Teaching quality optimization method and system based on AI intelligent correction and analysis

Abstract

The invention relates to the technical field of AI data analysis, in particular to a teaching quality optimization method and system based on AI intelligent correction and analysis, and the method comprises the steps: dividing a homework explanation video stream into a plurality of correction analysis time periods according to a plurality of explanation correction feature frames in the homework explanation video stream; the method comprises the steps of determining a correction content tendency category of a homework explanation video stream according to a content value coefficient difference condition of each correction analysis period, determining an evaluation factor of the homework explanation video stream based on the correction content tendency category, determining whether configuration of teaching resources is reasonable or not according to the evaluation factor, and generating resource supplement information recommended to students, and finally, the generated resource supplement information is sent to the student terminal, so that a dual-dimension evaluation system of analysis explanation and analysis annotation is constructed, the reasonability of teaching resource configuration is scientifically judged, the resource supplement information is generated, and the change of an experience-driven decision mode of a smart classroom to a data-driven decision mode is realized.

Description

Technical Field

[0001] The present invention relates to the field of AI data analysis technology, and in particular to a teaching quality optimization method and system based on AI intelligent correction and analysis. Background Art

[0002] In the current education sector, traditional teaching resource recommendations have significant limitations. Amidst the rapid development of educational informatization, optimizing teaching quality remains a core concern in the field. Traditional teaching quality assessments rely primarily on teacher experience and limited manual analysis, making it difficult to comprehensively and accurately evaluate the rationality of the teaching process and resource allocation. During the assignment grading and explanation phases, teachers often struggle to quantify the importance of different explanation contents. Teaching resource allocation lacks scientific data support, resulting in students potentially not receiving timely, targeted learning resource supplements. Furthermore, traditional methods struggle to systematically analyze teaching content and quickly identify resource imbalances within the teaching process. With the increasing application of AI technology in education, leveraging AI to intelligently analyze and optimize the teaching process has become a research hotspot. However, existing AI-based teaching quality optimization solutions still have significant room for improvement in terms of in-depth analysis of homework explanation videos, accurate evaluation of teaching resource allocation, and personalized resource recommendations for students, in order to meet the urgent need for improved teaching quality in modern education.

[0003] For example, China Patent Publication No.: CN119672739A, the invention discloses a homework grading method based on image recognition and a large language model, which relates to the field of teaching and solves the problem of insufficient practicality of the homework grading method, including step S1: obtaining multiple homework texts and multiple answer texts to obtain preliminary homework collection data, step S2: dividing the multiple homework texts into a first type of homework text and a second type of homework text, and grading the first type of homework text and the second type of homework text respectively to obtain homework correction data, step S3: performing answer correctness analysis on the homework submitted by the target student to obtain a first homework quality evaluation coefficient and a second homework quality evaluation coefficient to obtain correctness analysis data, step S4: performing completion quality evaluation on the current batch of homework submitted by the target student by analyzing the correctness analysis data. The present invention can expand the function of the homework grading method and further improve practicality.

[0004] The following problems also exist in the prior art:

[0005] Existing technologies do not consider quantitative evaluation of the matching degree between teachers' explanation content and analytical annotations. Existing technologies cannot construct a dual-dimensional evaluation system of analytical explanation and analytical annotations. There is a lack of in-depth analysis of the homework explanation process, resulting in unreasonable allocation of teaching resources and students' inability to obtain targeted learning resources to supplement knowledge gaps. Summary of the Invention

[0006] To this end, the present invention provides a teaching quality optimization method and system based on AI intelligent correction and analysis to overcome the problem that the existing technology cannot construct a dual-dimensional evaluation system of analysis and explanation and analysis and annotation, and lacks in-depth analysis of the homework explanation process.

[0007] To achieve the above objectives, the present invention provides a teaching quality optimization method based on AI intelligent correction and analysis, comprising:

[0008] Obtaining a video stream of a teacher explaining homework during the teaching process, determining a number of explanation and correction feature frames in the homework explanation video stream, and dividing the homework explanation video stream into a number of correction analysis periods based on the explanation and correction feature frames;

[0009] Determining the content value coefficient of each correction and analysis period according to the duration of the correction and analysis period, and determining the correction content tendency category of the homework explanation video stream according to the difference in the content value coefficient;

[0010] Determining evaluation factors for the homework explanation video stream based on the correction content tendency category, the evaluation factors including the correlation between the first format information and the second format information within the characteristic correction analysis period, and the correlation between the first associated parameter and the second associated parameter within the correction analysis period;

[0011] Wherein, the feature correction and analysis period is determined based on the content value coefficient;

[0012] Determining whether the configuration of teaching resources is reasonable based on the evaluation factors, and generating resource supplementary information recommended to students in response to a determination result that the configuration of teaching resources is unreasonable, wherein the resource supplementary information is first recommended supplementary information generated based on the first format information, or second recommended supplementary information generated based on the content of the heterogeneous association correction and analysis period;

[0013] The heterosexual association correction and analysis period is determined based on the difference between the first association parameter and the second association parameter;

[0014] The generated resource supplement information is sent to the student terminal.

[0015] Furthermore, a method for determining a plurality of explanation and correction feature frames is to obtain images of consecutive frames at preset intervals in the homework explanation video stream, and determine the moment when the correction questions in the image are switched as the moment when the explanation and correction feature frames are located.

[0016] Furthermore, the process of determining the content value coefficient includes:

[0017] Determining the period between two adjacent explanation and correction feature frames as the correction and parsing period;

[0018] The duration of each correction and analysis period is determined, and the ratio of the duration to the total duration of the homework explanation video stream is determined as the content value coefficient.

[0019] Furthermore, the process of determining the correction content tendency category of the homework explanation video stream includes:

[0020] Calculate the standard deviation of the content value coefficient for each revision analysis period;

[0021] According to the comparison result that the standard deviation satisfies the video category distinction condition, determining that the correction content tendency category of the homework explanation video stream is a correction tendency video category;

[0022] According to the comparison result that the standard deviation does not meet the video category distinction condition, determining that the correction content tendency category of the homework explanation video stream is a correction non-tendency video category;

[0023] The video category distinction condition is that the standard deviation exceeds a preset standard deviation comparison value.

[0024] Furthermore, the evaluation factors of the homework explanation video stream are determined as follows:

[0025] If the homework explanation video stream is a correction tendency video category, determining the evaluation factor as the correlation between the first format information and the second format information within the characteristic correction analysis period;

[0026] If the homework explanation video stream is a non-correction-oriented video category, the evaluation factor is determined to be a correlation coefficient between the first correlation parameter and the second correlation parameter during the correction analysis period.

[0027] Furthermore, the process of determining the relevance between the first format information and the second format information includes:

[0028] Filter the correction and analysis period corresponding to the maximum value coefficient of the content in the homework explanation video stream;

[0029] Acquire the voice information within the correction and analysis period and determine voice keywords based on semantic analysis technology, and determine the voice keywords as the first format information;

[0030] Acquire images corresponding to the explanation and correction feature frames in the correction and analysis period based on a time sequence relationship, determine keywords of newly added text in images corresponding to the explanation and correction feature frames in the preceding and following time sequences using text recognition technology, and determine the keywords as the second format information;

[0031] The Euclidean distance between the first format information and the second format information is determined as the correlation.

[0032] Furthermore, the process of determining whether the configuration of teaching resources is reasonable according to the relevance and generating resource supplementary information includes:

[0033] According to the comparison result that the correlation exceeds a preset correlation threshold, determining that the configuration of the teaching resources is unreasonable;

[0034] In response to a result that the configuration is determined to be unreasonable according to the relevance, generating resource supplementary information sent to the student as first recommended supplementary information generated according to the first format information;

[0035] The Euclidean distance between the content keyword of the first recommended supplementary information and the first format information does not exceed a preset reference Euclidean distance.

[0036] Furthermore, the process of determining the correlation relationship between the first correlation parameter and the second correlation parameter includes:

[0037] Determining the number of words converted from the speech information into text within each correction and analysis period based on speech recognition, and determining the number of words as the first associated parameter;

[0038] Acquire images corresponding to explanation and correction feature frames of each correction and analysis period based on a time sequence relationship, determine the number of words in the newly added text in the images corresponding to the explanation and correction feature frames of the preceding and subsequent time sequences using a text recognition technique, and determine the number of words in the text as the second associated parameter;

[0039] The correction and analysis time periods are sorted in order from small to large according to the first correlation parameters, and the second correlation parameters of each correction and analysis time period are sequentially combined into a correlation relationship sequence according to the sorting.

[0040] Furthermore, the process of determining whether the configuration of teaching resources is reasonable and generating resource supplementary information based on the association relationship includes:

[0041] If the association relationship sequence is not an increasing sequence, it is determined that the configuration of teaching resources is unreasonable;

[0042] In response to a result that the configuration is unreasonable according to the association relationship, the resource supplementary information sent to the student is the second recommended supplementary information generated according to the content of the heterosexual association correction and parsing period;

[0043] According to the comparison result that the difference between the first associated parameter and the second associated parameter in the correction analysis period exceeds the difference comparison value, the correction analysis period is determined as a heterosexual correlation correction analysis period, and the difference comparison value is the average value of the difference between the first associated parameter and the second associated parameter in several correction analysis periods;

[0044] Wherein, the Euclidean distance between the content keyword of the second recommended supplementary information and the content of the opposite-sex association correction and analysis period does not exceed a preset reference Euclidean distance.

[0045] Furthermore, the present invention also provides a teaching quality optimization system based on AI intelligent correction and analysis, including:

[0046] An information acquisition unit includes a video recorder for acquiring a video stream of homework explanations and a microphone for acquiring audio information. The video recorder is also coupled to a video processor for dividing the video stream of homework explanations into a number of correction and analysis periods.

[0047] a feature extraction unit connected to the information acquisition unit, configured to determine the content value coefficient of each correction analysis period and to determine the correction content tendency category of the homework explanation video stream based on the difference in the content value coefficient;

[0048] an intelligent evaluation unit, connected to the information acquisition unit and the feature extraction unit, respectively, for determining evaluation factors of the homework explanation video stream based on the correction content tendency category;

[0049] a content recommendation generating unit, connected to the intelligent evaluation unit, for determining whether the configuration of teaching resources is reasonable based on the evaluation factors, and generating resource supplementary information recommended to students;

[0050] A transmission unit is connected to the content recommendation generation unit and is used to send the generated resource supplement information to the student terminal.

[0051] Compared with the prior art, the beneficial effect of the present invention lies in that the present invention divides the homework explanation video stream into several correction and analysis periods according to several explanation and correction feature frames in the homework explanation video stream, determines the correction content tendency category of the homework explanation video stream according to the difference in content value coefficients of each correction and analysis period, determines the evaluation factors of the homework explanation video stream based on the correction content tendency category, determines whether the configuration of teaching resources is reasonable according to the evaluation factors, and generates resource supplement information recommended to students, and finally sends the generated resource supplement information to the student terminal, thereby realizing the construction of a dual-dimensional evaluation system of analysis and explanation and analysis and annotation, scientifically determining the rationality of teaching resource configuration and generating resource supplement information, and realizing the transformation of smart classroom from experience-driven to data-driven decision-making mode.

[0052] Furthermore, the present invention obtains continuous frame images and determines the switching moment of correction questions as the explanation and correction feature frame. This method can clearly and accurately divide the homework explanation video stream into different correction and analysis time periods. By determining the time period between adjacent explanation and correction feature frames as the correction and analysis time period, and using the ratio of its duration to the total video duration as the content value coefficient, a quantitative evaluation of the importance of each correction and analysis time period in the entire homework correction process is achieved. By calculating the standard deviation of the content value coefficient of each correction and analysis time period, and determining the correction content tendency category based on the relationship between the standard deviation and the preset standard deviation comparison value, it is helpful to adopt more targeted teaching resource allocation strategies for different types of explanation videos.

[0053] Furthermore, the present invention targets the correction tendency video category and selects the feature correction analysis period corresponding to the maximum value coefficient of the content as the key to achieving the teaching goal. By focusing on this core period, it avoids wasting computing power due to paying too much attention to non-critical content. The voice keywords determined by semantic analysis technology during the correction analysis period are used as the first format information, and the keyword text of the newly added text in the explanation and correction feature frame image obtained based on text recognition technology is used as the second format information. Key information is extracted from two important dimensions: the teacher's explanation voice and the newly added text of the correction content update. The voice keywords reflect the key content conveyed orally by the teacher, while the keyword text of the newly added text reflects the key knowledge points presented in written form. This multi-dimensional method of extracting key information comprehensively and deeply covers the important content carried by different carriers in the teaching process, realizes the construction of a dual-dimensional evaluation system of analysis and explanation and analysis and annotation, and realizes the transformation of smart classroom from experience-driven to data-driven decision-making mode.

[0054] Furthermore, after determining that the configuration of teaching resources is unreasonable, the present invention generates first recommended supplementary information based on the first format information, fully considering the students' learning needs for the key content explained by the teacher. The first format information is the key information obtained by semantic analysis technology of the teacher's teaching voice, representing the core knowledge conveyed orally by the teacher. The resource supplementary information generated based on this can accurately supplement the deficiencies that may exist in students' understanding of the teacher's analysis and explanation of the homework, and requires that the Euclidean distance between the content keywords of the first recommended supplementary information and the first format information does not exceed the preset reference Euclidean distance, which ensures the close correlation between the supplementary information and the teacher's explanation points, and ensures the consistency of the supplementary resources with the teaching points in content, so that students can efficiently obtain valuable learning resources, form a dynamic and adaptive teaching resource ecosystem, and realize the transformation of smart classrooms from experience-driven to data-driven decision-making models.

[0055] Furthermore, for correcting non-tendency video categories, the present invention uses voice recognition and text recognition technology to obtain relevant data when determining the association relationship, and constructs an association relationship series after sorting the first association parameter, which can intuitively show the relationship between the teacher's explanation amount and the text update amount of homework correction, and prevent the teaching effect from being affected by configuration imbalance. In terms of generating resource supplementary information, the second recommended supplementary information is generated according to the content of the heterogeneous association correction analysis period, and the time period with abnormal information output during the teaching process is accurately located. These time periods are determined by calculating the average value of the difference in association parameters, which can effectively make up for the weak links in teaching. At the same time, the Euclidean distance between the keywords of the supplementary information content and the content of the heterogeneous association correction analysis period is required to meet the preset standard, which ensures that the supplementary information is closely related to the teaching content of the problem period, and realizes the transformation of the smart classroom from an experience-driven to a data-driven decision-making model. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a step diagram of a teaching quality optimization method based on AI intelligent correction and analysis according to an embodiment of the present invention;

[0057] Figure 2 A logic flow chart for determining the tendency category of the correction content of a homework explanation video stream according to an embodiment of the present invention;

[0058] Figure 3 A diagram illustrating the steps of determining the correlation between first format information and second format information according to an embodiment of the present invention;

[0059] Figure 4 A diagram showing steps for determining a correlation relationship between a first correlation parameter and a second correlation parameter according to an embodiment of the present invention;

[0060] Figure 5 It is a system block diagram of the teaching quality optimization system based on AI intelligent correction and analysis in an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0062] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0063] It should be noted that, in the description of the present invention, terms such as "upper", "lower", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0064] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "connected" and "connection" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; mechanical connections, electrical connections; direct connections, indirect connections through an intermediate medium, and internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0065] See also Figure 1 As shown in FIG, it is a step diagram of a teaching quality optimization method based on AI intelligent correction and analysis according to an embodiment of the present invention. The teaching quality optimization method based on AI intelligent correction and analysis according to the present invention includes:

[0066] Step S100: obtaining a video stream of a teacher explaining homework during the teaching process, determining a number of explanation and correction feature frames in the homework explanation video stream, and dividing the homework explanation video stream into a number of correction analysis periods based on the explanation and correction feature frames;

[0067] Step S200, determining the content value coefficient of each correction analysis period according to the duration of the correction analysis period, and determining the correction content tendency category of the homework explanation video stream according to the difference in the content value coefficient;

[0068] Step S300: determining evaluation factors for the homework explanation video stream based on the correction content tendency category, the evaluation factors including the correlation between the first format information and the second format information within the characteristic correction analysis period, and the correlation between the first correlation parameter and the second correlation parameter within the correction analysis period;

[0069] Wherein, the feature correction and analysis period is determined based on the content value coefficient;

[0070] Step S400: determining whether the configuration of teaching resources is reasonable based on the evaluation factors; and generating supplementary resource information recommended to the student in response to a determination that the configuration of teaching resources is unreasonable. The supplementary resource information may be first supplementary recommendation information generated based on the first format information, or second supplementary recommendation information generated based on the content of the heterogeneous association correction and analysis period.

[0071] The heterosexual association correction and analysis period is determined based on the difference between the first association parameter and the second association parameter;

[0072] Step S500: Send the generated resource supplement information to the student terminal.

[0073] Specifically, a method for determining several explanation and correction feature frames is to obtain images of consecutive frames at preset intervals in the homework explanation video stream, and determine the moment when the correction questions in the image are switched as the moment when the explanation and correction feature frames are located.

[0074] It will be understood by those skilled in the art that, in a video stream explaining homework, the switching of grading questions usually marks a change in the teaching content. For example, when transitioning from one grading question to another, continuous frame images are acquired at preset intervals in order to systematically and comprehensively scan the video content, and utilize the close connection between the switching of grading questions and the segmentation of teaching content to reasonably divide the video stream according to the changes in the actual teaching content.

[0075] In implementation, the preset interval length for obtaining continuous frame images can be set by technical personnel. In order to avoid missing the moment of switching between grading questions due to too long interval length, and excessive image analysis due to too short interval length, the preset interval length has a value range of [10, 30], and the interval unit is s. Preferably, the preset interval length can be set to 15s.

[0076] In implementation, the difference in question text can be identified by using fusion image recognition (OCR) technology on the acquired continuous frame images, and then the moment when the correction question is switched can be determined. The fusion image recognition (OCR) technology for identifying question text is an existing technology and will not be repeated here.

[0077] Specifically, the process of determining the content value coefficient includes:

[0078] Determining the period between two adjacent explanation and correction feature frames as the correction and parsing period;

[0079] The duration of each correction and analysis period is determined, and the ratio of the duration to the total duration of the homework explanation video stream is determined as the content value coefficient.

[0080] Understandably, the content value coefficient is determined by the ratio of the duration of the grading and analysis period to the total duration of the homework explanation video stream. The principle is that the time ratio can, to a certain extent, reflect the relative importance of the teaching content during that period in the overall teaching process. If a grading and analysis period lasts longer, it means that the teacher has invested more time in explanation during this period, and the teaching content during this period may have made a significant contribution to the achievement of the overall teaching objectives. This quantitative method converts the time factor in the teaching process into a measurable value.

[0081] Specifically, see Figure 2 As shown in FIG, it is a logic flow chart for determining the tendency category of the correction content of the homework explanation video stream. The process of determining the tendency category of the correction content of the homework explanation video stream includes:

[0082] Calculate the standard deviation of the content value coefficient for each revision analysis period;

[0083] According to the comparison result that the standard deviation satisfies the video category distinction condition, determining that the correction content tendency category of the homework explanation video stream is a correction tendency video category;

[0084] According to the comparison result that the standard deviation does not meet the video category distinction condition, determining that the correction content tendency category of the homework explanation video stream is a correction non-tendency video category;

[0085] The video category distinction condition is that the standard deviation exceeds a preset standard deviation comparison value.

[0086] In practice, the setting of the standard deviation comparison value requires screening out the correction and analysis periods with significantly longer durations. The standard deviation comparison value range is [0.07, 0.1]. Here, a preferred value of the standard deviation comparison value is provided, and the standard deviation comparison value is set to 0.08.

[0087] It is understandable that the larger the standard deviation of the content value coefficients of each correction and analysis period, the greater the difference in the content value coefficients of each correction and analysis period, that is, the content contribution of certain periods is outstanding. Conversely, the smaller the standard deviation, the closer the content value coefficients of each correction and analysis period, which means that the teaching content is relatively balanced in time distribution. By effectively classifying explanation videos with different teaching characteristics, it provides a basis for the subsequent formulation of targeted teaching resource allocation strategies.

[0088] Specifically, the present invention obtains continuous frame images at preset intervals and determines the switching moment of correction questions as the explanation and correction feature frame. This method can clearly and accurately divide the homework explanation video stream into different correction and analysis time periods. By determining the time period between adjacent explanation and correction feature frames as the correction and analysis time period, and taking the ratio of its duration to the total video duration as the content value coefficient, a quantitative evaluation of the importance of each correction and analysis time period in the entire homework correction process is achieved. By calculating the standard deviation of the content value coefficient of each correction and analysis time period, and determining the correction content tendency category based on the relationship between the standard deviation and the preset standard deviation comparison value, it is helpful to adopt more targeted teaching resource allocation strategies for different types of explanation videos.

[0089] Specifically, the evaluation factors of the homework explanation video stream are determined as follows:

[0090] If the homework explanation video stream is a correction tendency video category, determining the evaluation factor as the correlation between the first format information and the second format information within the characteristic correction analysis period;

[0091] If the homework explanation video stream is a non-correction-oriented video category, the evaluation factor is determined to be a correlation coefficient between the first correlation parameter and the second correlation parameter during the correction analysis period.

[0092] Specifically, see Figure 3 As shown in FIG. , which is a diagram showing the steps of determining the relevance between the first format information and the second format information according to an embodiment of the present invention, the process of determining the relevance between the first format information and the second format information includes:

[0093] Step S301, screening the correction and analysis period corresponding to the maximum value coefficient of the content in the homework explanation video stream;

[0094] Step S302, obtaining voice information within the correction and analysis period and determining voice keywords based on semantic analysis technology, and determining the voice keywords as the first format information;

[0095] Step S303: obtaining images corresponding to the explanation and correction feature frames in the correction and analysis period based on the time sequence relationship, determining keyword texts of newly added texts in the images corresponding to the explanation and correction feature frames in the preceding and subsequent time sequences using text recognition technology, and determining the keyword texts as the second format information;

[0096] Step S304: Determine the Euclidean distance between the first format information and the second format information as the correlation.

[0097] Specifically, in order to obtain voice information during the correction and analysis period and determine voice keywords based on semantic analysis technology, audio acquisition equipment, such as high-quality microphones, can be deployed in appropriate locations in the classroom to ensure clear capture of the teacher's voice during the correction and analysis period. With the help of existing mature speech recognition engines, the stored voice audio files of the correction and analysis period are input into the speech recognition engine. The engine uses acoustic models and language models to convert voice signals into text. Using semantic analysis technology tools in natural language processing technology, by calculating the PageRank values ​​of node words, words with higher PageRank values ​​are screened out as voice keywords. These keywords can accurately reflect the core content of the teacher's explanation during the correction and analysis period.

[0098] Specifically, text recognition technology and tools can be used to input the corresponding images of each saved explanation and correction feature frame into the text recognition tool in sequence. The tool converts the text in the image into text form through steps such as image preprocessing, character segmentation and character recognition. According to the time sequence relationship, the texts corresponding to the adjacent explanation and correction feature frames are compared, and a string matching algorithm is used to calculate the difference between the two texts. By calculating the similarity or difference of the texts, the newly added text content in the latter explanation and correction feature frame image is found, and natural language processing is performed on the determined newly added text. Keyword text that can accurately summarize its core content is extracted from the newly added text. This is an existing technology and will not be repeated here.

[0099] In implementation, the first format information and the second format information are vectorized respectively, and the two transformed vectors are substituted into the Euclidean distance formula to calculate the Euclidean distance between the first format information and the second format information. The Euclidean distance formula is widely used in semantic recognition analysis and will not be repeated here.

[0100] For example, the voice information of the teacher explaining the knowledge of "quadratic function" is obtained through a recording device. Assuming that the voice content includes the following keywords "quadratic function", "image opening direction" and "vertex coordinate formula", the word embedding model is used to map these keywords to a vector space of fixed dimension, and these keyword vectors are weighted averaged to obtain a comprehensive vector of the first format information, and the new text content is extracted from the correction question image through text recognition technology. Assuming that the correction question adds the following keywords "quadratic function expression", "symmetry axis position" and "function monotonicity" when explaining "quadratic function", the word embedding model is also used to map these keywords to three-dimensional vectors, and these keyword vectors are weighted averaged to obtain a comprehensive vector of the second format information, and the Euclidean distance formula is used to calculate the distance between the two vectors. The vectorization of text information is an existing technology and will not be described here.

[0101] It is understandable that semantic analysis technology can understand natural language, and by processing voice information during the correction and parsing period, it can convert the continuous voice stream into text, and extract keywords reflecting the core content from it to form the first format information, which represents the focus of the teacher's oral explanation; for the explanation and correction feature frame images, text recognition technology can detect and recognize the text content in the image, obtain the previous and next explanation and correction feature frame images based on the time sequence relationship, and extract the keyword text of the newly added text as the second format information, which reflects the focus of updating written content in the teaching process.

[0102] In this paper, the first and second format information are treated as two vectors, and their correlation is quantified by calculating their Euclidean distance. From a teaching perspective, if the teacher's explanation is closely related to the written content, then the corresponding voice keywords and the newly added text keywords will be semantically and logically close, which will be reflected in a smaller Euclidean distance value, indicating a high correlation. Conversely, if the two are significantly different, the Euclidean distance value will be large, indicating a low correlation. This quantitative method provides an intuitive way to judge whether the allocation of teaching resources is reasonable.

[0103] Specifically, the present invention targets the correction tendency video category, selects the feature correction analysis period corresponding to the maximum value coefficient of the content as the key to achieving the teaching goal, and by focusing on this core period, avoids wasting computing power due to paying too much attention to non-critical content. The voice keywords determined by semantic analysis technology during the correction analysis period are used as the first format information, and the keyword text of the newly added text in the explanation and correction feature frame image obtained based on text recognition technology is used as the second format information. Key information is extracted from two important dimensions: the teacher's explanation voice and the newly added text of the correction content update. The voice keywords reflect the key content conveyed orally by the teacher, while the keyword text of the newly added text reflects the key knowledge points presented in written form. This multi-dimensional method of extracting key information comprehensively and deeply covers the important content carried by different carriers in the teaching process, realizes the construction of a dual-dimensional evaluation system of analysis and explanation and analysis and annotation, and realizes the transformation of smart classrooms from experience-driven to data-driven decision-making mode.

[0104] Specifically, the process of determining whether the configuration of teaching resources is reasonable based on the relevance and generating resource supplementary information includes:

[0105] According to the comparison result where the correlation exceeds the preset correlation threshold, it is determined that the configuration of the teaching resources is unreasonable; according to the comparison result where the correlation does not exceed the preset correlation threshold, it is determined that the configuration of the teaching resources is reasonable;

[0106] In response to a result that the configuration is determined to be unreasonable according to the relevance, generating resource supplementary information sent to the student as first recommended supplementary information generated according to the first format information;

[0107] The Euclidean distance between the content keyword of the first recommended supplementary information and the first format information does not exceed a preset reference Euclidean distance.

[0108] In implementation, the correlation threshold and the reference Euclidean distance can be determined based on experimental data. The following is the Euclidean distance experimental data of 8 sets of experiments between the first format information and the second format information.

[0109]

[0110] In order to ensure that the setting of the correlation threshold can screen out the correction and analysis periods with a smaller Euclidean distance between the first format information and the second format information as periods with unreasonable configuration of teaching resources, the correlation threshold can be set to the average value of the Euclidean distance between the first format information and the second format information in the experimental data, that is, 0.64.

[0111] The preset reference Euclidean distance needs to ensure that the content keywords of the first recommended supplementary information are highly relevant to the first format information to ensure the validity of the supplementary content. Based on this, the value range of the reference Euclidean distance can be set to [0.4, 0.55]. Here, an optimal value of the reference Euclidean distance is 0.5.

[0112] Specifically, in daily teaching, teachers use multimedia devices to record the process of explaining homework, forming a video stream of homework explanations. The system extracts continuous frame images at preset intervals, identifies the characteristic frames of switching between correction questions, divides the video into multiple correction analysis periods, and determines the content value coefficient by calculating the duration of each period. If the content value coefficient of a certain period is significantly higher than that of other periods and the standard deviation exceeds the preset value, it is determined to be a correction-oriented video category. The system focuses on this period and extracts voice keywords as the first format information. At the same time, it identifies the newly added text keywords on the blackboard as the second format information and calculates the Euclidean distance between the two. If the distance exceeds the threshold, it is determined that the allocation of teaching resources is unreasonable.

[0113] Specifically, after determining that the configuration of teaching resources is unreasonable, the present invention generates first recommended supplementary information based on the first format information, fully considering the students' learning needs for the key content explained by the teacher. The first format information is the key information obtained by semantic analysis technology of the teacher's teaching voice, representing the core knowledge conveyed orally by the teacher. The resource supplementary information generated based on this can accurately supplement the deficiencies that may exist in students' understanding of the teacher's analysis and explanation of the homework, and requires that the Euclidean distance between the content keywords of the first recommended supplementary information and the first format information does not exceed the preset reference Euclidean distance, which ensures the close correlation between the supplementary information and the teacher's explanation focus, and ensures the consistency of the content of the supplementary resources with the teaching focus, so that students can efficiently obtain valuable learning resources, form a dynamic and adaptive teaching resource ecosystem, and realize the transformation of smart classrooms from experience-driven to data-driven decision-making mode.

[0114] Specifically, see Figure 4 As shown in FIG, it is a step diagram of determining the correlation relationship between the first correlation parameter and the second correlation parameter according to an embodiment of the present invention. The process of determining the correlation relationship between the first correlation parameter and the second correlation parameter includes:

[0115] Step S311, determining the number of words converted from the speech information into text within each correction and analysis period based on speech recognition, and determining the number of words as the first correlation parameter;

[0116] Step S312: obtaining images corresponding to the explanation and correction feature frames of each correction and analysis period based on the time sequence relationship, determining the number of words in the newly added text in the images corresponding to the explanation and correction feature frames of the preceding and subsequent time sequences using text recognition technology, and determining the number of words in the text as the second correlation parameter;

[0117] Step S313 , sorting the correction and analysis time periods in ascending order of the first correlation parameters, and combining the second correlation parameters of each correction and analysis time period into a correlation relationship sequence according to the sorting.

[0118] It can be understood that the voice information of the teacher during each correction and analysis period is converted into text based on speech recognition technology. This process uses acoustic models and language models. The acoustic model extracts features and recognizes patterns on the speech signal and converts it into a corresponding phoneme sequence. The language model further converts the phoneme sequence into understandable text based on grammatical and semantic rules. The first associated parameter is determined by counting the number of words in these converted texts. The principle is that the amount of content explained by the teacher can be quantified to a certain extent by the number of words in the text converted from speech. The more analysis and explanation, the more words in the converted text. Therefore, this parameter can intuitively reflect the amount of information the teacher verbally conveys during each correction and analysis period. The text recognition technology in image recognition is used to detect, segment and recognize the text in the image, and convert the text information in the image into text form. The principle is that the update of the correction questions during the teaching process often represents the presentation of new content. The number of newly added text words reflects the change in the written teaching content. This method can provide another quantitative dimension for analyzing the teaching process from the perspective of written information update. Corresponding to the first associated parameter, they together constitute a comprehensive quantitative description of the information output of the teaching process.

[0119] Specifically, the process of determining whether the configuration of teaching resources is reasonable and generating resource supplementary information based on the association relationship includes:

[0120] If the association relationship sequence is not an increasing sequence, it is determined that the configuration of the teaching resources is unreasonable; if the association relationship sequence is an increasing sequence, it is determined that the configuration of the teaching resources is reasonable;

[0121] In response to a result that the configuration is unreasonable according to the association relationship, the resource supplementary information sent to the student is the second recommended supplementary information generated according to the content of the heterosexual association correction and parsing period;

[0122] According to the comparison result that the difference between the first associated parameter and the second associated parameter in the correction analysis period exceeds the difference comparison value, the correction analysis period is determined as a heterosexual correlation correction analysis period, and the difference comparison value is the average value of the difference between the first associated parameter and the second associated parameter in several correction analysis periods;

[0123] Wherein, the Euclidean distance between the content keyword of the second recommended supplementary information and the content of the opposite-sex association correction and analysis period does not exceed a preset reference Euclidean distance.

[0124] In implementation, in order to ensure that the content keywords of the second recommended supplementary information are highly relevant to the content of the opposite-sex association correction and analysis period within the preset reference Euclidean distance, the value range of the reference Euclidean distance can be set to [0.4, 0.55]. Here, an optimal value of 0.5 for the reference Euclidean distance is provided.

[0125] It is understandable that reasonable allocation of teaching resources should ensure the synergy between teachers' explanations and written content updates. When the correlation series shows an increasing trend, it reflects that the output of oral and written information is synchronized during the teaching process, and the allocation of teaching resources is relatively reasonable. If the correlation series is not an increasing series, it indicates that there are problems in the coordinated allocation of oral and written information in teaching resources.

[0126] In the present invention, after determining the time period for heterosexual correlation correction and analysis, the second recommended supplementary information is generated based on these time periods. The principle is that there is an unreasonable allocation of teaching resources in these time periods, and the deficiencies in the teaching process are compensated by supplementing relevant learning resources. The Euclidean distance between the content keywords of the second recommended supplementary information and the content of the heterosexual correlation correction and analysis time period meets the preset reference Euclidean distance. The Euclidean distance is used to measure the similarity between the supplementary information keywords and the teaching content of the abnormal time period in the feature space. When the Euclidean distance meets the preset standard, it indicates that the supplementary information can address the problems of student understanding difficulties caused by the unreasonable allocation of teaching resources in this time period, and provide targeted and relevant learning materials to help students better understand and master the teaching content of this time period, thereby optimizing the teaching effect.

[0127] Specifically, for correcting non-tendency video categories, the present invention uses voice recognition and text recognition technology to obtain relevant data when determining the association relationship, and constructs an association relationship series after sorting the first association parameter. It can intuitively show the relationship between the teacher's explanation amount and the text update amount of homework correction, and prevent the teaching effect from being affected by configuration imbalance. In terms of generating resource supplementary information, the second recommended supplementary information is generated according to the content of the heterogeneous association correction analysis period, and the time period with abnormal information output during the teaching process is accurately located. These time periods are determined by calculating the average value of the difference in association parameters, which can effectively make up for the weak links in teaching. At the same time, the Euclidean distance between the keywords of the supplementary information content and the content of the heterogeneous association correction analysis period is required to meet the preset standard, which ensures that the supplementary information is closely related to the teaching content of the problem period, and realizes the transformation of the smart classroom from experience-driven to data-driven decision-making mode.

[0128] See also Figure 5 As shown in FIG, which is a system block diagram of a teaching quality optimization system based on AI intelligent correction and analysis according to an embodiment of the present invention, the present invention also provides a teaching quality optimization system based on AI intelligent correction and analysis, including:

[0129] An information acquisition unit includes a video recorder for acquiring a video stream of homework explanations and a microphone for acquiring audio information. The video recorder is also coupled to a video processor for dividing the video stream of homework explanations into a number of correction and analysis periods.

[0130] Specifically, the present invention does not limit the specific structure of the video recorder. It can be a camera for obtaining the video of the teacher correcting homework, which will not be described in detail here.

[0131] Specifically, the present invention does not limit the specific structure of the microphone. It only needs to realize the function of obtaining the teacher's voice information and converting it into a voice file, which will not be elaborated here.

[0132] a feature extraction unit connected to the information acquisition unit, configured to determine the content value coefficient of each correction analysis period and to determine the correction content tendency category of the homework explanation video stream based on the difference in the content value coefficient;

[0133] Specifically, the present invention does not limit the feature extraction unit. Preferably, the feature extraction unit can be a microcomputer processor, which will not be described in detail here.

[0134] an intelligent evaluation unit, connected to the information acquisition unit and the feature extraction unit, respectively, for determining evaluation factors of the homework explanation video stream based on the correction content tendency category;

[0135] Specifically, the intelligent evaluation unit in the present invention can be a data storage and a processor connected to the data storage. The processor calls the data in the data storage according to the execution instruction, which will not be described in detail here.

[0136] a content recommendation generating unit, connected to the intelligent evaluation unit, for determining whether the configuration of teaching resources is reasonable based on the evaluation factors, and generating resource supplementary information recommended to students;

[0137] Specifically, the content recommendation generation unit in the present invention can be an information processor connected to the network, which is used to connect to the network to screen qualified online teaching resources or local teaching resources according to the executed instructions, and will not be described in detail here.

[0138] a transmission unit connected to the content recommendation generation unit and configured to send the generated resource supplement information to the student terminal;

[0139] Specifically, the present invention does not limit the transmission unit. Preferably, it can be a data transmitter for sending resource supplementary information to the terminal, which will not be described in detail here.

[0140] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0141] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A teaching quality optimization method based on AI intelligent correction and analysis, characterized in that: include: Obtaining a video stream of a teacher explaining homework during the teaching process, determining a number of explanation and correction feature frames in the homework explanation video stream, and dividing the homework explanation video stream into a number of correction analysis periods based on the explanation and correction feature frames; Determining the content value coefficient of each correction and analysis period according to the duration of the correction and analysis period, and determining the correction content tendency category of the homework explanation video stream according to the difference in the content value coefficient; Determining evaluation factors for the homework explanation video stream based on the correction content tendency category, the evaluation factors including the correlation between the first format information and the second format information within the characteristic correction analysis period, and the correlation between the first associated parameter and the second associated parameter within the correction analysis period; Wherein, the feature correction and analysis period is determined based on the content value coefficient; Determining whether the configuration of teaching resources is reasonable based on the evaluation factors, and generating resource supplementary information recommended to students in response to a determination result that the configuration of teaching resources is unreasonable, wherein the resource supplementary information is first recommended supplementary information generated based on the first format information, or second recommended supplementary information generated based on the content of the heterogeneous association correction and analysis period; The heterosexual association correction and analysis period is determined based on the difference between the first association parameter and the second association parameter; The generated resource supplement information is sent to the student terminal.

2. The teaching quality optimization method based on AI intelligent correction and analysis according to claim 1 is characterized in that: The method of determining several explanation and correction feature frames is to obtain images of consecutive frames at preset intervals in the homework explanation video stream, and determine the time when the correction questions in the image are switched as the time when the explanation and correction feature frames are located.

3. The teaching quality optimization method based on AI intelligent correction and analysis according to claim 2 is characterized in that: The process of determining the content value coefficient includes: Determining the period between two adjacent explanation and correction feature frames as the correction and parsing period; The duration of each correction and analysis period is determined, and the ratio of the duration to the total duration of the homework explanation video stream is determined as the content value coefficient.

4. The teaching quality optimization method based on AI intelligent correction and analysis according to claim 3 is characterized in that: The process of determining the correction content tendency category of the homework explanation video stream includes: Calculate the standard deviation of the content value coefficient for each revision analysis period; According to the comparison result that the standard deviation satisfies the video category distinction condition, determining that the correction content tendency category of the homework explanation video stream is a correction tendency video category; According to the comparison result that the standard deviation does not meet the video category distinction condition, determining that the correction content tendency category of the homework explanation video stream is a correction non-tendency video category; The video category distinction condition is that the standard deviation exceeds a preset standard deviation comparison value.

5. The teaching quality optimization method based on AI intelligent correction and analysis according to claim 4 is characterized in that: The evaluation factors of the homework explanation video stream are determined as follows: If the homework explanation video stream is a correction tendency video category, determining the evaluation factor as the correlation between the first format information and the second format information within the characteristic correction analysis period; If the homework explanation video stream is a non-correction-oriented video category, the evaluation factor is determined to be a correlation coefficient between the first correlation parameter and the second correlation parameter during the correction analysis period.

6. The teaching quality optimization method based on AI intelligent correction and analysis according to claim 5 is characterized in that: The process of determining the relevance between the first format information and the second format information includes: Filter the correction and analysis period corresponding to the maximum value coefficient of the content in the homework explanation video stream; Acquire the voice information within the correction and analysis period and determine voice keywords based on semantic analysis technology, and determine the voice keywords as the first format information; Acquire images corresponding to the explanation and correction feature frames in the correction and analysis period based on a time sequence relationship, determine keywords of newly added text in images corresponding to the explanation and correction feature frames in the preceding and following time sequences using text recognition technology, and determine the keywords as the second format information; The Euclidean distance between the first format information and the second format information is determined as the correlation.

7. The teaching quality optimization method based on AI intelligent correction and analysis according to claim 6 is characterized in that: The process of determining whether the configuration of teaching resources is reasonable according to the relevance and generating resource supplementary information includes: According to the comparison result that the correlation exceeds a preset correlation threshold, determining that the configuration of the teaching resources is unreasonable; In response to a result that the configuration is determined to be unreasonable according to the relevance, generating resource supplementary information sent to the student as first recommended supplementary information generated according to the first format information; The Euclidean distance between the content keyword of the first recommended supplementary information and the first format information does not exceed a preset reference Euclidean distance.

8. The teaching quality optimization method based on AI intelligent correction and analysis according to claim 5 is characterized in that: The process of determining the correlation relationship between the first correlation parameter and the second correlation parameter includes: Determining the number of words converted from the speech information into text within each correction and analysis period based on speech recognition, and determining the number of words as the first associated parameter; Acquire images corresponding to explanation and correction feature frames of each correction and analysis period based on a time sequence relationship, determine the number of words in the newly added text in the images corresponding to the explanation and correction feature frames of the preceding and subsequent time sequences using a text recognition technique, and determine the number of words in the text as the second associated parameter; The correction and analysis time periods are sorted in order from small to large according to the first correlation parameters, and the second correlation parameters of each correction and analysis time period are sequentially combined into a correlation relationship sequence according to the sorting.

9. The teaching quality optimization method based on AI intelligent correction and analysis according to claim 8 is characterized in that: The process of determining whether the configuration of teaching resources is reasonable based on the association relationship and generating resource supplementary information includes: If the association relationship sequence is not an increasing sequence, it is determined that the configuration of teaching resources is unreasonable; In response to a result that the configuration is unreasonable according to the association relationship, the resource supplementary information sent to the student is the second recommended supplementary information generated according to the content of the heterosexual association correction and parsing period; According to the comparison result that the difference between the first associated parameter and the second associated parameter in the correction analysis period exceeds the difference comparison value, the correction analysis period is determined as a heterosexual correlation correction analysis period, and the difference comparison value is the average value of the difference between the first associated parameter and the second associated parameter in several correction analysis periods; Wherein, the Euclidean distance between the content keyword of the second recommended supplementary information and the content of the opposite-sex association correction and analysis period does not exceed a preset reference Euclidean distance.

10. A teaching quality optimization system based on AI intelligent correction and analysis, used to implement the teaching quality optimization method based on AI intelligent correction and analysis according to any one of claims 1 to 9, characterized in that: include: An information acquisition unit includes a video recorder for acquiring a video stream of homework explanations and a microphone for acquiring audio information. The video recorder is also coupled to a video processor for dividing the video stream of homework explanations into a number of correction and analysis periods. a feature extraction unit connected to the information acquisition unit, configured to determine the content value coefficient of each correction analysis period and to determine the correction content tendency category of the homework explanation video stream based on the difference in the content value coefficient; an intelligent evaluation unit, connected to the information acquisition unit and the feature extraction unit, respectively, for determining evaluation factors of the homework explanation video stream based on the correction content tendency category; a content recommendation generating unit, connected to the intelligent evaluation unit, for determining whether the configuration of teaching resources is reasonable based on the evaluation factors, and generating resource supplementary information recommended to students; A transmission unit is connected to the content recommendation generation unit and is used to send the generated resource supplement information to the student terminal.

Images