An online intelligent marking system based on big data

Abstract

The present invention belongs to the technical field of online intelligent marking of students' homework. Specifically, an online intelligent marking system based on big data is provided, including: identifying each blurred character in each student's homework, comparing the stroke writing characteristics and arrangement patterns of each blurred character with those of its similar characters, and determining the calibrated characters of each blurred character in each student's homework; comparing the corresponding answer writing outlines of the subjective questions in each student's homework with each standard answer on the network public platform and with the reference answers at the same time, identifying the irrationality indexes of the similar parts between the subjective question answers of each student's homework and the network standard answers, and accordingly determining the network citation score of the student; comparing the homework answers among students with each other to screen out each suspected student whose homework answers may have cross-citation, and then detecting the homework progress index by identifying the historical homework submission records of each suspected student, and determining the homework cross-citation score of each student.

Description

Technical Field

[0001] The present invention belongs to the technical field of online intelligent marking of students' homework, and relates to an online intelligent marking system based on big data. Background Art

[0002] With the booming rise of online education, the traditional teaching mode and homework marking method are difficult to meet the growing personalized learning needs and efficient management requirements. In the traditional education system, teachers need to spend a lot of time manually correcting homework, which not only affects teaching efficiency but also limits teachers' investment in teaching innovation, personalized tutoring, etc. Against this background, the technology of "online intelligent marking of students' homework" has emerged. The online intelligent marking technology can automatically process a large number of basic marking tasks, effectively reducing the burden on teachers and enabling them to focus more on improving teaching quality and guiding students' personalized development. It has become an important invention and breakthrough in the field of education.

[0003] Some related solutions for online homework marking in the prior art, for example, a Chinese patent application for an online homework intelligent marking system and method with the publication number CN109903014A. It plans the homework format so that the big data analysis module can accurately extract data from the data storage module for processing, quickly generate different charts, enabling students, parents, teachers, and principals to quickly master the learning situation. In the wrong question bank, students can retrieve unqualified knowledge points for review according to chapters / units / sections or chronological order; by logging in through different ports such as the student login port, teacher login port, parent login port, and principal login port, it meets the operation needs of different personnel and is convenient to use. This can provide materials for big data analysis, collect data in a companion manner, accurately portrait the learning situation of students, and thus form a learning situation report for a certain time or stage, enabling students, parents, teachers, and principals to clearly understand the learning and education situation.

[0004] Another Chinese patent application for an objective question automatic scoring method and system based on blockchain with the patent number CN114298875A. It creates an automatic scoring platform based on the blockchain framework. The operation node reads the student information and homework information, screens out the student information to be marked and the objective question information to be marked, and writes the student information to be marked and the objective question information to be marked into the automatic scoring platform; the marking node reads the objective question information to be marked and the correct answers, corrects the objective question information to be marked based on bitwise operations, and obtains the marking result; the statistical node calculates the objective question score based on the marking result, generates an objective question score table in combination with the student information to be marked, and writes the score table into the automatic scoring platform.

[0005] Although the above solutions propose some methods for online marking of students' homework, there are still the following limitations: 1. The existing technology lacks the ability to correct and evaluate subjective questions. When evaluating the answers to questions, it mainly compares the written answers of students with the reference answers, without considering the impact of the unclear handwriting of students on the answer evaluation, which may lead to students ignoring the value of handwriting.

[0006] 2. It does not identify the citation of online answers and interactive answers among students in their homework, which may lead to students ignoring the value of academic integrity. Summary of the Invention

[0007] In view of this, to solve the problems raised in the above background technology, an online intelligent marking system based on big data is proposed.

[0008] The object of the present invention can be achieved by the following technical solutions: The present invention provides an online intelligent marking system based on big data, which includes: a homework information extraction module: used to count the homework sets of all students and classify the questions in each student's homework, including subjective questions and objective questions, where the objective questions include fill-in-the-blank questions and multiple-choice questions, and the student numbers are 1, 2,...r..., n.

[0009] A handwriting recognition module: used to scan and obtain the answers to various questions in each student's homework, extract the writing outline of the answers, and then identify the written scores of each student's homework.

[0010] An identification accuracy correction module: used to screen out the unclear characters and letters in each student's homework that do not meet the standards, and then accurately calibrate the outline of the fonts that do not meet the standards.

[0011] A network citation scoring module: used to match the subjective question answers of each student's homework with the publicly available network answers, list the similar parts and their sources in the subjective question answers of each student's homework, evaluate the irrationality index of the similar parts in the subjective question answers of each student's homework, and determine the network citation score D' of the subjective question answers of each student's homework r .

[0012] A student interaction scoring module: used to compare the subjective question answers of each student with each other, and determine the homework interaction citation score D″ of each student r .

[0013] A result feedback module: used to obtain the contour similarity between the answers to various questions in each student's homework and the reference answers, and determine the final score of each student's homework.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) By identifying the blurred characters in each student's homework that do not meet the standard of normalization, and comparing the stroke writing characteristics and arrangement patterns of each blurred character with those of its similar characters, the calibration characters of each blurred character in each student's homework are determined, reducing the impact of the blurred handwriting of the students themselves on the answer evaluation, and thus increasing the accuracy of the student homework grading.

[0015] (2) By identifying the answer writing normalization indicators of each student's homework, the written scores of each student's homework are determined. Students can better understand their writing habits and weaknesses, and thus cultivate the ability of self-reflection, which is very important for personal growth and development.

[0016] (3) By comparing the corresponding answer writing outlines of the subjective question answers in each student's homework with the standard answers on the network public platform, and at the same time comparing the writing outlines of the reference answers, the irrationality indicators of the similar parts between the subjective question answers in each student's homework and the standard answers on the network public platform are identified, which can objectively evaluate the students' citation situations, avoid the errors caused by subjective judgment, make the evaluation process more fair, and accordingly determine the network citation scores of the students. This evaluation method encourages students to rely more on their own understanding and analysis when completing their homework, reduces the reliance on direct copying and pasting, thereby promoting the improvement of critical thinking and independent thinking abilities, and effectively detecting whether there is plagiarism or dishonest behavior in the students' homework.

[0017] (4) By comparing the homework answers among students, the suspected students whose homework answers may have cross-references are screened, and then by identifying the historical homework submission records of each suspected student, the homework progress indicators are detected, and accordingly the homework cross-reference scores of each student are determined, increasing the reliability of the final grading of the student homework. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0019] Figure 1 It is a schematic diagram of the connection of the system modules of the present invention.

[0020] Figure 2 It is a schematic diagram of the blurred characters of the present invention.

[0021] Figure 3 It is a schematic diagram of the blurred letters of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] Please refer to Figure 1 As shown, the present invention provides an online intelligent marking system based on big data. The system includes: a homework information extraction module, a writing recognition module, a recognition accuracy correction module, a network citation scoring module, a student interaction scoring module, and result feedback.

[0024] The homework information extraction module is connected to the writing recognition module, the writing recognition module is connected to the recognition accuracy correction module, the recognition accuracy correction module is connected to the network citation scoring module, the network citation scoring module is connected to the student interaction scoring module, and the student interaction scoring module is connected to the result feedback.

[0025] The homework information extraction module is used to count the homework sets of all students and classify the questions of each student's homework, including subjective questions and objective questions. Among them, the objective questions include fill-in-the-blank questions and multiple-choice questions, and the student numbers are 1, 2,... r..., n.

[0026] The writing recognition module is used to scan and obtain the answers to various questions in each student's homework, extract the writing outline of the answers, and then recognize the written score of each student's homework.

[0027] In a preferred implementation manner, the specific method for recognizing the written score of each student's homework is as follows: Match the corresponding answer writing outline of the answers to various questions in each student's homework with all the characters and letters in the font library. If a certain character or letter in the corresponding answer writing outline of a certain type of question answer in a student's homework has a unique matching character or letter in the font library, the normative weight of this character or letter is recorded as A1, such as A1 = 2. If there are multiple matching characters or letters, the normative weight of this character or letter is recorded as A2, such as A2 = 1. If there is no matching character or letter, the normative weight of this character or letter is recorded as A3, such as A3 = 0.

[0028] The recognition standard for the matching of characters or letters in the answer writing outline with the font library is that the similarity of the stroke writing outline is within a specified range.

[0029] The method for obtaining the similarity of the stroke writing outlines is as follows: match each stroke of each character in the answer writing outline with each stroke of each Chinese character in the font library, extract the comprehensive difference values of features such as the length, width, and inclination angle of the strokes, and then obtain the inverse ratio of the comprehensive difference value to the preset reference benchmark difference value, which is recorded as the similarity of the stroke writing outlines of each character in the answer writing outline and each Chinese character in the font library.

[0030] Statistically obtain the comprehensive normative weights corresponding to all characters and letters in the answer writing outlines of all questions in each student's homework, and obtain the ratio of it to the preset reference normative weight, which is recorded as the answer writing normativity index of each student's homework.

[0031] Further, match the answer writing normativity index of each student's homework with the corresponding written scores within the preset range of each writing normativity index to determine the written scores of each student's homework.

[0032] The present invention determines the written scores of each student's homework by identifying the answer writing normativity index of each student's homework. Students can better understand their writing habits and weaknesses, and then cultivate the ability of self-reflection, which is very important for personal growth and development.

[0033] Please refer to Figure 2 、 3 As shown, the recognition accuracy correction module is used to screen out each fuzzy character and each fuzzy letter in each student's homework that do not meet the norms, and then accurately calibrate the outlines of the fonts that do not meet the norms.

[0034] In a preferred implementation manner, the specific screening method for each fuzzy character and each fuzzy letter in each student's homework that do not meet the norms is as follows: extract the characters or letters with normative weights of A2 and A3 in the answer writing outlines of various questions in each student's homework, and then mark each character with a normative weight of A2 and A3 as each fuzzy character, and mark each letter with a normative weight of A2 and A3 as each fuzzy letter.

[0035] In a further preferred implementation manner, the accurate calibration of the outlines of the fonts that do not meet the norms includes: extracting each character in the font library that matches each fuzzy character in each student's homework that does not meet the norms and its outline, which is recorded as each similar character, and extracting the similarity between each fuzzy character in each student's homework and its corresponding similar character from the similarity of the stroke writing outlines of each character in the answer writing outline and each Chinese character in the font library.

[0036] Based on the contour edge detection technology, the habitual writing features and habitual arrangements of each stroke of each student are identified from the writing contours of the corresponding answers to various questions in the students' homework, such as horizontal, vertical, dot, left-falling stroke, right-falling stroke, hook, and turn strokes, writing features such as contour length, width, and inclination angle, arrangement methods such as the distance between adjacent strokes, the position of the starting and ending points of the pen compared to the overall area of the text, and the thickness and length of the connecting contour with adjacent strokes.

[0037] Obtain the writing contours corresponding to each fuzzy character that does not meet the standardization standards in each student's homework, extract the fixed-frame strokes corresponding to each fuzzy character in each student's homework, identify the writing features to which they belong, compare them with the inertial writing features of each stroke by the corresponding student, obtain the difference values between the fixed-frame strokes corresponding to each fuzzy character in each student's homework and the inertial writing features of each stroke, sum them up to obtain the comprehensive difference values between the fixed-frame strokes corresponding to each fuzzy character in each student's homework and each stroke, screen out the strokes with the minimum comprehensive difference value, and determine them as the sub-categorized writing trend strokes of the fixed-frame strokes corresponding to the corresponding fuzzy character in the corresponding student's homework, obtain the sub-categorized writing trend strokes of the fixed-frame strokes corresponding to each fuzzy character in each student's homework, reorganize and construct the sub-categorized writing trend strokes of the fixed-frame strokes corresponding to each fuzzy character in each student's homework, and obtain the sub-categorized writing trend characters of each fuzzy character in each student's homework.

[0038] From the writing contours corresponding to the fuzzy characters that do not meet the standardization standards in the assignments of each student, identify the arrangement modes of the fixed-frame strokes corresponding to each fuzzy character in each student's assignment, match them with the inertial arrangement modes of the strokes of the corresponding students, assign weights to the matched arrangement modes, and count the total weight assignments of the matches between the fixed-frame strokes corresponding to each fuzzy character in each student's assignment and the arrangement modes of each stroke, and then determine the stroke to which the maximum total weight assignment belongs as the supplementary writing trend stroke of the fixed-frame stroke of the corresponding fuzzy character in the assignment of the corresponding student, and obtain the supplementary writing trend strokes of the fixed-frame strokes corresponding to each fuzzy character in the assignments of each student, and then reconstruct and obtain the supplementary writing trend Chinese characters of each fuzzy character in the assignments of each student.

[0039] Specifically, the matching arrangement modes are weighted, such as: if an arrangement mode of a fixed-frame stroke corresponding to a fuzzy word in a student's homework matches a habitual arrangement mode of a stroke of the corresponding student, the weight of the arrangement mode is assigned to 1, and if it does not match, the weight is assigned to 0.

[0040] Assign similarity values to the Chinese characters with a tendency towards fine classification writing and the Chinese characters with a tendency towards supplementary writing, and accumulate them to the similarity between each fuzzy character and its corresponding similar characters in each student's assignment, to obtain the final similarity value between each fuzzy character and its corresponding similar characters in each student's assignment. Furthermore, take the similar character corresponding to the maximum value of the final similarity value of each fuzzy character in each student's assignment as the calibrated character for the corresponding fuzzy character in the corresponding student's assignment.

[0041] The assignment of similarity values to the Chinese characters with a tendency towards fine classification writing and the Chinese characters with a tendency towards supplementary writing is as follows: assign the similarity value of the Chinese characters with a tendency towards fine classification writing as z1, and assign the similarity value of the Chinese characters with a tendency towards supplementary writing as z2. z1 and z2 are set constants, for example, z1 = 0.4 and z2 = 0.6.

[0042] In a further preferred embodiment, the precise calibration of the font outlines that do not meet the standard of normativity further includes: extracting each letter and its outline in the font library that match each fuzzy letter in the student assignments that do not meet the standard of normativity, and recording them as each similar letter.

[0043] According to the calibration character recognition method for each fuzzy character in each student's assignment, similarly determine the calibrated letter for each fuzzy letter in each student's assignment.

[0044] The present invention determines the calibrated characters for each fuzzy character in each student's assignment by identifying each fuzzy character in each student's assignment that does not meet the standard of normativity, and comparing the stroke writing characteristics and arrangement methods of each fuzzy character with its corresponding similar characters, reducing the influence of the fuzzy writing of the student's own handwriting on the answer evaluation, and thus increasing the accuracy of the student assignment marking.

[0045] The network citation scoring module is used to match the subjective question answers of each student's assignment with the publicly available network answers, list each network similar part and its source in the subjective question answers of each student's assignment, evaluate the irrationality index of the similar parts of the subjective question answers of each student's assignment, and determine the network citation score D' of the subjective question answers of each student's assignment. r 。

[0046] In a preferred embodiment, the process of listing the network similar parts and their sources in the subjective question answers of each student's homework is as follows: Divide the corresponding answer writing outlines of the subjective question answers of each student's homework into text groups to obtain the corresponding text group answer writing outlines of the subjective question answers of each student's homework. Count the standard answers of each subjective question on the network public platform, such as the answers obtained by searching according to different combination modes of each keyword in the question on chat, and the answers obtained by searching various question-solving software, and integrate them into the text outline of the standard answers of each subjective question. Match it with the corresponding text group answer writing outlines of the subjective question answers of each student's homework. According to the similarity acquisition method of the stroke writing outline, similarly obtain the similarity between the corresponding text group answer writing outlines of the subjective question answers of each student's homework and the text outlines of the standard answers of the corresponding questions. Screen out the text group outline parts where the similarity between the subjective question answers of each student's homework and the text outlines of the standard answers of the corresponding questions is within the preset range, which are the network similar parts in the subjective question answers of each student's homework, and mark the text outline of the corresponding standard answer as the source of the corresponding network similar part.

[0047] In a further preferred embodiment, the irrationality index for evaluating the similar parts of the subjective question answers of each student's homework is specifically as follows: Obtain various types of reference answers pre-submitted by teachers on the grading platform, scan to obtain the writing outlines of various types of reference answers, and compare them with the network similar parts in the subjective question answers of each student's homework. According to the similarity acquisition method of the stroke writing outline, similarly identify the similarity between the corresponding answer writing outlines of the network similar parts in the subjective question answers of each student's homework and the writing outlines of the corresponding parts in the reference answers of the corresponding questions. If the similarity between the corresponding answer writing outline of a network similar part in the subjective question answer of a certain student's homework and the writing outline of the corresponding part in the reference answer of the corresponding question is within the set standard range, mark the network similar part in the subjective question answer of this student as similar and reasonable; otherwise, perform citation weighting on the network similar part in the subjective question answer of this student.

[0048] Set the corresponding weighting value g when performing citation weighting on the network similar parts in the subjective question answers of students. g is a set constant, such as g = 1. Based on this, obtain the citation weighting values of all network similar parts in the subjective question answers of all students, sum them to obtain the comprehensive citation weighting value of the subjective question answers of each student, obtain the ratio of it to the preset reference citation weight value, and obtain the irrationality index of the similar parts of the subjective question answers of each student's homework.

[0049] In a further preferred embodiment, the formula for determining the network citation score of the subjective question answers of each student's assignment is as follows: Set the corresponding citation impact factors for each source of the network similarity part, match them with the sources of each network similarity part in the subjective question answers of each student's assignment, obtain the citation impact factors of each network similarity part in the subjective question answers of each student's assignment, and sum them to obtain the comprehensive citation impact factor U of the network similarity part in the subjective question answers of each student's assignment r , and then determine the network citation score of the subjective question answers of each student's assignment In the formula, δ r is the irrationality index of the similar part of the subjective question answer of the r-th student's assignment, represents rounding down, and e is the natural constant

[0050] By comparing the corresponding answer writing outlines of the subjective question answers of each student's assignment with the standard answers on the network public platform, and at the same time comparing the writing outlines of the reference answers, the present invention identifies the irrationality index of the similar parts between the subjective question answers of each student's assignment and the standard answers on the network public platform, can objectively evaluate the citation situation of students, avoid the errors caused by subjective judgment, make the evaluation process more fair, and accordingly determine the network citation score of the students. This evaluation method encourages students to rely more on their own understanding and analysis when completing their assignments, reduces the dependence on direct copying and pasting, thereby promoting the improvement of critical thinking and independent thinking abilities, and effectively detecting whether there is plagiarism or dishonest behavior in students' assignments

[0051] The student interaction scoring module is used to compare the subjective question assignment answers of each student with each other to determine the assignment interaction citation score D″ of each student r .

[0052] In a preferred embodiment, the content of determining the assignment interaction citation score of each student is as follows: Compare the corresponding answer writing outlines of each group of answers in the subjective question answers of each student's assignment with each other, identify the similar text group parts in the corresponding answer writing outlines of the subjective question answers of each student's assignment whose similarity with the assignments of other students is within the specified range, and then, in the same way as the analysis method of the irrationality index of the similar parts of the subjective question answers of each student's assignment, analyze the irrationality indexes of all similar text group parts between each student and the assignments of other students, and screen out the students whose irrationality indexes of all similar text group parts with the assignments of other students exceed the preset irrationality index threshold, and record them as each suspected student

[0053] Extract the submission records of each historical assignment of each suspected student from the marking platform, based on the preset texture features of erased handwriting, identify the total number M′x of erased handwriting in the current submitted assignment of each suspected student, and obtain the total number M of erased handwriting in the submission records of each historical assignment of each suspected student xj , and compare and evaluate the assignment progress index of each suspected student Where M x(j+1) is the total number of handwriting alterations in the submission record of the x-th suspected student's historical assignment for the (j + 1)-th time, ΔM is the preset reference modification difference of the total number of handwriting alterations, x is the number of the suspected student, x = 1, 2,..., c, j is the number of the historical assignment submission record, j = 1, 2,..., q, and q is the number of historical assignment submission records.

[0054] Obtain the simplicity τ of the current assignment pre-submitted by the teacher in the grading platform, and analyze the abnormal evaluation coefficients of the assignments submitted by each suspected student is the preset benchmark abnormal evaluation coefficient of the student's assignment, and record the abnormal evaluation coefficients of the assignments submitted by other students except each suspected student as

[0055] Obtain the abnormal evaluation coefficients of the assignments submitted by each student or Match it with the corresponding assignment cross-reference scoring in the preset abnormal evaluation coefficient ranges to determine the assignment cross-reference scores of each student.

[0056] The result feedback module is used to obtain the contour similarity between the answers to various questions in each student's assignment and the reference answers, and determine the final score of each student's assignment.

[0057] In a preferred embodiment, the determination of the final score of each student's assignment is specifically as follows: After the grading platform accurately calibrates the corresponding answer writing contours of the answers to various questions in each student's assignment, match the corresponding answer writing contours of the answers to various questions in each student's assignment with the corresponding question reference answers pre-submitted by the teacher in the grading platform, and in the same way as the similarity acquisition method of the stroke writing contours, identify the contour similarity μ between the answers to various questions in each student's assignment and the reference answers ri , where i is the number of the assignment question type, i = 1, 2,..., h.

[0058] Determine the final score of each student's assignment Where is the written score of the r-th student's assignment, and h is the number of assignment question types.

[0059] The present invention screens each suspected student whose assignment answers may have cross-references by comparing the assignment answers among students, and then detects the assignment progress indicators by identifying the historical assignment submission records of each suspected student, and accordingly determines the assignment cross-reference scores of each student, increasing the reliability of the final grading score of the student's assignment.

[0060] The above content is only an example and illustration of the concept of the present invention. Those skilled in the art of this technology can make various modifications, supplements, or use similar methods to replace the specific embodiments described, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. An online intelligent marking system based on big data, characterized in that, The system includes: an assignment information extraction module: counting the assignment sets of all students, and classifying the questions in the assignments of each student, including subjective questions and objective questions, and the student number is ; Writing recognition module: Scan and obtain various question answers of each student's homework, extract the writing outlines of their answers, and then recognize the written scores of each student's homework; Recognition accuracy correction module: Screen out the blurred characters and blurred letters in each student's homework that do not meet the standard of normalization, including: if there are zero or more matching characters in the character library for a certain character in the answer writing outline, then mark this character as a blurred character, and the screening method for blurred letters is the same; Perform accuracy calibration on the outlines of the fonts that do not meet the standard of normalization, including: extract each character in the character library that matches each blurred character and its outline, record them as each similar character, and extract the similarity between each blurred character and its each similar character; Recognize each student's inertial writing characteristics and inertial arrangement methods for each stroke from the answer writing outline; Extract each frozen stroke corresponding to each blurred character, recognize its respective writing characteristics, compare it with each student's inertial writing characteristics for each stroke, determine the sub-classified writing tendency strokes of each frozen stroke, and reconstruct them to obtain the sub-classified writing tendency characters of each blurred character in each student's homework; Recognize the arrangement methods to which each frozen stroke belongs, determine the supplementary writing tendency strokes of each frozen stroke, and then reconstruct them to obtain the supplementary writing tendency Chinese characters of each blurred character in each student's homework; Assign similarity values to the sub-classified writing tendency Chinese characters and the supplementary writing tendency Chinese characters, accumulate them to the similarity between each blurred character and its each similar character, obtain the final similarity value, and use the similar character with the maximum final similarity value as the calibrated character for the corresponding blurred character; Network reference scoring module: used to match the subjective question answers of each trainee's assignment with the publicly available network answers, list each network similar part and its source in the subjective question answers of each trainee's assignment, and evaluate the irrationality index of the similar parts of the subjective question answers of each trainee's assignment , and determine the network reference score of the subjective question answers of each trainee's assignment ; The evaluation method is as follows: obtain the reference answers of various types of questions submitted in advance, scan to obtain the corresponding writing outlines, compare them with the network-similar parts in the subjective answers of each student's homework, identify the similarity between the writing outlines of the corresponding answers in the network-similar parts of the subjective answers of each student's homework and the writing outlines of the corresponding parts in the reference answers of the corresponding questions, and accordingly judge whether the network-similar parts in the subjective answers of each student are similar and reasonable or weighted by citation. Statistically calculate the comprehensive citation weighted value of the subjective question answers of each student, obtain the ratio of the value to the preset reference citation weight value, and obtain ; Trainee Interaction Scoring Module: Used to compare the subjective question assignment answers of each trainee with each other to determine the interaction reference score of each trainee's assignment : Result feedback module: Used to obtain the outline similarity between various question answers of each student's homework and the reference answers, and determine the final score of each student's homework.

2. An online intelligent marking system based on big data according to claim 1, characterized in that The written scoring for identifying the assignments of each student is specifically as follows: Match the corresponding answer writing outlines of various questions in the assignments of each student with all the characters and letters in the font library. If a certain character or letter in the corresponding answer writing outline of a certain type of question in a student's assignment has a unique match with the characters or letters in the font library, then record the standardization weight of this character or letter as , if there are multiple matching characters or letters, then record the standardization weight of this character or letter as , if there is no matching character or letter, then record the standardization weight of this character or letter as ; Statistically obtain the comprehensive normalization weights corresponding to all characters and letters of the answer writing outlines of all question answers of each student's homework, obtain the ratio of it to the preset reference normalization weight, and record it as the answer writing normalization index of each student's homework; Further match the answer writing normalization index of each student's homework with the corresponding written scores within the preset range of each writing normalization index to determine the written scores of each student's homework.

3. An online intelligent marking system based on big data according to claim 1, characterized in that, The content of performing accuracy calibration on the outlines of the fonts that do not meet the standard of normalization also includes: extract each letter in the character library that matches each blurred letter in each student's homework that does not meet the standard of normalization and its outline, and record them as each similar letter; Determine the calibrated letters of each blurred letter in each student's homework in the same way as the calibration character recognition method for each blurred character in each student's homework.

4. An online intelligent marking system based on big data according to claim 1, characterized in that, The process of listing each network similar part and its source in the subjective question answers of each student's assignment is as follows: conduct text group division on the corresponding answer writing outlines of the subjective question answers of each student's assignment to obtain the corresponding text group answer writing outlines of the subjective question answers of each student's assignment, count each standard answer of the subjective question on the network public platform, integrate them into the text outline of each standard answer of the subjective question, match it with the corresponding text group answer writing outlines of the subjective question answers of each student's assignment, and obtain each text group outline part with a similarity within the preset range between the subjective question answers of each student's assignment and the text outlines of each standard answer of the corresponding question, which is each network similar part in the subjective question answers of each student's assignment, and mark the text outline of its corresponding standard answer as the source of the corresponding network similar part.

5. An online intelligent marking system based on big data according to claim 1, characterized in that, The network citation score for determining the subjective question answers of each trainee's assignment is calculated by the formula: determine the citation impact factor of each network similarity part in the subjective question answers of each trainee's assignment, and sum them to obtain the comprehensive citation impact factor of the network similarity parts in the subjective question answers of each trainee's assignment , and then determine the network citation score of the subjective question answers of each trainee's assignment , where is the irrationality index of the similarity part of the subjective question answers of the th trainee's assignment, represents rounding down, and e is the natural constant 6. An online intelligent marking system based on big data according to claim 4, characterized in that, The content of determining the interactive citation score of each student's assignment is as follows: compare the corresponding text group answer writing outlines of the subjective question answers of each student's assignment with each other, identify each similar text group part with a similarity within the specified range in the corresponding answer writing outline of the subjective question answers of each student, and then, in the same way as the analysis method of the irrationality index of the similar parts of the subjective question answers of each student's assignment, analyze the irrationality indexes of all similar text group parts of each student and other students' assignments, and screen out the students whose irrationality indexes of all similar text group parts with other students' assignments exceed the preset irrationality index threshold, which are recorded as each suspected student; Extract the submission records of each historical assignment of each suspected student, and identify the total number of handwriting alterations in the current assignment submitted by each suspected student and obtain the total number of handwriting alterations in the submission records of each historical assignment of each suspected student and compare and evaluate the assignment progress indicators of each suspected student with each other wherein is the number of the suspected student, , is the number of the submission records of historical assignments, , is the quantity of the submission records of historical assignments; Obtain the current assignment simplicity pre-submitted by the teacher in the marking platform , analyze the abnormal evaluation coefficients of the assignments submitted by each suspected student , is the preset benchmark abnormal evaluation coefficient for students' assignments, and record the abnormal evaluation coefficients of the assignments submitted by other students except each suspected student as ; Obtain the abnormal evaluation coefficients of the assignments submitted by each student , , match it with the assignment cross-reference scores corresponding to the preset abnormal evaluation coefficient ranges to determine the assignment cross-reference scores of each student.

7. An online intelligent marking system based on big data according to claim 1, characterized in that, The determination of the final scores of the assignments of each student is specifically as follows: After the grading platform accurately calibrates the writing outlines of the answers to various questions in the assignments of each student, it independently identifies the contour similarity between the answers to various questions in the assignments of each student and the reference answers. , is the number of the type of assignment question, ; Determine the final scores of each trainee's assignments , where is the written score of the th trainee's assignment, and is the number of types of assignment topics.

Images