Smart classroom management method and device based on big data analysis

Abstract

The invention discloses a smart classroom management method and device based on big data analysis, and the method comprises the steps: collecting corresponding data, and generating corresponding sample data; based on the first dimension, performing clustering analysis on the specified sample through big data analysis to obtain a plurality of class clusters; for each class cluster, data distribution conditions corresponding to a third dimension and a fourth dimension corresponding to each specified sample are determined through big data analysis, and a first classroom management scheme corresponding to the third dimension and a second classroom management scheme corresponding to the fourth dimension are generated; and recommending a first classroom management scheme to the teacher before the current online teaching, and recommending a second classroom management scheme to the teacher in the current online teaching. Different teaching environments are classified in advance, and based on the current teaching environment, the teaching environment classification most conforming to the current teacher is determined. Based on big data analysis, the use conditions of other teachers for teaching materials and teaching equipment in a similar environment are obtained, so that recommendation is performed to the teacher, and the teacher can quickly adapt to the current teaching environment.

Description

Technical Field

[0001] The present application relates to the field of smart teaching, and specifically to a smart classroom management method and equipment based on big data analysis. Background Art

[0002] With the development of technology, smart teaching has gradually come into people's vision. Smart teaching is an important part of supporting talent training in the construction of smart campuses. It mainly includes smart teaching environment, digital teaching resources, smart teaching management, etc.

[0003] The construction, management and maintenance of smart classrooms are important components of the construction of smart teaching environments. Smart classrooms aim to improve teaching effectiveness and learning experience through advanced equipment and systems. Smart classrooms not only rely on traditional teaching aids, but also use a variety of digital tools and technologies to support teaching and learning activities.

[0004] The traditional smart classroom management process usually relies on the access of various technical equipment and educational software. The degree of intelligence is often only concentrated on the integration of hardware and software, which makes the traditional smart classroom only able to provide basic functions such as local data collection, integration, and display. It does not draw on the teaching experience of others, making it difficult to achieve effective smart teaching.

[0005] For example, in some smart classrooms, teachers can use smart devices and online teaching tools in the classroom to conduct remote online teaching. However, due to the large differences between the online teaching environment and the actual teaching environment, and even between different online teaching environments, it may be difficult for teachers to quickly adapt to the new online teaching environment. Summary of the invention

[0006] In order to solve the above problems, this application proposes a smart classroom management method based on big data analysis, which is characterized by including:

[0007] For the online teaching process in the smart classroom, corresponding teaching data is collected to generate corresponding sample data, wherein the sample data includes multiple dimensions, the first dimension is used to represent the teaching environment, the second dimension is used to represent the teaching feedback, the third dimension is used to represent the preparation of teaching materials, and the fourth dimension is used to represent the use of teaching equipment during the teaching process;

[0008] Based on the second dimension, sample data whose teaching feedback meets expectations are selected from all sample data as designated samples;

[0009] Based on the first dimension, cluster analysis is performed on the designated samples through big data analysis to obtain multiple clusters, each cluster containing multiple designated samples;

[0010] For each cluster, determine the data distribution of the third dimension and the fourth dimension corresponding to each designated sample through big data analysis, and generate a first classroom management plan corresponding to the third dimension and a second classroom management plan corresponding to the fourth dimension according to the data distribution;

[0011] Determine the current teaching environment of this online teaching, determine the cluster to which this online teaching belongs in the cluster analysis based on the current teaching environment, and the first classroom management plan and the second classroom management plan corresponding to the cluster, and recommend the first classroom management plan to the teacher before the online teaching, and recommend the second classroom management plan to the teacher during the online teaching.

[0012] On the other hand, the present application also proposes a smart classroom management device based on big data analysis, including:

[0013] at least one processor; and,

[0014] a memory communicatively connected to the at least one processor; wherein,

[0015] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: implement a smart classroom management method based on big data analysis as described in any of the above examples.

[0016] The online intelligent teaching method proposed in this application can bring the following beneficial effects:

[0017] Different teaching environments are classified in advance, and based on the current teaching environment, the most suitable teaching environment classification for the current teacher is determined. Based on big data analysis, the use of teaching materials and teaching equipment by other teachers in similar environments is obtained, and recommendations are made to teachers so that they can quickly adapt to the current teaching environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0019] Figure 1 This is a flow chart of the online intelligent teaching method in the embodiment of the present application;

[0020] Figure 2 This is a schematic diagram of the interface of the online teaching platform client in the embodiment of the present application;

[0021] Figure 3 This is a schematic diagram of an online smart teaching device in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0023] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0024] like Figure 1 As shown, the embodiment of the present application provides a smart classroom management method based on big data analysis, including:

[0025] S101: For the online teaching process in the smart classroom, corresponding data is collected and corresponding sample data is generated, wherein the sample data includes multiple dimensions, a first dimension is used to represent the teaching environment, a second dimension is used to represent the teaching feedback, a third dimension is used to represent the preparation of teaching materials, and a fourth dimension is used to represent the use of the online teaching platform.

[0026] The collection of relevant data can be done by collecting historical records of smart classrooms in an area and generating relevant data. During the collection process, a sample data may not contain data of all dimensions, so the relevant data can be marked as empty. After the data is collected, it can be sorted and summarized accordingly to eliminate abnormal data.

[0027] Specifically, the sub-dimensions of the first dimension include at least: subject, grade, total teaching experience of teachers, online teaching experience of teachers, degree of intelligence of smart classrooms, etc. Among them, total teaching experience of teachers and online teaching experience of teachers can be counted in years, and the degree of intelligence of smart classrooms can be determined based on the number of smart devices deployed in the smart classroom where the online teaching process is located, the control software of the smart devices, etc.

[0028] The sub-dimensions of the second dimension include at least: questionnaire feedback data and teaching achievement feedback data. The questionnaire feedback data can be obtained through paper questionnaires or electronic questionnaires, interviews with students, auditors, etc., while the teaching achievement feedback can be the accuracy rate of homework or the accuracy rate of in-class tests.

[0029] The sub-dimensions of the third dimension include at least: the preparation of smart devices and the preparation of teaching courseware. Among them, smart devices may include electronic whiteboards, projectors, experimental instruments, etc., and teaching courseware may include PPT files, video files, etc. The preparation mainly indicates whether the smart device or teaching courseware needs to be used.

[0030] like Figure 2 As shown, the online teaching platform includes: a main window module, a sub-window module, a teaching tool module, and an auxiliary explanation module. The teaching tool module includes multiple teaching tools for teachers to use.

[0031] Classrooms can log in to the online teaching platform through the teacher-side APP. The client interface contains multiple virtual modules, each with corresponding functions. Teachers can configure the size, position, and internal functions of each virtual module based on their needs.

[0032] The main window module is used to display the teacher's live content or the display content corresponding to the teaching tools, and the secondary window module is used to display the student status (which may include the student's name, whether he is online, or the student's live video, etc.). Generally speaking, students cannot see the secondary window module when they are doing online teaching.

[0033] The teaching tool module includes multiple teaching tools for teachers to use. Teachers can match them by themselves or use the default settings. The auxiliary explanation module automatically assists in explaining the knowledge points explained by the teacher to facilitate students' understanding.

[0034] At this time, the sub-dimensions of the fourth dimension at least include: the usage of each teaching tool in the online teaching platform. The usage of the teaching tool includes whether the teaching tool is used in this online live broadcast, and of course, it can also include the usage time, etc.

[0035] S102: Based on the second dimension, sample data whose teaching feedback meets expectations are selected from all sample data as designated samples.

[0036] Teaching feedback that meets expectations usually refers to some sample data with good teaching feedback. For example, the overall score of questionnaire feedback data is "high", or the sample data with a high accuracy rate in teaching achievement feedback. When there is no teaching feedback in the sample data, that is, when there is no second dimension, if the specified sample size is too small, sample data that at least partially does not contain the second dimension is selected as the specified sample. If the specified sample size is sufficient, the sample data that does not contain the second dimension is discarded.

[0037] Through screening, we mainly select sample data with better teaching feedback as guidance and recommendation data for subsequent teachers.

[0038] S103: Based on the first dimension, cluster analysis is performed on the designated samples through big data analysis to obtain a plurality of clusters, each of which contains a plurality of designated samples.

[0039] The first dimension contains multiple sub-dimensions, and based on the specific values ​​of each sub-dimension, a multidimensional vector corresponding to the first dimension is generated. Among them, continuous values ​​such as grade, teaching experience, and degree of intelligence can be obtained through the corresponding grade, duration, and score, while for non-continuous values ​​such as subjects, corresponding values ​​can be assigned to each subject in advance to represent them.

[0040] Specifically, for the intelligence level of the smart classroom, the configuration status of the smart devices that have been set up in the smart classroom and the concentration level of the control software of the smart devices can be obtained. Among them, the configuration status refers to whether the corresponding smart devices are configured, and the concentration level refers to whether all smart devices are controlled by several control software, and the degree of intelligence is preliminarily judged from the software and hardware. These two can judge the degree of intelligence in a relatively general and basic way at the hardware and software levels. Of course, the judgment level can also be added based on demand, such as adding the diversity of functional modules of the control software, the number of occurrences of smart device anomalies, etc.

[0041] Different smart devices are given different scores. For example, higher scores are given to devices with more obvious intelligence, such as electronic whiteboards and smart speakers, and lower scores are given to devices with less intelligence, such as printers and projectors. At this time, the first score corresponding to the configuration state is obtained by summing the scores corresponding to each of the smart devices that have been set up. The second score corresponding to the degree of concentration is obtained based on the number of control software for each smart device that has been set up. Among them, the lower the number of control software, the higher the degree of concentration, and the higher the second score.

[0042] Based on the sum of the first score and the second score, the total score is used as the intelligence level of the smart classroom. To describe the intelligence level of a smart classroom, the mean and median of the intelligence levels of smart classrooms of a large number of specified samples can be used to determine it.

[0043] At this time, multiple intervals are divided based on the degree of intelligence, and the degree of intelligence in each interval is different, and weights are assigned to each sub-dimension of the first dimension in each interval. Among them, in intervals with higher degrees of intelligence, the requirements for teachers' online teaching experience are higher, so the weight is higher. At the same time, because the difference between classrooms with higher degrees of intelligence and other classrooms is greater, the requirements for teachers' control ability are also higher, so the weight of the intelligence degree of the smart classroom itself is higher. Correspondingly, the weights of other sub-dimensions can be reduced.

[0044] At this time, for each interval, based on the weights of each dimension of the first dimension, clustering is performed using the K-means clustering algorithm to obtain multiple clusters. Of course, in addition to the K-means clustering algorithm, clustering algorithms such as DBSCAN and CFSFDP can also be used. When calculating the current teaching environment later, you can also first determine the interval corresponding to its degree of intelligence, and then determine the cluster to which it belongs.

[0045] S104: For each cluster, determine the data distribution of the third dimension and the fourth dimension corresponding to each specified sample through big data analysis, and generate a first classroom management plan corresponding to the third dimension and a second classroom management plan corresponding to the fourth dimension according to the data distribution.

[0046] Data distribution refers to the distribution of data in each sub-dimension in the third and fourth dimensions. Among them, for the third dimension, the preparation between different smart devices and teaching courseware may have a strong correlation, that is, the correlation between the sub-dimensions is strong, so when considering the third dimension, each data of the third dimension includes the specific values ​​of all the sub-dimensions it contains. For example, a certain data of the third dimension can be: using smart device A, using smart device B, using smart device C, using teaching courseware D. The fourth dimension may involve some specific usage situations, and the correlation between them is not strong, so each data of the fourth dimension only selects the specific value of any sub-dimension it contains. For example, a certain data of the fourth dimension can be: using teaching tool E and using teaching tool F in the online teaching platform.

[0047] Based on this, when determining the data distribution, for each cluster, the third dimension and the fourth dimension corresponding to each specified sample are determined. At this time, for each data item in the third dimension and the fourth dimension, the corresponding frequency is determined. The composition of each data item in the third dimension and the fourth dimension has been described above.

[0048] Generate cross tables corresponding to the third and fourth dimensions based on the frequency, and determine whether there is a correlation between the data in the third and fourth dimensions based on the chi-square test.

[0049] An example of a cross table can be shown in Table 1 below, where the values ​​are specific sample numbers:

[0050]

[0051] Table 1

[0052] Among them, the chi-square test is a hypothesis testing method, which can statistically calculate the correlation between various data. Taking Table 1 as an example, it can be calculated that the correlation between "using teaching tool F" and "using equipment A, C and courseware E" is high.

[0053] Continue to sort based on the frequency of each data in the third dimension to obtain the distribution of the first data corresponding to the third dimension. Generally speaking, sort from high to low according to the frequency.

[0054] Sorting is performed based on the frequency of each item of data in the fourth dimension, and the sorting is also performed from high to low. Of course, if there is a single item of data in the fourth dimension, there is a correlation between it and the single item of data in the third dimension, and the ranking of the single item of data in the third dimension in the first data distribution is located before the first preset order (based on actual demand setting), for example, "using F teaching tool" in Table 1 is associated with "using equipment A, C and courseware E", and "using equipment A, C and courseware E" is located before the first preset order, which means that even if the single item of data in the fourth dimension is rarely used, it should be ranked higher because the single item of data in the third dimension associated with it is easy to use, and the ranking of the single item of data in the fourth dimension should be raised to the second preset order (based on actual demand setting, usually due to the complexity of the teaching tools, the number before the second preset order is more than the first preset order) to obtain the second data distribution corresponding to the fourth dimension.

[0055] At this time, according to the distribution of the first data, several single data items sorted before the first preset order are selected as the first classroom management plan; and according to the distribution of the second data, several single data items sorted before the second preset order are selected as the second classroom management plan.

[0056] That is to say, the first classroom management plan includes the most commonly used teaching materials in a single cluster, while the second classroom management plan includes the most commonly used teaching tools of the online teaching platform in a single cluster, as well as the teaching tools that are associated with the first classroom management plan.

[0057] S105: Determine the current teaching environment of this online teaching, determine the cluster to which this online teaching belongs in the cluster analysis based on the current teaching environment, and the first classroom management plan and the second classroom management plan corresponding to the cluster, and recommend the first classroom management plan to the teacher before this online teaching, and recommend the second classroom management plan to the teacher during this online teaching.

[0058] When the teacher conducts this online teaching, the corresponding current teaching environment is first collected and the cluster to which it belongs is determined. The teaching environments of other sample data in the cluster are similar to the current teaching environment, and the teaching feedback is good. Therefore, the more common first classroom management plan and second classroom management plan are selected, and the first classroom management plan is recommended to the teacher before this online teaching, so that the teacher can prepare the corresponding teaching materials in advance (previous teaching materials can be stored and downloaded through the cloud, of course, teachers can also choose their own teaching materials). In the teaching process, commonly used teaching tools can also be recommended to teachers to help them get started quickly.

[0059] Different teaching environments are classified in advance, and based on the current teaching environment, the most suitable teaching environment classification for the current teacher is determined. Based on big data analysis, the use of teaching materials and teaching equipment by other teachers in similar environments is obtained, and recommendations are made to teachers so that they can quickly adapt to the current teaching environment.

[0060] In one embodiment, when the teacher performs online teaching and triggers the teaching tool module, the teaching tool included in the second classroom management plan is recommended to the teacher in the teaching tool module so that the teacher can quickly get started with the teaching tool. Of course, in addition to whether to use the teaching tool, the second classroom management plan can also include an introduction to the use of the teaching tool.

[0061] At this time, in order to facilitate teachers to further use teaching tools and improve the quality of teaching through teaching tools, the usage status of teaching tools can be obtained in real time, and when the usage status hits the preset target state, data collection is performed on the main window module to obtain teaching data.

[0062] Then, the teaching data is analyzed to obtain the knowledge points currently being taught, and corresponding auxiliary explanation information is added to the auxiliary explanation module according to the knowledge points, so that the auxiliary explanation information is displayed to students, so that students can understand the corresponding knowledge points more vividly.

[0063] Specifically, teaching tools can be divided into several categories according to their functions, including: sharing (e.g., file sharing), interaction (e.g., online voting, interactive games), auxiliary (e.g., electronic paintbrushes, electronic whiteboards, virtual lab benches) and management (e.g., classroom management).

[0064] Among them, file sharing means that teachers can share PPT, videos, web pages and other files on their own screens so that students can see the corresponding files. In online voting, teachers can edit the questions and answers of the vote to let students choose. Interactive games cover a wide range, allowing students to perform corresponding actions according to instructions. Electronic brushes can annotate shared files on the basis of file sharing. In the electronic whiteboard, teachers can write and draw in real time and share the screen with students through the main window module. After starting the virtual laboratory table (for example, PhET), the teacher jumps to the virtual laboratory table and shares the screen with students. Classroom management refers to some management content of the class by the teacher, such as modifying the class title, starting student sign-in, and setting up class access.

[0065] Since sharing-type auxiliary tools are used frequently and in a wide range, covering a lot of content, it is difficult to extract knowledge points. Moreover, management-type auxiliary tools are mainly used for the management of students and do not involve the explanation of knowledge points.

[0066] Therefore, in actual use, only the usage status of the specified type of teaching tools is obtained in real time; the specified types include: interactive type, auxiliary type; the interactive type includes: online voting, interactive games, and the auxiliary type includes: electronic brush, electronic whiteboard, virtual laboratory table.

[0067] When the usage status of the teaching tool is started, its corresponding usage status is obtained in real time according to the work log of the teaching tool, and a preset target status is set for each specified type of teaching tool. When the usage status hits the target status, the collection of teaching data begins.

[0068] Furthermore, for online voting, after the voting results are generated, a screenshot of the main window module is taken, and OCR recognition is performed based on the screenshot to obtain the first text content. At this time, if the main window module is a live broadcast screen of the teacher, a screenshot is not required to obtain the first text content. If it is an interface for sharing files, a screenshot is taken to obtain the first text content. At the same time, the second text content included in the online voting is determined. The second text content may include: the text content of the online voting question and the text content of each option. The first text content and the second text content are used as teaching data.

[0069] For interactive games, when the interactive game is started, the voice data generated by the teacher during the execution of the interactive game is collected, and the voice data is subjected to voice recognition to obtain the third text content. At the same time, the fourth text content corresponding to the text description of the interactive game is determined, and the third text content and the fourth text content are used as teaching data. The third text content reflects what the teacher said during the interactive game, and the interactive game is usually created based on knowledge points, so the discourse usually contains the content of knowledge points. Similarly, the fourth text content may also contain the content of knowledge points.

[0070] For the electronic whiteboard, when the erase function or close function of the electronic whiteboard is activated and has not been executed, that is, the teacher has already written a large section of content on the electronic whiteboard and is about to erase it and write new content, or has finished organizing the knowledge points and is about to close the electronic whiteboard, at this time, the teacher has already clicked the erase function or close function button, first takes a screenshot of the electronic whiteboard in the main window module, and then executes the function corresponding to the button, which plays the role of saving handwriting through screenshots, and performs OCR recognition based on the screenshots to obtain the fifth text content as teaching data.

[0071] For the electronic brush, unlike the initial blank state of the electronic whiteboard, the electronic brush usually annotates the existing file content (for example, the file that has been shared in the main screen window). Therefore, when the use time of the electronic brush exceeds the preset time (for example, 2 minutes), it can be considered that the teacher has annotated a lot of content, or annotated more important content, rather than casually annotating unimportant knowledge points. At this time, a screenshot is taken of the main window module containing the electronic brush drawing traces, and OCR recognition is performed based on the screenshot to obtain the sixth text content as teaching data. In the sixth text content, in addition to the content of the file shared, it also includes the content written by the electronic brush annotation. Or, when the drawing pattern of the electronic brush meets the preset pattern requirements (for example, a single horizontal line, double horizontal lines, wavy lines, circles, etc.), it means that it has annotated some important parts of the file. At this time, a screenshot is taken of the main window module, and the content contained in the drawing pattern (for the drawing pattern is a circle) and the upper preset range (for the drawing pattern is a single horizontal line, double horizontal lines, wavy lines) in the screenshot is OCR recognized to obtain the seventh text content as teaching data. At this time, the seventh text content mainly includes the part that the teacher has focused on drawing.

[0072] For virtual lab benches, since they come in various forms and are difficult to quantify, the voice data generated by teachers during the execution of the virtual lab benches is collected, and speech recognition is performed on the voice data to obtain the eighth text content as teaching data, which usually includes explanations and repetitions of key knowledge points.

[0073] At this point, the collection of teaching data has been completed, and the teaching data can be analyzed to obtain the knowledge points currently being taught.

[0074] Specifically, for different teaching tools, teaching data has been collected based on their usage status and converted into corresponding text content. At this time, the knowledge points currently being taught are extracted based on the text content, that is, the knowledge points when using the teaching tool.

[0075] It should be noted that teachers will also explain some knowledge points when they are not using teaching tools. However, if knowledge points are extracted throughout the entire teaching process, it will not only require more computing power, but also cause more noise. When teachers use teaching tools, it is often because they need to explain certain knowledge points more vividly. Therefore, when the use state of the teaching tool hits the target state, there is a greater probability that knowledge points will be extracted.

[0076] Based on this, when extracting knowledge points, due to the different teaching tools, the sources of the extracted text content are also different. Therefore, the text content extracted above is first classified. The first text content, the second text content, the third text content, the fourth text content, and the eighth text content are directly collected texts, or the texts converted through voice data. Their texts have no format attributes, so they are used as the first data to be analyzed; and the fifth text content, the sixth text content, and the seventh text content are text contents recognized through screenshots. Each text content has format attributes in the screenshot, so they are used as the second data to be analyzed.

[0077] At this time, pattern recognition is performed on the second data to be analyzed to determine the structural position of the second data to be analyzed in the screenshot, wherein the structural position includes: title, text, chart, and notes. For example, if the screenshot belongs to a PPT file shared by a file, or a blackboard content hand-drawn by a teacher, it usually has corresponding structural attributes, so its structural attributes are identified here. The process of identifying structural attributes can be based on the thickness, size, length, position, etc. of the text. For example, the title is usually above the screenshot, and it is thicker, the font size is larger, and the length is shorter.

[0078] Based on the entity recognition algorithm, knowledge points are extracted from the first data to be analyzed and the second data to be analyzed. For the seventh text content, since it is annotated for the content in the shared file, this part is highly likely to be a knowledge point. If no knowledge point is extracted for the seventh text content, the preset range is expanded and OCR recognition is performed again to obtain the seventh text content and extract knowledge points again.

[0079] At least some of the knowledge points extracted are selected to add corresponding auxiliary explanation information in the auxiliary explanation module. The auxiliary explanation information can be a pre-defined explanation video or explanation courseware related to the knowledge point. Taking the explanation video as an example, it can be stored in the cloud and add corresponding knowledge point tags. When the knowledge point is selected, the corresponding explanation video is found in the cloud and added to the auxiliary explanation module. The explanation video can be played instantly based on the teacher's click.

[0080] Multiple knowledge points may be extracted in the same period. For example, the teacher uses multiple teaching tools in the same period and extracts knowledge points from all of them (for example, online voting or interactive games are used while using a virtual laboratory bench). Alternatively, multiple knowledge points are extracted during the use of a teaching tool (for example, multiple knowledge points are identified in the content circled according to a drawing pattern). At this time, adding corresponding auxiliary explanation information to all knowledge points may affect students' memory and learning. Therefore, it may be necessary to select some knowledge points, for example, only select one or two knowledge points at most in the same period to add corresponding auxiliary explanation information.

[0081] Among them, the selection priority of knowledge points is as follows: in the first data to be analyzed, it is believed that the content in the screenshots not only fits the knowledge points, but also often summarizes them, so the first text content has the highest priority. Secondly, it is believed that the teacher's voice data is also used to describe knowledge points, but it is not summarized and may contain noise, so the third text data and the eighth text content have lower priority than the first text content. Secondly, the text content of online voting and interactive games may involve knowledge points, but there is a certain probability that they will not be involved, so the second text content and the fourth text content have lower priority than the third text data and the eighth text content. Among them, the scope of interactive games is wider, and the association with knowledge points may be unstable, while online voting is the content input by the teacher, so it is believed that the second text content has a higher priority than the fourth text content. In the second data to be analyzed, the order of priority is: the title has a higher priority than the text, and the text has a higher priority than charts and notes.

[0082] Furthermore, when extracting knowledge points, the first data to be analyzed and the second data to be analyzed are cleaned to obtain cleaned data, wherein the data cleaning includes: removing special characters, removing punctuation marks, removing stop words, and performing word segmentation operations.

[0083] Through the pre-trained entity recognition model, the cleaned data is subjected to entity recognition to obtain the first entity; and the cleaned data is subjected to keyword recognition based on the word vector to obtain the first keyword. Among them, the entity recognition model is also called the NER model, which is mainly used to identify entities in text data, and the word vector can use the word2vec tool to identify keywords.

[0084] Based on the domain dictionary corresponding to the teaching subject, the first entity and the first keyword are filtered respectively to obtain the second entity and the second keyword. Different teaching customers have different domain dictionaries, and the entities and keywords are filtered through the dictionary to find only the entities and keywords that match the teaching subject.

[0085] Based on the semantic association, the second entity and the second keyword are combined to extract the knowledge point. The semantic association can use word vectors (for example, word2vec tool) to represent the semantic association between words, and by calculating the similarity between word vectors, semantically related words are determined and combined.

[0086] In addition, it is also mentioned above that speech recognition is performed on the speech data to extract the third text content and the eighth text content. However, in some cases, for example, if this online teaching is an experimental class or an interactive game class, the third text content and the eighth text content will have too much data volume, thereby generating a lot of noise and affecting the extraction of knowledge points.

[0087] Based on this, when the duration of voice data is longer than a preset duration (for example, 10 minutes), the voice data is cleaned and converted to obtain standard data. Among them, data cleaning can include checking data consistency, processing invalid values ​​and missing values, etc. Data conversion can standardize the data.

[0088] Feature extraction is performed on the standard data to obtain speech features, and the speech features are reduced in dimension. Speech features include fundamental frequency features, energy features, and spectrum features. Among them, the fundamental frequency feature, also known as the fundamental frequency, is the lowest frequency component in the sound signal and can be used to analyze and identify different tones or the voice characteristics of the speaker. The energy feature refers to the energy distribution in the sound signal. By measuring the energy level of the signal, the intensity information of the sound signal can be obtained. The spectrum feature refers to the distribution of the sound signal in the frequency domain. The spectrum feature can be obtained by Fourier transforming the signal, which provides the energy distribution of the sound signal at different frequencies and can be used to identify different audio effects.

[0089] Generally speaking, when a teacher is talking about a knowledge point, his or her voice characteristics may change accordingly. For example, when talking about a knowledge point, the teacher's tone will be heavier and the speaking speed will be relatively slower.

[0090] Based on this, the support vector machine obtained through supervised training is used to identify the speech features after feature dimensionality reduction to determine the corresponding speech tags. The speech tags include: speech rate tags and intonation tags, which are used to indicate the speech rate and emotion of the speech. In the support vector machine, the radial basis kernel function is used as the kernel function, the penalty parameter C is set to 1.0, the kernel function parameter gamma is set to scale, the convergence tolerance is set to 0.001, the kernel function cache size is set to 100MB, and the cross-validation fold is set to 10. In actual recognition, the speech data can be divided. For example, whenever the teacher's pause time (that is, the time without speaking) exceeds 3 seconds, a new interval is obtained by division, so as to identify the speech tags in each interval.

[0091] At this time, the teaching data is divided on the time axis based on the voice tags to obtain multiple time periods, wherein each time at least one of the voice tags changes, a new time period is obtained, that is, at least one label is different in every two adjacent time periods. Of course, the minimum duration of the time period can be set, and if a time period with too short duration appears, it is regarded as part of the previous time period, effectively reducing the excessive division of time periods and reducing errors.

[0092] For each time period, a corresponding weight is assigned to the text content of speech recognition within the time period based on the speech label corresponding to the time period; wherein, the slower the speech rate label and the stronger the intonation label, the higher the corresponding weight. The speech recognition results within the time period whose weight is lower than the preset threshold are deleted, that is, the part with particularly fast speech rate and / or particularly flat intonation is deleted, and it is considered that the probability of knowledge points existing in this part of the content is small, so that the remaining speech recognition results are used as the third text content or the eighth text content, thereby completing the voice screening.

[0093] like Figure 3 As shown, the embodiment of the present application also provides a smart classroom management device based on big data analysis, including:

[0094] at least one processor; and,

[0095] a memory communicatively connected to the at least one processor; wherein,

[0096] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: implement a smart classroom management method based on big data analysis as described in any of the above embodiments.

[0097] The embodiment of the present application also provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured as: a smart classroom management method based on big data analysis as described in any of the above embodiments.

[0098] The above is only the embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A smart classroom management method based on big data analysis, characterized in that: include: For the online teaching process in the smart classroom, corresponding data is collected to generate corresponding sample data, wherein the sample data includes multiple dimensions, the first dimension is used to represent the teaching environment, the second dimension is used to represent the teaching feedback, the third dimension is used to represent the preparation of teaching materials, and the fourth dimension is used to represent the use of the online teaching platform; Based on the second dimension, sample data whose teaching feedback meets expectations are selected from all sample data as designated samples; Based on the first dimension, cluster analysis is performed on the designated samples through big data analysis to obtain multiple clusters, each cluster containing multiple designated samples; For each cluster, determine the data distribution of the third dimension and the fourth dimension corresponding to each designated sample through big data analysis, and generate a first classroom management plan corresponding to the third dimension and a second classroom management plan corresponding to the fourth dimension according to the data distribution; Determine the current teaching environment of this online teaching, determine the cluster to which this online teaching belongs in the cluster analysis based on the current teaching environment, and the first classroom management plan and the second classroom management plan corresponding to the cluster, and recommend the first classroom management plan to the teacher before the online teaching, and recommend the second classroom management plan to the teacher during the online teaching.

2. The method according to claim 1, characterized in that The online teaching platform includes: a main window module, a secondary window module, a teaching tool module, and an auxiliary explanation module. The teaching tool module includes a plurality of teaching tools for teachers to use; The sub-dimensions of the first dimension include at least: subject, grade, teacher’s total teaching experience, teacher’s online teaching experience, and the degree of intelligence of the smart classroom; The sub-dimensions of the second dimension at least include: questionnaire feedback data, teaching achievement feedback data; The sub-dimensions of the third dimension include at least: the preparation of smart devices and the preparation of teaching courseware; The sub-dimensions of the fourth dimension include at least: the usage of each teaching tool in the online teaching platform.

3. The method according to claim 2, characterized in that Based on the first dimension, cluster analysis is performed on the designated samples through big data analysis to obtain multiple clusters, specifically including: Obtain the configuration status of the smart devices already set up in the smart classroom and the degree of centralization of the control software of the smart devices; A first score corresponding to the configuration state is obtained based on the sum of scores corresponding to each of the set smart devices, and a second score corresponding to the concentration level is obtained based on the number of control software of each of the set smart devices; wherein the lower the number of control software, the higher the second score; Obtaining a degree of intelligence of the smart classroom based on the sum of the first score and the second score; Divide multiple intervals based on the intelligence level, and assign weights to each sub-dimension of the first dimension in each interval; wherein, in intervals with higher intelligence levels, the weights of teachers' online teaching experience and the intelligence level of smart classrooms are higher; For each interval, clustering is performed using a K-means clustering algorithm based on the weights of each dimension of the first dimension to obtain a plurality of clusters.

4. The method according to claim 2, characterized in that: For each cluster, the data distribution of the third dimension and the fourth dimension corresponding to each specified sample is determined through big data analysis, and according to the data distribution, a first classroom management plan corresponding to the third dimension and a second classroom management plan corresponding to the fourth dimension are generated respectively, specifically including: For each cluster, determine the third dimension and the fourth dimension corresponding to each specified sample therein; For each item of data in the third dimension and the fourth dimension, determine the corresponding frequency; wherein each item of data in the third dimension includes specific values ​​of all sub-dimensions contained therein, and each item of data in the fourth dimension only selects the specific value of any sub-dimension contained therein; Generate a cross table corresponding to the third dimension and the fourth dimension according to the frequency, and determine whether there is a correlation relationship between the data in the third dimension and the fourth dimension based on a chi-square test; Sorting based on the frequency of each data item in the third dimension to obtain the first data distribution corresponding to the third dimension; Sorting is performed based on the frequency of each item of data in the fourth dimension, and when there is a single item of data in the fourth dimension, which is associated with the single item of data in the third dimension, and the ranking of the single item of data in the third dimension in the first data distribution is before the first preset order, then the ranking of the single item of data in the fourth dimension is increased to before the second preset order, so as to obtain the second data distribution corresponding to the fourth dimension; According to the first data distribution, several single data items sorted before the first preset order are selected as the first classroom management plan; and according to the second data distribution, several single data items sorted before the second preset order are selected as the second classroom management plan.

5. The method according to claim 4, characterized in that In this online teaching, the second classroom management plan is recommended to teachers, including: When the teacher performs the online teaching and triggers the teaching tool module, the teaching tool included in the second classroom management solution is recommended to the teacher in the teaching tool module; Acquire the use status of the teaching tool in real time, and when the use status hits a preset target status, collect data for the main window module to obtain teaching data; The teaching data is analyzed to obtain the knowledge points currently being taught, and corresponding auxiliary explanation information is added to the auxiliary explanation module according to the knowledge points.

6. The method according to claim 5, characterized in that Acquiring the use status of the teaching tool in real time, and when the use status hits the preset target status, collecting data for the main window module to obtain teaching data, specifically including: For each teaching tool in the learning tool module, determine its corresponding type; the types include: sharing type, interactive type, auxiliary type, and management type; Real-time acquisition of the usage status of a specified type of teaching tool; the specified type includes: interactive type, auxiliary type; the interactive type includes: online voting, interactive games, the auxiliary type includes: electronic paintbrush, electronic whiteboard, virtual laboratory table; For the online voting, after the voting result is generated, the main window module is screenshotted, OCR recognition is performed according to the screenshot to obtain the first text content, and the second text content included in the online voting is determined, and the first text content and the second text content are used as teaching data; For the interactive game, when the interactive game is started, voice data generated by the teacher during the execution of the interactive game is collected, voice recognition is performed on the voice data to obtain a third text content, and a fourth text content corresponding to the text description of the interactive game is determined, and the third text content and the fourth text content are used as teaching data; For the electronic whiteboard, when the erase function or the close function of the electronic whiteboard is activated and before being executed, a screenshot of the electronic whiteboard in the main window module is taken, and OCR recognition is performed based on the screenshot to obtain fifth text content as teaching data; For the electronic brush, when the usage time of the electronic brush exceeds a preset time, a screenshot is taken of the main window module containing the drawing traces of the electronic brush, and OCR recognition is performed based on the screenshot to obtain sixth text content as teaching data; or, when the drawing pattern of the electronic brush meets the preset pattern requirements, a screenshot is taken of the main window module, and OCR recognition is performed on the content contained in the drawing pattern and in the preset range above the screenshot to obtain seventh text content as teaching data; With respect to the virtual experiment platform, voice data generated by the teacher during the execution of the virtual experiment platform is collected, and voice recognition is performed on the voice data to obtain an eighth text content as teaching data.

7. The method according to claim 6, characterized in that According to the teaching data, the knowledge points currently being taught are analyzed, including: The first text content, the second text content, the third text content, the fourth text content, and the eighth text content are used as first data to be analyzed; the fifth text content, the sixth text content, and the seventh text content are used as second data to be analyzed; Performing pattern recognition on the second data to be analyzed to determine the structural position of the second data to be analyzed in the screenshot, wherein the structural position includes: title, text, chart, and remarks; Based on the entity recognition algorithm, knowledge points are extracted from the first data to be analyzed and the second data to be analyzed; wherein, if no knowledge points are extracted for the seventh text content, the preset range is expanded and OCR recognition is performed again to obtain the seventh text content; Selecting at least some of the knowledge points extracted, and adding corresponding auxiliary explanation information in the auxiliary explanation module; Among them, the selection priority of the knowledge points is: in the first data to be analyzed, the first text content> the third text data= the eighth text content> the second text content> the fourth text content; in the second data to be analyzed, the title> the text> the chart=the notes.

8. The method according to claim 7, characterized in that Based on the entity recognition algorithm, knowledge points are extracted from the first data to be analyzed and the second data to be analyzed, specifically including: Cleaning the first data to be analyzed and the second data to be analyzed to obtain cleaned data, wherein the data cleaning includes: removing special characters, removing punctuation marks, removing stop words, and performing word segmentation operations; Performing entity recognition on the cleaned data through a pre-trained entity recognition model to obtain a first entity; and performing keyword recognition on the cleaned data based on a word vector to obtain a first keyword; Based on the domain dictionary corresponding to the teaching subject, the first entity and the first keyword are filtered respectively to obtain a second entity and a second keyword; Based on semantic association, the second entity and the second keyword are combined to extract knowledge points.

9. The method according to claim 7, characterized in that: After performing speech recognition on the speech data, the method further includes: When the duration of the voice data is longer than a preset duration, performing data cleaning and data conversion on the voice data to obtain standard data; Extracting features from the standard data to obtain speech features, and performing feature dimension reduction on the speech features, wherein the speech features include fundamental frequency features, energy features, and spectrum features; The support vector machine obtained by supervised training is used to identify the speech features after feature dimensionality reduction to determine the corresponding speech tags, wherein the speech tags include: speech rate tags and intonation tags; Dividing the teaching data on a time axis based on the voice tags to obtain a plurality of time periods, wherein each time at least one of the voice tags changes, a new time period is obtained; For each time period, based on the speech label corresponding to the time period, a corresponding weight is assigned to the text content of speech recognition within the time period; the slower the speech rate label and the stronger the intonation label, the higher the corresponding weight; The speech recognition results within the time period in which the weight is lower than the preset threshold are deleted, so that the remaining speech recognition results are used as the third text content or the eighth text content.

10. A smart classroom management device based on big data analysis, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: implement the smart classroom management method based on big data analysis as described in any one of claims 1 to 9.

Images