or physical entity, or by a product with certain functions, to perform the above-described intelligent evaluation method for handwriting input ability and achieve the same technical effect as the method described above. A typical embodiment is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, an in-vehicle human-computer interaction device, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0122] In another exemplary embodiment, the present invention also provides a computer-readable storage medium including program instructions, which, when executed by a processor, implement the steps of the intelligent evaluation method for handwriting ability oriented towards handwriting input in any of the above embodiments. For example, the computer-readable storage medium may be the memory including the program instructions described above, which may be executed by the processor of the evaluation system to complete the intelligent evaluation method for handwriting ability oriented towards handwriting input described above, and achieve the same technical effects as the method described above.

[0123] In summary, the intelligent evaluation method and system for handwriting ability based on handwriting input provided by the embodiments of the present invention have the following beneficial effects:

[0124] 1. This invention includes not only semantic analysis of text, but also handwritten character recognition technology;

[0125] 2. This invention introduces multiple deep learning models for evaluation, which solves problems such as careless handwriting and illegible fonts, and is compatible with various handwriting situations. While ensuring accuracy, it greatly improves the robustness and practicality of the entire evaluation method.

[0126] The above provides a detailed description of the intelligent evaluation method and system for handwriting input provided by this invention. Any obvious modifications made by those skilled in the art without departing from the essence of this invention will constitute an infringement of the patent rights of this invention and will incur corresponding legal liability.

Claims

1. A handwriting ability intelligent evaluation method for handwriting input, characterized by, The method comprises the following steps: acquiring a handwritten character image input by a user on an electronic screen; performing text recognition on the handwritten character image based on a preset text recognition model to convert the handwritten character image into a candidate text; the text recognition process comprises: performing image feature extraction on the handwritten character image based on a preset text detection model to extract an end-to-end text writing region of an indefinite size; performing direction recognition on the text writing region based on a preset text direction recognition model to recognize the text direction of the text writing region; wherein the text direction comprises at least four directions of up, down, left and right; performing character recognition on the text writing region based on a preset character recognition model, and sorting the recognized characters based on the recognized text direction to output a correctly sorted candidate text; specifically comprising: performing normalization processing on the cut image after the text direction is corrected through a third image transformation module; performing feature extraction on the correctly oriented cut image after normalization processing based on a convolutional recurrent neural network through a third network feature extraction module to extract an indefinite-length character feature matrix; performing feature enhancement on the extracted indefinite-length character feature matrix through a third feature enhancement module; outputting the characters sorted in the correct text direction through a third network output module to form the correctly sorted candidate text; wherein the text detection model, the text direction recognition model and the character recognition model jointly constitute the text recognition model; performing text correction on the candidate text based on a preset text correction model to correct the candidate text into a formal text; the text correction process specifically comprises: when a character with illegible handwriting is encountered during the process of recognizing characters by the text recognition model, multiple candidate characters are given according to the confidence level to form a candidate character set; based on the pre-trained text correction model, the characters with illegible handwriting in the candidate text are predicted to be hollow to predict multiple predicted characters in the hollow position to form a predicted character set; if the candidate character set and the predicted character set have an intersection, the intersection is taken as the character recognition result; if the candidate character set and the predicted character set have no intersection, the candidate character in the candidate character set that makes the sentence smooth is selected as the character recognition result; if all candidate characters in the candidate character set cannot make the sentence smooth, the predicted character in the predicted character set that makes the sentence smooth is selected as the character recognition result; the above process is repeated to correct all characters with illegible handwriting in the candidate text, so as to correct the candidate file into a formal text; performing text analysis on the formal text based on a preset text analysis model to output a text analysis result for the user; performing intelligent writing ability evaluation on the user according to the text analysis result.

2. The method of claim 1, wherein the writing ability intelligent assessment method is characterized by performing image feature extraction on the handwritten character based on a preset text detection model to extract an end-to-end text writing region of an indefinite size specifically comprises: performing normalization processing on the handwritten character image through a first image transformation module; The first network feature extraction module extracts features of the normalized handwritten text picture based on a convolutional recurrent neural network to obtain a text writing area feature matrix; The first feature enhancement module enhances features of the text writing area feature matrix; The first network output module outputs the text writing area after feature enhancement.

3. The method of claim 2, wherein the writing ability intelligent assessment method is characterized by The text writing area is subjected to direction recognition based on a preset text direction recognition model to recognize the text direction of the text writing area, which specifically includes: The second image transformation module normalizes the cut picture of the text writing area; The second network feature extraction module extracts features of the normalized cut picture of the text writing area based on a convolutional recurrent neural network to obtain a text writing direction feature matrix; The second feature enhancement module enhances features of the text writing direction feature matrix; The second network output module outputs the text direction of the cut picture of the text writing area after feature enhancement.

4. The intelligent writing ability evaluation method of claim 3, characterized in that: The third network output module increases the dimension from to wherein n represents indefinite length of text, and 5529 represents the prediction probability of each character of Chinese characters, punctuation, and English. The character with the maximum probability is selected as the recognition result, and a CTC loss function is used to optimize the function of the text writing area to output the text direction of the correct sequence of characters; The CTC loss function optimizes the function of the text writing area as follows: The probability of the output of the LSTM given the input x is: ​ π e B -1 (l) represents all paths π that are transformed by B into π; For any path π, there is: For the path π1 of T = 12, there is: The CTC loss function adjusts the parameters w of the LSTM by gradient, so that the input sample π ∈ B -1 (l) when p(l|x) is maximum; The value y of a certain one of the input y matrix of the CTC loss function tk The probability: where α t (l k ), β t (l k ) are constants computed by recursion.

5. The intelligent assessment method of writing ability as claimed in claim 1, wherein The text analysis process specifically includes: According to the dependency relation label of the formal text output by the syntax analysis model, the dependency relation label at least includes subject-verb relation SBV, verb-object relation VOB, preposition-object relation POB, core relation HED and double object relation DOB; According to the part-of-speech tag of the formal text output by the part-of-speech tagging model, the part-of-speech tag at least includes noun n, verb v, pronoun r, adjective a, adverb d and punctuation w; According to the dependency relation label and the part-of-speech tag, it is determined whether the formal text is grammatically correct and meets the requirements of the scale question according to the preset rule. If the result is yes, the user scores 1 point in the corresponding item of the scale, and if the result is no, the user scores 0 point in the corresponding item of the scale.

6. The intelligent assessment method of writing ability as claimed in claim 5, wherein The preset rule includes: If the dependency relation label of the formal text includes any one of the combinations of SBV+VOB, SBV+POB, SBV+DOB and SBV+HED, and the part-of-speech tag of SBV belongs to n or r, the part-of-speech tag of VOB, POB and DOB belongs to n or r, and the part-of-speech tag of HED belongs to n or r or v, the result is yes; otherwise, the result is no.

7. A handwriting ability intelligent evaluation system for handwriting input, characterized by, The processor reads the computer program in the memory for executing the writing ability intelligent evaluation method of any one of claims 1-6. The processor reads the computer program in the memory for executing the writing ability intelligent evaluation method of any one of claims 1-6.

Images