Intelligent higher vocational English dictation system convenient for real-time statistics

Abstract

The invention relates to the technical field of higher vocational English dictation statistics, and relates to an intelligent higher vocational English dictation system convenient for real-time statistics, which comprises a dictation test paper acquisition module, a dictation test paper screening module, an error test paper classification module, a dictation report evaluation module and a cloud database. The dictation test papers corresponding to the students are further screened, the students corresponding to the wrong test papers are marked as target students, error reason analysis is carried out on the wrong test papers corresponding to the target students, error reason classification is carried out on the target students, and finally dictation evaluation reports corresponding to the students are generated. The method effectively guarantees the accuracy and efficiency of summarizing and classifying English dictation errors, effectively guarantees the scientificity and reliability of test paper text analysis through intelligent analysis of dictated test paper texts, and provides accurate data for subsequent error reason classification and analysis.

Description

Technical Field

[0001] The invention belongs to the technical field of higher vocational English dictation statistics, and relates to an intelligent higher vocational English dictation system which is convenient for real-time statistics. Technical Background

[0002] Listening is the first of the five skills of English learning: listening, speaking, reading, writing and translating. Dictation is a must for every student on the road to learning English and is the most difficult skill in English learning. Therefore, listening has long been a weak point and bottleneck in students' English learning.

[0003] However, the dictation process involves the application of multiple abilities and knowledge, such as language perception and recognition ability, vocabulary, short-term memory ability, etc. Dictation can effectively improve students' comprehension of English listening and improve students' pronunciation level. People spend 45% of their time in language communication on listening. College students should not only train their reading and writing skills, but also pay attention to the cultivation of dictation ability. It can be seen that to improve listening level, dictation ability cannot be ignored.

[0004] At present, there are still some areas that need to be optimized in the summary processing of English dictation errors, which are specifically reflected in the following aspects:

[0005] 1. Currently, it is impossible to effectively guarantee the accuracy and efficiency of summarizing and classifying English dictation errors. At the same time, the current technology does not perform intelligent analysis of the dictation test paper text, and cannot effectively guarantee the scientificity and reliability of the test paper text analysis, and it is also impossible to provide accurate data for the subsequent classification analysis of the causes of errors.

[0006] 2. The current summary and classification of English dictation errors is a general assessment. The assessment dimension is too single and the reference value is not strong. It cannot provide a reliable reference for students to correct their learning methods, and thus cannot provide a reliable direction for students' subsequent development. It cannot improve the effectiveness of subsequent classroom teaching, and the level of intelligence and automation is not high. Summary of the invention

[0007] In view of the above problems existing in the prior art, the present invention provides an intelligent higher vocational English dictation system that is convenient for real-time statistics, which is used to solve the above technical problems.

[0008] In order to achieve the above purpose and other purposes, the technical solution adopted by the present invention is as follows:

[0009] The present invention provides an intelligent higher vocational English dictation system that is convenient for real-time statistics. The system comprises a dictation test paper acquisition module, a dictation test paper screening module, an error test paper classification module, a dictation report evaluation module and a cloud database;

[0010] The dictation test paper acquisition module is used for each student to log in to the corresponding dictation client device, connect the dictation client device corresponding to each student to the server, and obtain the student dictation test paper corresponding to each dictation client device through the server, mark it as the dictation test paper corresponding to each student, and synchronously obtain the student number corresponding to each student;

[0011] The dictation test paper screening module is used to screen the dictation test papers corresponding to each student, and further screen them into full score test papers and error test papers;

[0012] The error test paper classification module is used to obtain the student number corresponding to each error test paper, mark the student corresponding to each error test paper as a target student, and then analyze the error cause of the error test paper corresponding to each target student, thereby classifying the error cause of each target student;

[0013] The dictation report evaluation module is used to generate a dictation evaluation report corresponding to each student;

[0014] The cloud database is used to store the target dictation text, the standard shape image of each standard letter, the number of standard bends and the standard bend angle corresponding to each bend, and the number of standard strokes, and is also used to store each reference dictation evaluation report template.

[0015] As a preferred solution, the dictation test papers corresponding to each student are screened, and the specific screening process is as follows:

[0016] Obtain the dictation test paper image corresponding to each student according to the server, and extract the test paper text image corresponding to each student;

[0017] Normalize the text size of the test paper text image corresponding to each student to obtain the test paper text processing image corresponding to each student, perform text separation on the image to obtain the shape and number of strokes of each letter written in the test paper text corresponding to each student, and obtain the number of bends of each letter written in the test paper text corresponding to each student and the corresponding bend angle of each bend from the shape of each letter in the test paper text corresponding to each student;

[0018] Extract the standard shape image of each standard letter from the cloud database, compare it with the shape of each written letter of each student's corresponding test paper text, obtain the comparison overlap between each written letter of each student's corresponding test paper text and each standard letter, and arrange them in descending order, and then select the top five standard letters in the comparison overlap as the reference letters of each written letter of each student's corresponding test paper text, thereby obtaining each reference letter of each written letter of each student's corresponding test paper text;

[0019] According to the number of standard bends of each standard letter, the standard bend angle corresponding to each bend, and the number of standard strokes stored in the cloud database, the number of standard bends of each reference letter, the standard bend angle corresponding to each bend, and the number of standard strokes are obtained, and the letter similarity coefficient of each student's corresponding test paper text corresponding to each reference letter is obtained through analysis;

[0020] The letter similarity coefficients of the reference letters corresponding to the written letters of the corresponding test paper texts of each student are arranged in descending order, and the reference letters with the highest letter similarity coefficient are used as the standard letter styles of the written letters of the corresponding test paper texts of each student, and recorded as the actual written letters of the corresponding test paper texts of each student, thereby obtaining the actual text content of the corresponding test paper texts of each student;

[0021] The target dictation text is extracted from the cloud database, and the words in the target dictation text are further obtained and numbered one by one. At the same time, the actual written letters of the corresponding test paper text of each student are summarized into actual written words, and they are numbered one by one. If the words corresponding to each order in the target dictation text are consistent with the actual written words of each order in the corresponding test paper text of each student, the corresponding test paper text of the student is recorded as a full-mark test paper, otherwise the corresponding test paper text of the student is recorded as an incorrect test paper. In this screening method, full-mark test papers and incorrect test papers are obtained.

[0022] As a preferred solution, the analysis obtains the letter similarity coefficients of each student's corresponding test paper text and each reference letter. The specific analysis process is as follows:

[0023] The number of standard bends of each reference letter, the standard bend angle corresponding to each bend, and the number of standard strokes are marked as Wherein, c represents the reference letter number, c=1, 2, ... 5, w represents the number of bends, w=1, 2, ... q;

[0024] Mark the number of strokes, number of bends and the corresponding bending angle of each bend for each letter in the test paper text of each student as h represents the student's number, h = 1, 2, ... t, r represents the number of the written letter, r = 1, 2, ... p;

[0025] The allowable bending angle Δθ of each bend is corresponding to each predefined reference letter. cw , by analyzing the formula , the analysis results show that the letter similarity coefficients of each student's corresponding test paper text and each reference letter e represents a natural constant, q represents the total number of bends, a1, a2 and a3 represent the weights of the letter similarity coefficients corresponding to the set number of bends, bending angles and number of strokes respectively, and a1+a2+a3=1.

[0027] As a preferred solution, the wrong test paper classification module includes a vocabulary analysis subunit and a speech analysis subunit.

[0028] As a preferred solution, the vocabulary analysis subunit analyzes the causes of errors in the wrong test papers corresponding to each target student. The specific analysis process is as follows:

[0029] The words corresponding to each order in the target dictation text are compared with the actual written words of each order in the corresponding test paper text of each target student. If the words corresponding to a certain order in the target dictation text are inconsistent with the actual written words of a certain order in the corresponding test paper text of a target student, the actual written words corresponding to the order are recorded as incorrect words, thereby obtaining the incorrect words of the corresponding test paper text of each target student;

[0030] Get the number of letters K of each wrong word in the test paper text of each target student hd , d represents the number of the wrong word, d = 1, 2...n, and according to the order of the wrong words in the corresponding test paper text of each target student, the reference letter number K′ of the words in the corresponding order in the target dictation text is synchronously obtained, and each letter in each wrong word in the corresponding test paper text of each target student is compared with each letter in the words in the corresponding order in the target dictation text, and the number of overlapping letters between each wrong word in the corresponding test paper text of each target student and the words in the corresponding order in the target dictation text is obtained;

[0031] pass Calculate the possible evaluation coefficient of vocabulary error for each wrong word in the test paper text of each target student φ hd , b1 and b2 represent the influence factors of the predefined number of letters and the number of overlapping letters, respectively, and b1+b2=1, ΔK represents the predefined allowable difference in the number of letters, K′ hd represents the number of overlapping letters between the dth incorrect word in the test paper text of the hth target student and the corresponding order word in the target dictation text, and cot represents the cotangent function;

[0032] The possible vocabulary error assessment coefficient of each incorrect word in the corresponding test paper text of each target student is compared with the preset vocabulary error reference assessment coefficient. If the possible vocabulary error assessment coefficient of a certain incorrect word in the corresponding test paper text of a target student is greater than or equal to the preset vocabulary error reference assessment coefficient, the error cause of the incorrect word in the corresponding test paper text of the target student is marked as a vocabulary reason.

[0033] As a preferred solution, the speech analysis subunit performs error cause analysis on the error test papers corresponding to each target student, and the specific analysis process is as follows:

[0034] Step 1: According to the incorrect words in the test paper text of each target student, import them into the intelligent speech library, obtain the speech samples of the incorrect words in the test paper text of each target student, and divide them into frames to further obtain the original speech signal value of each frame speech signal corresponding to each incorrect word in the test paper text of each target student And the frequency index value g represents the number of each frame of speech signal, g = 1, 2...b;

[0035] It should be noted that the frequency index value of each frame of speech signal corresponding to each wrong word in the test paper text of each target student is specifically obtained as follows:

[0036] The sampling frequency of each frame of speech signal corresponding to each incorrect word in the test paper text of each target student is obtained, and the reference length of FFT is extracted from the cloud database. The sampling frequency of each frame of speech signal corresponding to each incorrect word in the test paper text of each target student is used as the numerator of the fraction, and the reference length of FFT is used as the denominator of the fraction, and then the frequency resolution of each frame of speech signal corresponding to each incorrect word in the test paper text of each target student is calculated;

[0037] According to the order corresponding to each incorrect word in the test paper text corresponding to each target student, the standard frequency value corresponding to the order of each incorrect word in the test paper text corresponding to each target student is obtained, and the standard frequency value is multiplied by the sampling frequency of each frame of speech signal corresponding to each incorrect word in the test paper text corresponding to each target student, and the calculated result is divided by the reference length of FFT, thereby calculating the frequency index value of each frame of speech signal corresponding to each incorrect word in the test paper text corresponding to each target student.

[0038] Step 2: Use the Fourier transform formula to calculate the frequency domain signal value of each frame of speech signal corresponding to the speech sample of each wrong word in the test paper text of each target student. i(j) represents the preset window function, and FFT represents the fast Fourier transform calculation formula;

[0039] Step 3: Import the frequency index value of each frame of speech signal corresponding to each erroneous word in the test paper text of each target student into the Mel filter group, and number the Mel filters in the Mel filter group according to the index order, and obtain the center frequency value f(m) of each Mel filter, where m represents the index number of each Mel filter, m=1,2,...z, and further calculate the output value of each Mel filter corresponding to each frame of speech signal corresponding to each erroneous word speech sample in the test paper text of each target student

[0040] Step 4: Calculate the total energy value of each Mel filter corresponding to each wrong word speech sample in the test paper text of each target student

[0041] Step 5: Calculate the cepstrum coefficients of the speech samples of each wrong word in the test paper text of each target student. z represents the total number of Mel filters in the Mel filter bank;

[0042] The cepstral coefficients of the speech samples of the wrong words in the test paper text of each target student are combined to obtain the feature vector Y′ of the speech samples of the wrong words in the test paper text of each target student, expressed as Y′=[Ε h1 ,Ε h2 ...Ε hd ,...Ε hn ];

[0043] Step 6: Calculate the feature vector Y″ of the speech sample of the order of the wrong words in the target dictation text in the same way as the feature vector of the speech sample of the wrong word in the test paper text of each target student;

[0044] Step 7: Based on the analysis formula The cosine similarity ψ between the speech samples of each wrong word in the test paper text and the speech samples of the corresponding wrong word order in the target dictation text is obtained by analysis. hd , where ||Y′|| represents the Euclidean norm of the feature vector of the speech sample of each incorrect word in the test paper text of each target student, and ||Y″|| represents the Euclidean norm of the feature vector of the speech sample of the incorrect word order in the target dictation text;

[0045] Step 8: Compare the cosine similarity of the speech samples of each incorrect word in the test paper text of each target student and the speech samples of the corresponding incorrect word order in the target dictation text with the predefined reference cosine similarity. If the cosine similarity of the speech samples of each incorrect word in the test paper text of a target student and the speech samples of the corresponding incorrect word order in the target dictation text is greater than or equal to the predefined reference cosine similarity, then mark the error cause of the incorrect word in the test paper text of the target student as a speech error.

[0046] As a preferred solution, the calculation obtains each Mel filter output value corresponding to each frame of speech signal corresponding to each wrong word speech sample in the test paper text of each target student, and the calculation formula is as follows:

[0047] pass Calculate the output value of each Mel filter for each wrong word speech sample in the test paper text corresponding to each target student, each frame speech signal f(m-1), f(m+1), and f(m+2) represent the center frequency values ​​of the m-1th, m+1th, and m+2th Mel filters, respectively.

[0048] As a preferred solution, the error causes of each target student are classified, and the specific classification process is as follows:

[0049] According to the wrong words in the test paper text of each target student, the total number of wrong words in the test paper text of each target student is counted

[0050] Get the number of wrong words for each target student's test paper text whose error reasons are vocabulary reasons and phonetic errors, respectively, and mark them as and

[0051] pass The proportion of incorrect words due to vocabulary errors in the corresponding test paper text of each target student is calculated as ε′ h ,pass The proportion of incorrect words due to phonetic errors in the corresponding test paper text of each target student is calculated as ε″ h ;

[0052] If the proportion of the number of incorrect words due to vocabulary reasons in the corresponding test paper text of a target student is greater than or equal to the predetermined standard proportion of the number of incorrect words, the cause of the error of the target student is classified as vocabulary; if the proportion of the number of incorrect words due to phonetic reasons in the corresponding test paper text of a target student is greater than or equal to the predetermined standard proportion of the number of incorrect words, the cause of the error of the target student is classified as phonetic; if the proportion of the number of incorrect words due to phonetic reasons and the proportion of incorrect words due to vocabulary reasons in the corresponding test paper text of a target student are both within the predefined reference proportion range, the cause of the error of the target student is classified as vocabulary and phonetic.

[0053] As a preferred solution, the dictation evaluation report corresponding to each student is generated, and the specific generation process is as follows:

[0054] If the corresponding test paper of a student is identified as a full-mark test paper, the reference dictation report of the full-mark test paper is selected from the reference dictation evaluation report templates stored in the cloud database and recorded as the dictation evaluation report of the student;

[0055] If the cause of a target student's error is identified as phonetic, the reference dictation evaluation report templates stored in the cloud database are selected as the dictation evaluation report for the target student that needs to strengthen phonetic training;

[0056] If the cause of a target student's error is identified as vocabulary, the reference dictation evaluation report templates stored in the cloud database are selected as the dictation evaluation report for the target student that needs to strengthen vocabulary training;

[0057] If the cause of a target student's error is identified as vocabulary and phonetics, based on the reference dictation evaluation report templates stored in the cloud database, the reference dictation reports that require enhanced vocabulary and phonetics training are selected as the dictation evaluation reports for the target student.

[0058] As described above, the intelligent higher vocational English dictation system that facilitates real-time statistics provided by the present invention has at least the following beneficial effects:

[0059] (1) The present invention provides an intelligent higher vocational English dictation system that is convenient for real-time statistics. By obtaining the dictation test papers of each student, the dictation test papers corresponding to each student are further screened, and the students corresponding to each wrong test paper are marked as target students. Then, the error causes of the wrong test papers corresponding to each target student are analyzed, and the error causes of each target student are classified. Finally, a dictation evaluation report corresponding to each student is generated, which effectively solves the problem that there are still certain limitations in the current classification of student dictation test paper errors, effectively guarantees the accuracy and efficiency of English dictation error summary and classification, and at the same time, by intelligently analyzing the dictation test paper text, effectively guarantees the scientificity and reliability of the test paper text analysis, and also provides accurate data for subsequent error cause classification analysis.

[0060] (2) The embodiment of the present invention avoids the general evaluation of summarizing and classifying English dictation errors, improves the evaluation dimension, enhances the reference value of the analysis results, and provides a reliable reference basis for students to correct their learning methods, thereby providing a reliable direction for their subsequent development, improving the subsequent classroom teaching effect, and improving the intelligence and automation level of summarizing and classifying English dictation errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0062] Figure 1 It is a schematic diagram of the connection of various modules of the system of the present invention. DETAILED DESCRIPTION

[0063] The above contents in combination with the implementation of the present invention are merely examples and explanations of the concept of the present invention. The technical personnel in the relevant technical field may make various modifications or supplements to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they shall all fall within the protection scope of the present invention.

[0064] See also Figure 1 As shown, an intelligent higher vocational English dictation system that is convenient for real-time statistics, the system includes a dictation test paper acquisition module, a dictation test paper screening module, an error test paper classification module, a dictation report evaluation module and a cloud database;

[0065] The dictation test paper screening module is connected to the dictation test paper acquisition module and the error test paper classification module respectively, the dictation report evaluation module is connected to the error test paper classification module, and the cloud database is connected to the dictation test paper screening module, the error test paper classification module and the dictation report evaluation module respectively.

[0066] The dictation test paper acquisition module is used for each student to log in to the corresponding dictation client device, connect the dictation client device corresponding to each student to the server, and obtain the student dictation test paper corresponding to each dictation client device through the server, mark it as the dictation test paper corresponding to each student, and synchronously obtain the student number corresponding to each student;

[0067] The dictation test paper screening module is used to screen the dictation test papers corresponding to each student, and further screen them into full score test papers and error test papers;

[0068] Preferably, based on the above scheme, the dictation test papers corresponding to each student are screened, and the specific screening process is as follows:

[0069] Obtain the dictation test paper image corresponding to each student according to the server, and extract the test paper text image corresponding to each student;

[0070] It should be noted that the specific extraction process of extracting the test paper text content corresponding to each student is as follows:

[0071] First, the test paper image is preprocessed to improve the accuracy of OCR. This includes denoising, adjusting the brightness and contrast of the image, and performing image binarization to convert the text part to black and the background part to white;

[0072] Use the OCR library or algorithm to perform text detection on the preprocessed image, and then determine the text area in the image. Use OCR technology to perform text recognition on each detected text area, thereby extracting the test paper text content corresponding to each student.

[0073] Normalize the text size of the test paper text image corresponding to each student to obtain the test paper text processing image corresponding to each student, perform text separation on the image to obtain the shape and number of strokes of each letter written in the test paper text corresponding to each student, and obtain the number of bends of each letter written in the test paper text corresponding to each student and the corresponding bend angle of each bend from the shape of each letter in the test paper text corresponding to each student;

[0074] Extract the standard shape image of each standard letter from the cloud database, compare it with the shape of each written letter of each student's corresponding test paper text, obtain the comparison overlap between each written letter of each student's corresponding test paper text and each standard letter, and arrange them in descending order, and then select the top five standard letters in the comparison overlap as the reference letters of each written letter of each student's corresponding test paper text, thereby obtaining each reference letter of each written letter of each student's corresponding test paper text;

[0075] It should be noted that the specific process of obtaining the degree of coincidence between each written letter and each standard letter of the corresponding test paper text of each student is as follows:

[0076] The standard shape image of each standard letter is compared with the shape of each written letter of the corresponding test paper text of each student, and the overlapping area of ​​each written letter of the corresponding test paper text of each student and each standard letter is obtained;

[0077] According to the shape of each written letter in the corresponding test paper text of each student, the shape area of ​​each written letter in the corresponding test paper text of each student is used as the denominator of the fraction, and the overlapping area of ​​each written letter in the corresponding test paper text of each student and each standard letter is used as the numerator of the fraction. The comparison and overlap degree of each written letter in the corresponding test paper text of each student and each standard letter is further calculated through the fraction.

[0078] According to the number of standard bends of each standard letter, the standard bend angle corresponding to each bend, and the number of standard strokes stored in the cloud database, the number of standard bends of each reference letter, the standard bend angle corresponding to each bend, and the number of standard strokes are obtained, and the letter similarity coefficient of each student's corresponding test paper text corresponding to each reference letter is obtained through analysis;

[0079] The letter similarity coefficients of the reference letters corresponding to the written letters of the corresponding test paper texts of each student are arranged in descending order, and the reference letters with the highest letter similarity coefficient are used as the standard letter styles of the written letters of the corresponding test paper texts of each student, and recorded as the actual written letters of the corresponding test paper texts of each student, thereby obtaining the actual text content of the corresponding test paper texts of each student;

[0080] The target dictation text is extracted from the cloud database, and the words in the target dictation text are further obtained and numbered one by one. At the same time, the actual written letters of the corresponding test paper text of each student are summarized into actual written words, and they are numbered one by one. If the words corresponding to each order in the target dictation text are consistent with the actual written words of each order in the corresponding test paper text of each student, the corresponding test paper text of the student is recorded as a full-mark test paper, otherwise the corresponding test paper text of the student is recorded as an incorrect test paper. In this screening method, full-mark test papers and incorrect test papers are obtained.

[0081] Based on the above scheme, the analysis is preferably performed to obtain the letter similarity coefficients of each student's corresponding test paper text and each reference letter. The specific analysis process is as follows:

[0082] The number of standard bends of each reference letter, the standard bend angle corresponding to each bend, and the number of standard strokes are marked as Wherein, c represents the reference letter number, c=1, 2, ... 5, w represents the number of bends, w=1, 2, ... q;

[0083] Mark the number of strokes, number of bends and the corresponding bending angle of each bend for each letter in the test paper text of each student as h represents the student's number, h = 1, 2, ... t, r represents the number of the written letter, r = 1, 2, ... p;

[0084] The allowable bending angle Δθ of each bend is corresponding to each predefined reference letter. cw , by analyzing the formula , the analysis results show that the letter similarity coefficients of each student's corresponding test paper text and each reference letter e represents a natural constant, q represents the total number of bends, a1, a2 and a3 represent the weights of the letter similarity coefficients corresponding to the set number of bends, bending angles and number of strokes respectively, and a1+a2+a3=1.

[0086] The error test paper classification module is used to obtain the student number corresponding to each error test paper, mark the student corresponding to each error test paper as a target student, and then analyze the error cause of the error test paper corresponding to each target student, thereby classifying the error cause of each target student;

[0087] Preferably, based on the above scheme, the wrong test paper classification module includes a vocabulary analysis subunit and a speech analysis subunit.

[0088] Preferably, based on the above scheme, the vocabulary analysis subunit analyzes the causes of errors in the wrong test papers corresponding to each target student, and the specific analysis process is as follows:

[0089] The words corresponding to each order in the target dictation text are compared with the actual written words of each order in the corresponding test paper text of each target student. If the words corresponding to a certain order in the target dictation text are inconsistent with the actual written words of a certain order in the corresponding test paper text of a target student, the actual written words corresponding to the order are recorded as incorrect words, thereby obtaining the incorrect words of the corresponding test paper text of each target student;

[0090] Get the number of letters K of each wrong word in the test paper text of each target student hd , d represents the number of the wrong word, d = 1, 2...n, and according to the order of the wrong words in the corresponding test paper text of each target student, the reference letter number K′ of the words in the corresponding order in the target dictation text is synchronously obtained, and each letter in each wrong word in the corresponding test paper text of each target student is compared with each letter in the words in the corresponding order in the target dictation text, and the number of overlapping letters between each wrong word in the corresponding test paper text of each target student and the words in the corresponding order in the target dictation text is obtained;

[0091] pass Calculate the possible evaluation coefficient of vocabulary error for each wrong word in the test paper text of each target student φ hd , b1 and b2 represent the influence factors of the predefined number of letters and the number of overlapping letters, respectively, and b1+b2=1, ΔK represents the predefined allowable difference in the number of letters, K′ hd represents the number of overlapping letters between the dth incorrect word in the test paper text of the hth target student and the corresponding order word in the target dictation text, and cot represents the cotangent function;

[0092] The possible vocabulary error assessment coefficient of each incorrect word in the corresponding test paper text of each target student is compared with the preset vocabulary error reference assessment coefficient. If the possible vocabulary error assessment coefficient of a certain incorrect word in the corresponding test paper text of a target student is greater than or equal to the preset vocabulary error reference assessment coefficient, the error cause of the incorrect word in the corresponding test paper text of the target student is marked as a vocabulary reason.

[0093] Preferably, based on the above scheme, the speech analysis subunit analyzes the causes of errors in the wrong test papers corresponding to each target student, and the specific analysis process is as follows:

[0094] Step 1: According to the incorrect words in the test paper text of each target student, import them into the intelligent speech library, obtain the speech samples of the incorrect words in the test paper text of each target student, and divide them into frames to further obtain the original speech signal value of each frame speech signal corresponding to each incorrect word in the test paper text of each target student And the frequency index value g represents the number of each frame of speech signal, g = 1, 2...b;

[0095] Step 2: Use the Fourier transform formula to calculate the frequency domain signal value of each frame of speech signal corresponding to the speech sample of each wrong word in the test paper text of each target student. i(j) represents the preset window function, and FFT represents the fast Fourier transform calculation formula;

[0096] Step 3: Import the frequency index value of each frame of speech signal corresponding to each erroneous word in the test paper text of each target student into the Mel filter group, and number the Mel filters in the Mel filter group according to the index order, and obtain the center frequency value f(m) of each Mel filter, where m represents the index number of each Mel filter, m=1,2,...z, and further calculate the output value of each Mel filter corresponding to each frame of speech signal corresponding to each erroneous word speech sample in the test paper text of each target student

[0097] Step 4: Calculate the total energy value of each Mel filter corresponding to each wrong word speech sample in the test paper text of each target student

[0098] Step 5: Calculate the cepstrum coefficients of the speech samples of each wrong word in the test paper text of each target student. z represents the total number of Mel filters in the Mel filter bank;

[0099] The cepstral coefficients of the speech samples of the wrong words in the test paper text of each target student are combined to obtain the feature vector Y′ of the speech samples of the wrong words in the test paper text of each target student, expressed as Y′=[Ε h1 ,Ε h2 ...Ε hd ,...Ε hn ];

[0100] Step 6: Calculate the feature vector Y″ of the speech sample of the order of the wrong words in the target dictation text in the same way as the feature vector of the speech sample of the wrong word in the test paper text of each target student;

[0101] Step 7: Based on the analysis formula The cosine similarity ψ between the speech samples of each wrong word in the test paper text and the speech samples of the corresponding wrong word order in the target dictation text is obtained by analysis. hd , where ||Y′|| represents the Euclidean norm of the feature vector of the speech sample of each incorrect word in the test paper text of each target student, and ||Y″|| represents the Euclidean norm of the feature vector of the speech sample of the incorrect word order in the target dictation text;

[0102] Step 8: Compare the cosine similarity of the speech samples of each incorrect word in the test paper text of each target student and the speech samples of the corresponding incorrect word order in the target dictation text with the predefined reference cosine similarity. If the cosine similarity of the speech samples of each incorrect word in the test paper text of a target student and the speech samples of the corresponding incorrect word order in the target dictation text is greater than or equal to the predefined reference cosine similarity, then mark the error cause of the incorrect word in the test paper text of the target student as a speech error.

[0103] Based on the above scheme, the calculation obtains the output value of each Mel filter corresponding to each frame of speech signal corresponding to each wrong word speech sample in the test paper text of each target student, and the calculation formula is as follows:

[0104] pass Calculate the output value of each Mel filter for each wrong word speech sample in the test paper text corresponding to each target student, each frame speech signal f(m-1), f(m+1), and f(m+2) represent the center frequency values ​​of the m-1th, m+1th, and m+2th Mel filters, respectively.

[0105] Based on the above scheme, the error causes of each target student are classified, and the specific classification process is as follows:

[0106] According to the wrong words in the test paper text of each target student, the total number of wrong words in the test paper text of each target student is counted

[0107] Get the number of wrong words for each target student's test paper text whose error reasons are vocabulary reasons and phonetic errors, respectively, and mark them as and

[0108] pass The proportion of incorrect words due to vocabulary errors in the corresponding test paper text of each target student is calculated as ε′ h ,pass The proportion of incorrect words due to phonetic errors in the corresponding test paper text of each target student is calculated as ε″ h ;

[0109] If the proportion of the number of incorrect words due to vocabulary reasons in the corresponding test paper text of a target student is greater than or equal to the predetermined standard proportion of the number of incorrect words, the cause of the error of the target student is classified as vocabulary; if the proportion of the number of incorrect words due to phonetic reasons in the corresponding test paper text of a target student is greater than or equal to the predetermined standard proportion of the number of incorrect words, the cause of the error of the target student is classified as phonetic; if the proportion of the number of incorrect words due to phonetic reasons and the proportion of incorrect words due to vocabulary reasons in the corresponding test paper text of a target student are both within the predefined reference proportion range, the cause of the error of the target student is classified as vocabulary and phonetic.

[0110] The embodiment of the present invention avoids the general evaluation of summarizing and classifying English dictation errors, improves the evaluation dimension, enhances the reference of the analysis results, and provides a reliable reference basis for students to correct their learning methods, thereby providing a reliable direction for their subsequent development, and also improves the subsequent classroom teaching effect, and improves the intelligence and automation level of summarizing and classifying English dictation errors.

[0111] The dictation report evaluation module is used to generate a dictation evaluation report corresponding to each student;

[0112] Preferably, based on the above scheme, the dictation evaluation report corresponding to each student is generated, and the specific generation process is as follows:

[0113] If the corresponding test paper of a student is identified as a full-mark test paper, the reference dictation report of the full-mark test paper is selected from the reference dictation evaluation report templates stored in the cloud database and recorded as the dictation evaluation report of the student;

[0114] It should be noted that the specific screening process for selecting reference dictation reports with full-score test papers is as follows:

[0115] According to each reference dictation evaluation report template, text extraction is performed. If a reference dictation evaluation report template contains multiple words such as full marks and excellent, the reference dictation evaluation report template will be used as the dictation evaluation report of the student.

[0116] If the cause of a target student's error is identified as phonetic, the reference dictation evaluation report templates stored in the cloud database are selected as the dictation evaluation report for the target student that needs to strengthen phonetic training;

[0117] According to each reference dictation evaluation report template, text extraction is performed. If a reference dictation evaluation report template contains multiple words such as voice training, grammar, semantics, etc., the reference dictation evaluation report template will be used as the dictation evaluation report of the student.

[0118] If the cause of a target student's error is identified as vocabulary, the reference dictation evaluation report templates stored in the cloud database are selected as the dictation evaluation report for the target student that needs to strengthen vocabulary training;

[0119] According to each reference dictation evaluation report template, extract the text. If there are multiple words such as strengthening vocabulary, reading, letters, etc. in a reference dictation evaluation report template, use the reference dictation evaluation report template as the student's dictation evaluation report.

[0120] If the cause of a target student's error is identified as vocabulary and phonetics, based on the reference dictation evaluation report templates stored in the cloud database, the reference dictation reports that require enhanced vocabulary and phonetics training are selected as the dictation evaluation reports for the target student.

[0121] The cloud database is used to store the target dictation text, the standard shape image of each standard letter, the number of standard bends and the standard bend angle corresponding to each bend, and the number of standard strokes, and is also used to store each reference dictation evaluation report template.

[0122] The present invention provides an intelligent higher vocational English dictation system that is convenient for real-time statistics. The system obtains the dictation test papers of each student, further screens the dictation test papers corresponding to each student, marks the students corresponding to each wrong test paper as target students, and then analyzes the causes of errors in the wrong test papers corresponding to each target student, classifies the causes of errors for each target student, and finally generates a dictation evaluation report corresponding to each student. The system effectively solves the problem that there are still certain limitations in the current classification of errors in student dictation test papers, effectively guarantees the accuracy and efficiency of summarizing and classifying English dictation errors, and at the same time, through the intelligent analysis of the dictation test paper text, effectively guarantees the scientificity and reliability of the test paper text analysis, and also provides accurate data for subsequent error cause classification analysis.

[0123] The above contents are merely examples and explanations of the concept of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. An intelligent higher vocational English dictation system that facilitates real-time statistics, characterized in that: The system includes: The dictation test paper acquisition module is used for each student to log in to the corresponding dictation client device, connect the dictation client device corresponding to each student to the server, obtain the student dictation test paper corresponding to each dictation client device through the server, mark it as the dictation test paper corresponding to each student, and synchronously obtain the student number corresponding to each student; The dictation test paper screening module is used to screen the dictation test papers corresponding to each student, and then screen them into full-score test papers and error test papers; The error test paper classification module is used to obtain the student number corresponding to each error test paper, mark the student corresponding to each error test paper as a target student, and then analyze the error cause of the error test paper corresponding to each target student, thereby classifying the error cause of each target student; The dictation report evaluation module is used to generate a dictation evaluation report corresponding to each student; The cloud database is used to store the target dictation text, the standard shape image of each standard letter, the number of standard bends and the standard bending angle corresponding to each bend, and the number of standard strokes, and is also used to store each reference dictation evaluation report template.

2. According to claim 1, an intelligent higher vocational English dictation system that facilitates real-time statistics is characterized in that: The dictation test papers corresponding to each student are screened, and the specific screening process is as follows: Obtain the dictation test paper image corresponding to each student according to the server, and extract the test paper text image corresponding to each student; Normalize the text size of the test paper text image corresponding to each student to obtain the test paper text processing image corresponding to each student, perform text separation on the image to obtain the shape and number of strokes of each letter written in the test paper text corresponding to each student, and obtain the number of bends of each letter written in the test paper text corresponding to each student and the corresponding bend angle of each bend from the shape of each letter in the test paper text corresponding to each student; Extract the standard shape image of each standard letter from the cloud database, compare it with the shape of each written letter of each student's corresponding test paper text, obtain the comparison overlap between each written letter of each student's corresponding test paper text and each standard letter, and arrange them in descending order, and then select the top five standard letters in the comparison overlap as the reference letters of each written letter of each student's corresponding test paper text, thereby obtaining each reference letter of each written letter of each student's corresponding test paper text; According to the number of standard bends of each standard letter, the standard bend angle corresponding to each bend, and the number of standard strokes stored in the cloud database, the number of standard bends of each reference letter, the standard bend angle corresponding to each bend, and the number of standard strokes are obtained, and the letter similarity coefficient of each student's corresponding test paper text corresponding to each reference letter is obtained through analysis; The letter similarity coefficients of the reference letters corresponding to the written letters of the corresponding test paper texts of each student are arranged in descending order, and the reference letters with the highest letter similarity coefficient are used as the standard letter styles of the written letters of the corresponding test paper texts of each student, and recorded as the actual written letters of the corresponding test paper texts of each student, thereby obtaining the actual text content of the corresponding test paper texts of each student; The target dictation text is extracted from the cloud database, and the words in the target dictation text are further obtained and numbered one by one. At the same time, the actual written letters of the corresponding test paper text of each student are summarized into actual written words, and they are numbered one by one. If the words corresponding to each order in the target dictation text are consistent with the actual written words of each order in the corresponding test paper text of each student, the corresponding test paper text of the student is recorded as a full-mark test paper, otherwise the corresponding test paper text of the student is recorded as an incorrect test paper. In this screening method, full-mark test papers and incorrect test papers are obtained.

3. The intelligent higher vocational English dictation system that facilitates real-time statistics according to claim 2 is characterized in that: The analysis obtains the letter similarity coefficients of each student's corresponding test paper text and each reference letter. The specific analysis process is as follows: The number of standard bends of each reference letter, the standard bend angle corresponding to each bend, and the number of standard strokes are marked as Wherein, c represents the reference letter number, c=1, 2, ... 5, w represents the number of bends, w=1, 2, ... q; Mark the number of strokes, number of bends and the corresponding bending angle of each bend for each letter in the test paper text of each student as w, h represent the student's number, h = 1, 2, ... t, r represents the number of the written letter, r = 1, 2, ... p; The allowable bending angle Δθ of each bend is corresponding to each predefined reference letter. cw , by analyzing the formula , the analysis results show that the letter similarity coefficients of each student's corresponding test paper text and each reference letter e represents a natural constant, q represents the total number of bends, a1, a2 and a3 represent the weights of the letter similarity coefficients corresponding to the set number of bends, bending angles and number of strokes respectively, and a1+a2+a3=1.

4. The intelligent higher vocational English dictation system that facilitates real-time statistics according to claim 1 is characterized in that: The wrong test paper classification module includes a vocabulary analysis subunit and a speech analysis subunit.

5. The intelligent higher vocational English dictation system that facilitates real-time statistics according to claim 4 is characterized in that: The vocabulary analysis subunit analyzes the causes of errors in the wrong test papers corresponding to each target student. The specific analysis process is as follows: The words corresponding to each order in the target dictation text are compared with the actual written words of each order in the corresponding test paper text of each target student. If the words corresponding to a certain order in the target dictation text are inconsistent with the actual written words of a certain order in the corresponding test paper text of a target student, the actual written words corresponding to the order are recorded as incorrect words, thereby obtaining the incorrect words of the corresponding test paper text of each target student; Get the number of letters K of each wrong word in the test paper text of each target student hd , d represents the number of the wrong word, d = 1, 2...n, and according to the order of each wrong word in the corresponding test paper text of each target student, the reference number K′ of the order of the wrong words in the target dictation text is synchronously obtained, and each letter in each wrong word in the corresponding test paper text of each target student is compared with each letter in the word of the corresponding order in the target dictation text, and the number of overlapping letters between each wrong word in the corresponding test paper text of each target student and the word of the corresponding order in the target dictation text is obtained; pass Calculate the possible evaluation coefficient of vocabulary error for each wrong word in the test paper text of each target student φ hd , b1 and b2 represent the influence factors of the predefined number of letters and the number of overlapping letters, respectively, and b1+b2=1, ΔK represents the predefined allowable difference in the number of letters, K′ hd represents the number of overlapping letters between the dth incorrect word in the test paper text of the hth target student and the corresponding order word in the target dictation text, and cot represents the cotangent function; The possible vocabulary error assessment coefficient of each incorrect word in the corresponding test paper text of each target student is compared with the preset vocabulary error reference assessment coefficient. If the possible vocabulary error assessment coefficient of a certain incorrect word in the corresponding test paper text of a target student is greater than or equal to the preset vocabulary error reference assessment coefficient, the error cause of the incorrect word in the corresponding test paper text of the target student is marked as a vocabulary reason.

6. The intelligent higher vocational English dictation system that facilitates real-time statistics according to claim 5 is characterized in that: The speech analysis subunit analyzes the causes of errors in the wrong test papers corresponding to each target student. The specific analysis process is as follows: Step 1: According to the incorrect words in the test paper text of each target student, import them into the intelligent speech library, obtain the speech samples of the incorrect words in the test paper text of each target student, and divide them into frames to further obtain the original speech signal value of each frame speech signal corresponding to each incorrect word in the test paper text of each target student And the frequency index value g represents the number of each frame of speech signal, g = 1, 2...b; Step 2: Use the Fourier transform formula to calculate the frequency domain signal value of each frame of speech signal corresponding to the speech sample of each wrong word in the test paper text of each target student. i(j) represents the preset window function, and FFT represents the fast Fourier transform calculation formula; Step 3: Import the frequency index value of each frame of speech signal corresponding to each erroneous word in the test paper text of each target student into the Mel filter group, and number the Mel filters in the Mel filter group according to the index order, and obtain the center frequency value f(m) of each Mel filter, where m represents the index number of each Mel filter, m=1,2,...z, and further calculate the output value of each Mel filter corresponding to each frame of speech signal corresponding to each erroneous word speech sample in the test paper text of each target student Step 4: Calculate the total energy value of each Mel filter corresponding to each wrong word speech sample in the test paper text of each target student Step 5: Calculate the cepstrum coefficients of the speech samples of each wrong word in the test paper text of each target student. z represents the total number of Mel filters in the Mel filter bank; The cepstral coefficients of the speech samples of the wrong words in the test paper text of each target student are combined to obtain the feature vector Y′ of the speech samples of the wrong words in the test paper text of each target student, expressed as Y′=[Ε h1 ,Ε h2 ...Ε hd ,...Ε hn ]; Step 6: Calculate the feature vector Y″ of the speech sample of the order of the wrong words in the target dictation text in the same way as the feature vector of the speech sample of the wrong word in the test paper text of each target student; Step 7: Based on the analysis formula The cosine similarity ψ between the speech samples of each wrong word in the test paper text and the speech samples of the corresponding wrong word order in the target dictation text is obtained by analysis. hd , where ||Y′|| represents the Euclidean norm of the feature vector of the speech sample of each incorrect word in the test paper text of each target student, and ||Y″|| represents the Euclidean norm of the feature vector of the speech sample of the incorrect word order in the target dictation text; Step 8: Compare the cosine similarity of the speech samples of each incorrect word in the test paper text of each target student and the speech samples of the corresponding incorrect word order in the target dictation text with the predefined reference cosine similarity. If the cosine similarity of the speech samples of each incorrect word in the test paper text of a target student and the speech samples of the corresponding incorrect word order in the target dictation text is greater than or equal to the predefined reference cosine similarity, then mark the error cause of the incorrect word in the test paper text of the target student as a speech error.

7. The intelligent higher vocational English dictation system that facilitates real-time statistics according to claim 6 is characterized in that: The calculation results in that each target student's corresponding erroneous word voice sample in the test paper text corresponds to each frame of voice signal and each Mel filter output value, and the calculation formula is as follows: pass Calculate the output value of each Mel filter for each wrong word speech sample in the test paper text corresponding to each target student, each frame speech signal f(m-1), f(m+1), and f(m+2) represent the center frequency values ​​of the m-1th, m+1th, and m+2th Mel filters, respectively.

8. The intelligent higher vocational English dictation system that facilitates real-time statistics according to claim 7 is characterized in that: The error causes of each target student are classified, and the specific classification process is as follows: According to the wrong words in the test paper text of each target student, the total number of wrong words in the test paper text of each target student is counted Get the number of wrong words for each target student's test paper text whose error reasons are vocabulary reasons and phonetic errors, respectively, and mark them as and pass The proportion of incorrect words due to vocabulary errors in the corresponding test paper text of each target student is calculated as ε′ h ,pass The proportion of incorrect words due to phonetic errors in the corresponding test paper text of each target student is calculated as ε″ h ; If the proportion of the number of incorrect words due to vocabulary reasons in the corresponding test paper text of a target student is greater than or equal to the predetermined standard proportion of the number of incorrect words, the cause of the error of the target student is classified as vocabulary; if the proportion of the number of incorrect words due to phonetic reasons in the corresponding test paper text of a target student is greater than or equal to the predetermined standard proportion of the number of incorrect words, the cause of the error of the target student is classified as phonetic; if the proportion of the number of incorrect words due to phonetic reasons and the proportion of incorrect words due to vocabulary reasons in the corresponding test paper text of a target student are both within the predefined reference proportion range, the cause of the error of the target student is classified as vocabulary and phonetic.

9. The intelligent higher vocational English dictation system that facilitates real-time statistics according to claim 1 is characterized in that: The specific generation process of generating the dictation evaluation report corresponding to each student is as follows: If the corresponding test paper of a student is identified as a full-mark test paper, the reference dictation report of the full-mark test paper is selected from the reference dictation evaluation report templates stored in the cloud database and recorded as the dictation evaluation report of the student; If the cause of a target student's error is identified as phonetic, the reference dictation evaluation report templates stored in the cloud database are selected as the dictation evaluation report for the target student that needs to strengthen phonetic training; If the cause of a target student's error is identified as vocabulary, the reference dictation evaluation report templates stored in the cloud database are selected as the dictation evaluation report for the target student that needs to strengthen vocabulary training; If the cause of a target student's error is identified as vocabulary and phonetics, based on the reference dictation evaluation report templates stored in the cloud database, the reference dictation reports that require enhanced vocabulary and phonetics training are selected as the dictation evaluation reports for the target student.

Images