Ideological and political education monitoring method based on semantic analysis

Abstract

The invention discloses an ideological and political education monitoring method based on semantic analysis, and belongs to the technical field of semantic processing. The method comprises the following steps: extracting ideological and political education video texts, dividing into sequences, and recording corresponding voices; counting word frequency to construct a frequency set, calculating word following values according to the frequency and a voice amplitude fluctuation value, and forming a sequence; then abnormal and normal pair phrases are screened, and a sequence feature adjustment coefficient is calculated; and finally, inputting the phonetic word following sequence into the evaluation model, and performing feature adjustment and initial score optimization based on the adjustment coefficient to obtain an ideological and political education quality score.

Description

Technical Field

[0001] The present invention relates to the technical field of semantic processing, and in particular to an ideological and political education monitoring method based on semantic analysis. Background Art

[0002] With the rapid development of information technology, online education and video teaching have gradually become important forms of ideological and political education. However, how to effectively monitor and evaluate the teaching quality of ideological and political education videos. Existing ideological and political education monitoring technologies mainly rely on traditional questionnaires and classroom observations. These methods are often subjective and inefficient, and it is difficult to fully reflect the students' true learning status. In addition, when analyzing video content, existing technologies usually only focus on a single dimension of text or voice, and fail to make full use of the correlation between text and voice for comprehensive evaluation. This leads to a lack of accuracy in the evaluation of the quality of ideological and political education. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides an ideological and political education monitoring method based on semantic analysis, which solves the problem of low accuracy in evaluating the quality of ideological and political education in the prior art.

[0004] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a method for monitoring ideological and political education based on semantic analysis, comprising the following steps:

[0005] S1, extracting text from ideological and political education videos, dividing the text into text sequences by sentence, and recording the voice of reading the text sequence to obtain a voice set;

[0006] S2, counting the frequency of each word in the text sequence appearing in the ideological and political education text set, and constructing a frequency set;

[0007] S3, according to the fluctuation value of the frequency in the frequency set and the fluctuation value of the amplitude in the speech set, calculate the sound-word following value and construct the sound-word following sequence;

[0008] S4, according to the sound-word following value in the sound-word following sequence, screen out abnormal word pairs and normal word pairs, and calculate the sequence feature adjustment coefficient;

[0009] S5. Input the audio and word following sequence corresponding to the ideological and political education video into the ideological and political education quality assessment model, perform feature adjustment based on the sequence feature adjustment coefficient, and optimize the initial score based on the temporal change of the sequence feature adjustment coefficient to obtain the ideological and political education quality score.

[0010] Furthermore, S3 includes the following sub-steps:

[0011] S31, subtract the frequency of the i-th word in the frequency set from the frequency of the i+1-th word to obtain the i-th frequency fluctuation value, where i is a number;

[0012] S32, dividing each frequency fluctuation value by the maximum frequency fluctuation value to obtain a normalized frequency fluctuation value;

[0013] S33, extracting the speech segment corresponding to each word in the speech set, and extracting the amplitude mean value for each speech segment;

[0014] S34, subtracting the amplitude mean of the i-th word from the amplitude mean of the i+1-th word to obtain the i-th amplitude fluctuation value;

[0015] S35, dividing the i-th amplitude fluctuation value by the maximum amplitude fluctuation value to obtain a normalized amplitude fluctuation value;

[0016] S36, calculating the sound word following value according to the normalized frequency fluctuation value and the normalized amplitude fluctuation value of the same number;

[0017] S37, arranging the sound-word following values ​​in sequence to construct a sound-word following sequence.

[0018] Further, S36 includes the following sub-steps:

[0019] S361, calculating a difference coefficient according to the normalized frequency fluctuation value and the normalized amplitude fluctuation value of the same number;

[0020] S362, calculating a directional coefficient according to a normalized frequency fluctuation value and a normalized amplitude fluctuation value of the same number;

[0021] S363, multiply the direction coefficient by the difference coefficient to obtain the sound-word following value.

[0022] Furthermore, the formula for calculating the coefficient of difference in S361 is: , where ε i is the i-th difference coefficient, F f,i is the ith normalized frequency fluctuation value, F a,i is the ith normalized amplitude fluctuation value, | | is the absolute value operation;

[0023] The process of calculating the directional coefficient in S362 is specifically as follows: f,i and F a,i When both are greater than 0 or less than 0, the directional coefficient is assigned a value of 1; f,i and F a,i When one is negative and the other is positive, the directional coefficient is assigned a value of -1.

[0024] Further, S4 includes the following sub-steps:

[0025] S41, selecting two words corresponding to the sound-word following values ​​greater than 0 in the sound-word following sequence to form a normal pair of words;

[0026] S42, selecting two words corresponding to the sound-word following values ​​less than 0 in the sound-word following sequence to form an abnormal pair of words;

[0027] S43, searching for paragraphs where the phrases appear normally in the ideological and political education text set, and counting the frequency of the combination and the paragraph length ratio;

[0028] S44, searching for paragraphs where the abnormal pair of words appears in the ideological and political education text set, and counting the frequency of the combination and the paragraph length ratio;

[0029] S45. Calculate the ideological and political value index of normal phrases and abnormal phrases respectively according to the frequency of occurrence of the combination and the paragraph length ratio;

[0030] S46. Calculate the sequence feature adjustment coefficient based on the ideological and political value index of the normal phrase pair and the abnormal phrase pair, and the corresponding sound-word following value in the sound-word following sequence.

[0031] Furthermore, the formula for calculating the ideological and political value index of a normal phrase in S45 is: , where γ no,n is the ideological and political value index of the nth normal pair of phrases, f no,n,m is the frequency of the nth normal phrase combination appearing in the mth paragraph, P no,n,m is the ratio of the length of the mth paragraph in the nth normal phrase combination to the length of the ideological and political education text set, N no,n is the number of paragraphs where the nth normal pair of phrases appears, e is a natural constant, and m and n are positive integers;

[0032] The formula for calculating the ideological and political value index of abnormal phrases in S45 is: , where γ ab,n is the ideological and political value index of the nth abnormal pair of phrases, f ab,n,m is the frequency of the nth abnormal phrase combination appearing in the mth paragraph, P ab,n,m N is the ratio of the length of the mth paragraph in the ideological and political education text set where the nth abnormal phrase combination appears. ab,n The number of paragraphs in which the nth abnormal phrase combination appears.

[0033] Further, S46 includes the following sub-steps:

[0034] S461, calculating a normal valuation based on the normal ideological and political value index of the phrase and the corresponding sound-word following value in the sound-word following sequence;

[0035] S462, calculating anomaly valuation according to the ideological and political value index of the anomaly to the phrase and the corresponding sound-word following value in the sound-word following sequence;

[0036] S463. Add the normal estimate and the normal estimate to obtain the sequence characteristic adjustment coefficient.

[0037] Furthermore, the formula for calculating the normal valuation in S461 is: , where P no is the normal valuation, γ no,n is the ideological and political value index of the nth normal phrase, μ n is the sound-word following value corresponding to the nth normal phrase, M no is the number of normal phrases, n is a positive integer;

[0038] The formula for calculating the abnormal valuation in S462 is: , where P ab is the abnormal valuation, γ ab,k is the ideological and political value index of the kth abnormal pair of phrases, μ k is the phonetic word following value corresponding to the kth abnormal phrase, M ab is the number of abnormal phrases, and k is a positive integer.

[0039] Furthermore, the ideological and political education quality assessment model in S5 includes: multiple feature extraction modules, adders, and scoring optimization layers;

[0040] Each feature extraction module is used to process a sound-word following sequence, and adjust the sequence feature based on the sequence feature adjustment coefficient to obtain a sequence adjustment feature value;

[0041] The adder is used to add the adjusted feature values ​​of each sequence to obtain an initial score;

[0042] The scoring optimization layer is used to optimize the initial score according to the temporal changes of the characteristic adjustment coefficients of each sequence to obtain the ideological and political education quality score.

[0043] Furthermore, each feature extraction module includes a BiLSTM layer, a fully connected layer, and a sequence feature adjustment layer connected in sequence;

[0044] The expression of the sequence feature adjustment layer is: , where H is the sequence adjustment feature value output by the sequence feature adjustment layer, h is the sequence feature value output by the fully connected layer, tanh is the hyperbolic tangent function, and φ is the sequence feature adjustment coefficient;

[0045] The expression of the score optimization layer is: , where S is the ideological and political education quality score output by the score optimization layer, S in is the initial score, φA is the mean of the sequence characteristic adjustment coefficient in the second half of the period, φ B is the mean of the sequence characteristic adjustment coefficient in the first half of the period, and D is the normalization parameter.

[0046] The beneficial effects of the present invention are:

[0047] 1. The present invention extracts texts from ideological and political education videos and divides them into text sequences, while recording corresponding voices to obtain voice sets, and collects data from two dimensions: text and voice. This changes the previous limitation of focusing on only a single dimension, and can more comprehensively obtain information related to the ideological and political education process.

[0048] 2. The present invention calculates the sound-word following value and constructs the sound-word following sequence based on the fluctuation value of the frequency set and the voice set, making full use of the correlation between text and voice. Since the sound-word following situation can reflect the key content of education and the emphasis on voice, when teachers explain key knowledge, they will emphasize it through the weight of voice. By analyzing the sound-word following sequence, these key contents and the emphasis of teachers on voice can be clearly identified, so as to better understand the teaching process and provide a key basis for the evaluation of teaching effectiveness.

[0049] 3. The present invention selects abnormal and normal pairs of words based on the sound-word following sequence, and calculates the sequence feature adjustment coefficient, which can quantify the abnormal and normal states in the teaching process.

[0050] 4. The present invention inputs the sound-word following sequence into the ideological and political education quality evaluation model, combines the sequence feature adjustment coefficient to perform feature adjustment and initial score optimization, and comprehensively considers the quality of ideological and political education from multiple angles. By dynamically adjusting features and optimizing scores, the disadvantages of strong subjectivity and low efficiency of traditional evaluation methods are overcome, so that the final ideological and political education quality score can better reflect the actual teaching effect and improve the accuracy of the evaluation of ideological and political education quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flowchart of an ideological and political education monitoring method based on semantic analysis;

[0052] Figure 2 This is a structural diagram of the ideological and political education quality evaluation model. DETAILED DESCRIPTION

[0053] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0054] like Figure 1 As shown, a method for monitoring ideological and political education based on semantic analysis includes the following steps:

[0055] S1, extracting text from ideological and political education videos, dividing the text into text sequences by sentence, and recording the voice of reading the text sequence to obtain a voice set;

[0056] S2, counting the frequency of each word in the text sequence appearing in the ideological and political education text set, and constructing a frequency set;

[0057] S3, according to the fluctuation value of the frequency in the frequency set and the fluctuation value of the amplitude in the speech set, calculate the sound-word following value and construct the sound-word following sequence;

[0058] S4, according to the sound-word following value in the sound-word following sequence, screen out abnormal word pairs and normal word pairs, and calculate the sequence feature adjustment coefficient;

[0059] S5. Input the audio and word following sequence corresponding to the ideological and political education video into the ideological and political education quality assessment model, perform feature adjustment based on the sequence feature adjustment coefficient, and optimize the initial score based on the temporal change of the sequence feature adjustment coefficient to obtain the ideological and political education quality score.

[0060] In the present invention, each word in the text sequence in S2 refers to the words after removing the conjunctions. For example, "Vocational education aims to cultivate professional and skilled talents who can adapt to social needs". After removing the conjunctions "aims at", "of", and "adapt to", the words in the text sequence include: vocational education, training, social needs, and professional and skilled talents.

[0061] In this embodiment, S3 is specifically as follows: first, subtract the frequencies of adjacent words in the frequency set to obtain the frequency fluctuation value, then divide each frequency fluctuation value by the maximum frequency fluctuation value to complete normalization, then extract the amplitude mean of the speech segment corresponding to each word in the speech set, subtract the amplitude mean of adjacent words to obtain the amplitude fluctuation value and divide it by the maximum amplitude fluctuation value for normalization, then calculate the sound word following value using the normalized frequency fluctuation value and the normalized amplitude fluctuation value, and arrange each sound word following value in sequence to construct a sound word following sequence.

[0062] In this embodiment, S3 includes the following sub-steps:

[0063] S31, subtract the frequency of the i-th word in the frequency set from the frequency of the i+1-th word to obtain the i-th frequency fluctuation value, where i is a number;

[0064] S32, dividing each frequency fluctuation value by the maximum frequency fluctuation value to obtain a normalized frequency fluctuation value;

[0065] S33, extracting the speech segment corresponding to each word in the speech set, and extracting the amplitude mean value for each speech segment;

[0066] S34, subtracting the amplitude mean of the i-th word from the amplitude mean of the i+1-th word to obtain the i-th amplitude fluctuation value;

[0067] S35, dividing the i-th amplitude fluctuation value by the maximum amplitude fluctuation value to obtain a normalized amplitude fluctuation value;

[0068] S36, calculating the sound word following value according to the normalized frequency fluctuation value and the normalized amplitude fluctuation value of the same number;

[0069] S37, arranging the sound-word following values ​​in sequence to construct a sound-word following sequence.

[0070] In text analysis, the present invention can intuitively show the changes in vocabulary frequency in ideological and political education texts by calculating frequency fluctuation values, helping to identify the key transitions and core content of the content. In terms of speech analysis, extracting the mean amplitude of the speech segment and calculating the fluctuation value can reflect the changes in the teacher's speech intensity during the explanation process, which is usually related to the emphasized content. Normalization processing allows speech data and text data to be analyzed at the same scale. The present invention calculates the sound-word following value to closely link text and speech. When the sound-word following value is large, it means that the teacher's voice emphasis is highly consistent with the text focus, and the teaching effect is better. The sound-word following sequence finally constructed records the dynamic changes of this sound-word relationship during the teaching process.

[0071] In this embodiment, the first implementation method of S36 is specifically: take the absolute value of the difference between the normalized frequency fluctuation value and the amplitude fluctuation value, and take the inverse of the absolute value to obtain the difference coefficient. When the normalized frequency fluctuation value and the amplitude fluctuation value are both positive or negative, the direction coefficient is assigned to 1. When the normalized frequency fluctuation value and the amplitude fluctuation value are both positive and negative, the direction coefficient is assigned to -1. Multiply the difference coefficient and the direction coefficient to obtain the sound word following value.

[0072] Preferably, the second implementation method of S36 is as follows: the difference coefficient is calculated according to the normalized frequency fluctuation value and the amplitude fluctuation value; the direction coefficient is calculated according to the positive and negative values ​​of the two; and finally, the two are multiplied to obtain the sound word following value.

[0073] The second implementation method specifically includes the following steps:

[0074] S361, calculating a difference coefficient according to the normalized frequency fluctuation value and the normalized amplitude fluctuation value of the same number;

[0075] S362, calculating a directional coefficient according to a normalized frequency fluctuation value and a normalized amplitude fluctuation value of the same number;

[0076] S363. Multiply the direction coefficient by the difference coefficient to obtain the sound-word following value.

[0077] In this embodiment, the formula for calculating the difference coefficient in S361 is: , where ε i is the i-th coefficient of difference, F f,i is the ith normalized frequency fluctuation value, F a,i is the ith normalized amplitude fluctuation value, | | is the absolute value operation;

[0078] The process of calculating the directional coefficient in S362 is specifically as follows: f,i and F a,i When both are greater than 0 or less than 0, the directional coefficient is assigned a value of 1; f,i and F a,i When one is negative and the other is positive, the directional coefficient is assigned a value of -1.

[0079] The present invention can accurately quantify the degree of difference between the two by calculating the difference coefficient; when calculating the direction coefficient, it assigns a value based on the positive and negative relationship between the two, and meticulously describes the correlation between their changing directions.

[0080] In this embodiment, S4 is specifically as follows: according to the positive and negative values ​​in the sound-word following sequence, the corresponding words are respectively screened into normal phrases and abnormal phrases; then, the paragraphs in which the two types of phrases appear in combination in the ideological and political education text set are searched, and their occurrence frequency and paragraph length ratio are counted; the ideological and political value index of the normal phrases and the abnormal phrases are respectively calculated; finally, the sequence feature adjustment coefficient is calculated by combining the ideological and political value index with the sound-word following value.

[0081] S4 includes the following sub-steps:

[0082] S41, selecting two words corresponding to the sound-word following values ​​greater than 0 in the sound-word following sequence to form a normal pair of words;

[0083] S42, selecting two words corresponding to the sound-word following values ​​less than 0 in the sound-word following sequence to form an abnormal pair of words;

[0084] S43, searching for paragraphs where the phrases appear normally in the ideological and political education text set, and counting the frequency of the combination and the paragraph length ratio;

[0085] S44, searching for paragraphs where the abnormal pair of words appears in the ideological and political education text set, and counting the frequency of the combination and the paragraph length ratio;

[0086] S45. Calculate the ideological and political value index of normal phrases and abnormal phrases respectively according to the frequency of occurrence of the combination and the paragraph length ratio;

[0087] S46. Calculate the sequence feature adjustment coefficient based on the ideological and political value index of the normal phrase pair and the abnormal phrase pair, and the corresponding sound-word following value in the sound-word following sequence.

[0088] In the present invention, "combination" refers to two words in a normal phrase or an abnormal phrase appearing simultaneously in a sentence of the ideological and political education text set.

[0089] According to the sub-steps of step S3, the i-th sound-word following value corresponds to the i-th word and the i+1-th word. Therefore, the present invention can effectively distinguish between the content parts with good and poor coordination between voice and text in ideological and political education by screening the positive and negative sound-word following values, and the frequency of occurrence of phrases in the text collection and the ratio of paragraph length, and then calculate the ideological and political value index, thereby realizing the quantitative evaluation of the importance and influence of different phrases in ideological and political education.

[0090] In this embodiment, the formula for calculating the ideological and political value index of a normal phrase in S45 is: , where γ no,n is the ideological and political value index of the nth normal pair of phrases, f no,n,m is the frequency of the nth normal phrase combination appearing in the mth paragraph, P no,n,m is the ratio of the length of the mth paragraph in the nth normal phrase combination to the length of the ideological and political education text set, N no,n is the number of paragraphs where the nth normal pair of phrases appears, e is a natural constant, and m and n are positive integers;

[0091] The formula for calculating the ideological and political value index of abnormal phrases in S45 is: , where γ ab,n is the ideological and political value index of the nth abnormal pair of phrases, f ab,n,m is the frequency of the nth abnormal phrase combination appearing in the mth paragraph, P ab,n,m N is the ratio of the length of the mth paragraph in the ideological and political education text set where the nth abnormal phrase combination appears. ab,n The number of paragraphs in which the nth abnormal phrase combination appears.

[0092] The present invention comprehensively considers factors such as the frequency of occurrence of phrases in different paragraphs, the length ratio of paragraphs, and the number of paragraphs, and uses an index formula to quantify the value of normal and abnormal phrases in ideological and political education. Phrases with high frequency and large length ratio have higher corresponding indexes, which can accurately measure their importance and make the evaluation more objective and accurate.

[0093] In this embodiment, S46 includes the following sub-steps:

[0094] S461, calculating a normal valuation according to the ideological and political value index of the normal pair of phrases and the corresponding sound-word following value in the sound-word following sequence, wherein the normal valuation calculation process is: multiplying the ideological and political value index of each normal pair of phrases by its corresponding sound-word following value, and then summing the product results of all normal pair of phrases;

[0095] S462, calculating an abnormal valuation according to the ideological and political value index of the abnormal pair of phrases and the corresponding sound-word following value in the sound-word following sequence, wherein the abnormal valuation calculation process is: multiplying the ideological and political value index of each abnormal pair of phrases by its corresponding sound-word following value, and then summing the product results of all abnormal pair of phrases;

[0096] S463. Add the normal estimate and the normal estimate to obtain the sequence characteristic adjustment coefficient.

[0097] In this embodiment, the formula for calculating the normal valuation in S461 is: , where P no is the normal valuation, γ no,n is the ideological and political value index of the nth normal phrase, μ n is the sound-word following value corresponding to the nth normal phrase, M no is the number of normal phrases, n is a positive integer;

[0098] The formula for calculating the abnormal valuation in S462 is: , where P ab is the abnormal valuation, γ ab,k is the ideological and political value index of the kth abnormal pair of phrases, μ k is the phonetic word following value corresponding to the kth abnormal phrase, M ab is the number of abnormal phrases, and k is a positive integer.

[0099] The present invention combines the ideological and political value index and the phonetic-word following value of normal and abnormal pairs of phrases to calculate the normal valuation and the abnormal valuation respectively, comprehensively considering the value of the content in ideological and political education and the coordination of voice and text, and then comprehensively evaluates the teaching status, which can more accurately reflect the actual effect of ideological and political teaching.

[0100] like Figure 2 As shown, the ideological and political education quality assessment model in S5 includes: multiple feature extraction modules, adders and scoring optimization layers;

[0101] Each feature extraction module is used to process a sound-word following sequence, and adjust the sequence feature based on the sequence feature adjustment coefficient to obtain a sequence adjustment feature value;

[0102] The adder is used to add the adjusted feature values ​​of each sequence to obtain an initial score;

[0103] The scoring optimization layer is used to optimize the initial score according to the temporal changes of the characteristic adjustment coefficients of each sequence to obtain the ideological and political education quality score.

[0104] In this embodiment, each feature extraction module includes a BiLSTM layer, a fully connected layer, and a sequence feature adjustment layer connected in sequence;

[0105] The expression of the sequence feature adjustment layer is: , where H is the sequence adjustment feature value output by the sequence feature adjustment layer, h is the sequence feature value output by the fully connected layer, tanh is the hyperbolic tangent function, and φ is the sequence feature adjustment coefficient;

[0106] The expression of the score optimization layer is: , where S is the ideological and political education quality score output by the score optimization layer, S in is the initial score, φ A is the mean of the sequence characteristic adjustment coefficient in the second half of the period, φ B is the mean of the sequence characteristic adjustment coefficient in the first half of the period, and D is the normalization parameter.

[0107] In this embodiment, the normalization parameters are set according to experiments or requirements.

[0108] In an ideological and political education video, there may be text sequences at multiple time points when divided by sentences. Therefore, the time is divided into the first half and the second half.

[0109] The present invention uses each BiLSTM layer to process a sound-word following sequence, extracts eigenvalues, and then uses a fully connected layer to synthesize various eigenvalues ​​to obtain the sequence eigenvalues ​​output by the fully connected layer. The sequence feature adjustment layer adjusts h according to the size of the sequence feature adjustment coefficient φ, and uses a hyperbolic tangent function to map the sequence feature adjustment coefficient φ to between -1 and 1, so that the adjustment is adaptive. The larger the sequence feature adjustment coefficient φ is, the more significant the enhancement effect on the sequence eigenvalue h is, which means that the center of gravity of the speech segment is better matched with the text.

[0110] In this embodiment, the BiLSTM layer may also be replaced by an LSTM layer.

[0111] The present invention compares the mean of the characteristic adjustment coefficient of the second half of the time and the first half of the time series (φ A and φ B ), can effectively capture the dynamic changes of teaching conditions at different stages in the ideological and political education process. A -φ B ≥0, indicating that the teaching characteristics in the second half have changed positively compared with the first half, and the score will be improved accordingly; on the contrary, if φ A -φ BIf it is less than 0, it means that there may be problems in the second half of the teaching and the score will be lowered, thus dynamically reflecting the quality changes in the teaching process.

[0112] The present invention optimizes the initial scoring based on the dynamic changes of the teaching process, changing the limitations of a single static scoring method. This method takes into account the stage differences of teaching, making the final ideological and political education quality scoring more in line with the actual teaching effect and improving the accuracy and reliability of the scoring.

[0113] The present invention extracts texts from ideological and political education videos and divides them into text sequences, while recording corresponding voices to obtain voice sets, and collects data from two dimensions: text and voice. This changes the previous limitation of focusing on only a single dimension, and can more comprehensively obtain information related to the ideological and political education process.

[0114] The present invention calculates the sound-word following value and constructs the sound-word following sequence based on the fluctuation value of the frequency set and the voice set, making full use of the correlation between text and voice. Since the sound-word following situation can reflect the key content of education and the emphasis on voice, when teachers explain key knowledge, they will emphasize it through the weight of voice. By analyzing the sound-word following sequence, these key contents and the emphasis of teachers on voice can be clearly identified, so as to better understand the teaching process and provide a key basis for the evaluation of teaching effectiveness.

[0115] The present invention selects abnormal word pairs and normal word pairs based on the sound-word following sequence, and calculates the sequence feature adjustment coefficient, so as to quantify the abnormal and normal states in the teaching process.

[0116] The present invention inputs the sound-word following sequence into the ideological and political education quality evaluation model, combines the sequence feature adjustment coefficient to perform feature adjustment and initial score optimization, and comprehensively considers the quality of ideological and political education from multiple angles. By dynamically adjusting features and optimizing scores, the disadvantages of strong subjectivity and low efficiency of traditional evaluation methods are overcome, so that the final ideological and political education quality score can better reflect the actual teaching effect and improve the accuracy of the evaluation of ideological and political education quality.

[0117] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for monitoring ideological and political education based on semantic analysis, characterized in that: The following steps are involved: S1, extracting text from ideological and political education videos, dividing the text into text sequences by sentence, and recording the voice of reading the text sequence to obtain a voice set; S2, counting the frequency of each word in the text sequence appearing in the ideological and political education text set, and constructing a frequency set; S3, according to the fluctuation value of the frequency in the frequency set and the fluctuation value of the amplitude in the speech set, calculate the sound-word following value and construct the sound-word following sequence; S4, according to the sound-word following value in the sound-word following sequence, screen out abnormal word pairs and normal word pairs, and calculate the sequence feature adjustment coefficient; S5. Input the audio and word following sequence corresponding to the ideological and political education video into the ideological and political education quality assessment model, perform feature adjustment based on the sequence feature adjustment coefficient, and optimize the initial score based on the temporal change of the sequence feature adjustment coefficient to obtain the ideological and political education quality score.

2. The method for monitoring ideological and political education based on semantic analysis according to claim 1 is characterized in that: The S3 comprises the following sub-steps: S31, subtract the frequency of the i-th word in the frequency set from the frequency of the i+1-th word to obtain the i-th frequency fluctuation value, where i is a number; S32, dividing each frequency fluctuation value by the maximum frequency fluctuation value to obtain a normalized frequency fluctuation value; S33, extracting the speech segment corresponding to each word in the speech set, and extracting the amplitude mean value for each speech segment; S34, subtracting the amplitude mean of the i-th word from the amplitude mean of the i+1-th word to obtain the i-th amplitude fluctuation value; S35, dividing the i-th amplitude fluctuation value by the maximum amplitude fluctuation value to obtain a normalized amplitude fluctuation value; S36, calculating the sound word following value according to the normalized frequency fluctuation value and the normalized amplitude fluctuation value of the same number; S37, arranging the sound-word following values ​​in sequence to construct a sound-word following sequence.

3. The method for monitoring ideological and political education based on semantic analysis according to claim 2 is characterized in that: The S36 comprises the following sub-steps: S361, calculating a difference coefficient according to the normalized frequency fluctuation value and the normalized amplitude fluctuation value of the same number; S362, calculating a directional coefficient according to a normalized frequency fluctuation value and a normalized amplitude fluctuation value of the same number; S363. Multiply the direction coefficient by the difference coefficient to obtain the sound-word following value.

4. The method for monitoring ideological and political education based on semantic analysis according to claim 3 is characterized in that: The formula for calculating the coefficient of difference in S361 is: , where ε i is the i-th difference coefficient, F f,i is the ith normalized frequency fluctuation value, F a,i is the ith normalized amplitude fluctuation value, | | is the absolute value operation; The process of calculating the directional coefficient in S362 is specifically as follows: f,i and F a,i When both are greater than 0 or less than 0, the directional coefficient is assigned a value of 1; f,i and F a,i When one is negative and the other is positive, the directional coefficient is assigned a value of -1.

5. The method for monitoring ideological and political education based on semantic analysis according to claim 1 is characterized in that: The S4 comprises the following sub-steps: S41, selecting two words corresponding to the sound-word following values ​​greater than 0 in the sound-word following sequence to form a normal pair of words; S42, selecting two words corresponding to the sound-word following values ​​less than 0 in the sound-word following sequence to form an abnormal pair of words; S43, searching for paragraphs where the phrases appear normally in the ideological and political education text set, and counting the frequency of the combination and the paragraph length ratio; S44, searching for paragraphs where the abnormal pair of words appears in the ideological and political education text set, and counting the frequency of the combination and the paragraph length ratio; S45. Calculate the ideological and political value index of normal phrases and abnormal phrases respectively according to the frequency of occurrence of the combination and the paragraph length ratio; S46. Calculate the sequence feature adjustment coefficient based on the ideological and political value index of the normal phrase pair and the abnormal phrase pair, and the corresponding sound-word following value in the sound-word following sequence.

6. The method for monitoring ideological and political education based on semantic analysis according to claim 5 is characterized in that: The formula for calculating the ideological and political value index of a normal phrase in S45 is: , where γ no,n is the ideological and political value index of the nth normal pair of phrases, f no,n,m is the frequency of the nth normal phrase combination appearing in the mth paragraph, P no,n,m is the ratio of the length of the mth paragraph in the nth normal phrase combination to the length of the ideological and political education text set, N no,n is the number of paragraphs where the nth normal pair of phrases appears, e is a natural constant, and m and n are positive integers; The formula for calculating the ideological and political value index of the abnormal phrase in S45 is: , where γ ab,n is the ideological and political value index of the nth abnormal pair of phrases, f ab,n,m is the frequency of the nth abnormal phrase combination appearing in the mth paragraph, P ab,n,m N is the ratio of the length of the mth paragraph in the ideological and political education text set where the nth abnormal phrase combination appears. ab,n The number of paragraphs in which the nth abnormal phrase combination appears.

7. The method for monitoring ideological and political education based on semantic analysis according to claim 5 is characterized in that: The S46 comprises the following sub-steps: S461, calculating a normal valuation based on the normal ideological and political value index of the phrase and the corresponding sound-word following value in the sound-word following sequence; S462, calculating anomaly valuation according to the ideological and political value index of the anomaly to the phrase and the corresponding sound-word following value in the sound-word following sequence; S463. Add the normal estimate and the normal estimate to obtain the sequence characteristic adjustment coefficient.

8. The method for monitoring ideological and political education based on semantic analysis according to claim 7 is characterized in that: The formula for calculating the normal valuation in S461 is: , where P no is the normal valuation, γ no,n is the ideological and political value index of the nth normal phrase, μ n is the sound-word following value corresponding to the nth normal phrase, M no is the number of normal phrases, n is a positive integer; The formula for calculating the abnormal valuation in S462 is: , where P ab is the abnormal valuation, γ ab,k is the ideological and political value index of the kth abnormal pair of phrases, μ k is the phonetic word following value corresponding to the kth abnormal phrase, M ab is the number of abnormal phrases, and k is a positive integer.

9. The method for monitoring ideological and political education based on semantic analysis according to claim 1 is characterized in that: The ideological and political education quality assessment model in S5 includes: multiple feature extraction modules, adders and scoring optimization layers; Each of the feature extraction modules is used to process a sound-word following sequence, and to adjust the sequence features based on the sequence feature adjustment coefficient to obtain a sequence adjustment feature value; The adder is used to add the adjusted characteristic values ​​of each sequence to obtain an initial score; The scoring optimization layer is used to optimize the initial scoring according to the temporal changes of each sequence feature adjustment coefficient to obtain the ideological and political education quality score.

10. The method for monitoring ideological and political education based on semantic analysis according to claim 9 is characterized in that: Each of the feature extraction modules includes a BiLSTM layer, a fully connected layer, and a sequence feature adjustment layer connected in sequence; The expression of the sequence feature adjustment layer is: , where H is the sequence adjustment feature value output by the sequence feature adjustment layer, h is the sequence feature value output by the fully connected layer, tanh is the hyperbolic tangent function, and φ is the sequence feature adjustment coefficient; The expression of the score optimization layer is: , where S is the ideological and political education quality score output by the score optimization layer, S in is the initial score, φ A is the mean of the sequence characteristic adjustment coefficient in the second half of the period, φ B is the mean of the sequence characteristic adjustment coefficient in the first half of the period, and D is the normalization parameter.

Images