Cloud education platform management system based on big data

By using the big data management system on the cloud education platform, collecting and analyzing the voice interaction information between teachers and students, and using the attention semantic extraction model, the problem of inaccurate semantic extraction is solved, personalized tutoring and teaching adjustments are realized, and the improvement of Teaching effectiveness and student participation.

CN119067130AInactive Publication Date: 2024-12-03ANHUI YOUXIN EDUCATION TECH CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411179145.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the cloud education platform, the accuracy of semantic extraction when teachers interact with students is not high, resulting in teachers being unable to fully understand students' thinking defects and knowledge blind spots, and being unable to formulate personalized tutoring plans.

Method used

The cloud education platform management system based on big data is adopted to collect voice information between students and teachers through the data acquisition module, and the data processing module converts it into text information. The data analysis module uses the semantic extraction model with attention to perform semantic feature extraction. Generate personalized tutoring plans.

Benefits of technology

It realizes the detailed analysis and evaluation of the content of students' answers by teachers, provides personalized teaching tutoring suggestions and improvement strategies, and improves teaching effectiveness and students' learning enthusiasm.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067130A_ABST
    Figure CN119067130A_ABST
Patent Text Reader

Abstract

The invention discloses a cloud education platform management system based on big data, and relates to the field of cloud education platform management, and the system comprises a cloud server platform which is in communication connection with a data collection module, a data processing module, a data analysis module and a personalized tutoring module. An information base is arranged in the cloud server and is used for storing basic information of teachers and students, educational resource information, interactive voice information of the teachers and the students when the teachers carry out teaching and the like; the data acquisition module is used for acquiring voice information when students interact with teachers; the data processing module is used for processing voice information when the students interact with the teacher and converting the voice information into text information; the data analysis module is used for carrying out semantic feature extraction on the text information, obtaining complete logic of question answering of the students and carrying out semantic analysis; and the personalized tutoring module generates a personalized tutoring scheme according to the semantic analysis result and displays the personalized tutoring scheme on the cloud education platform for teachers and students to access.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of cloud education, and in particular to a cloud education platform management system based on big data. Background Art

[0002] The cloud education platform is an online education service platform based on cloud computing and network technology, providing a full range of educational services from online course management, teaching resource sharing to distance learning and examinations. The cloud education platform connects students, teachers and educational institutions through the Internet, provides flexible learning methods and tools, breaks through the limitations of time and space, and allows students to study online anytime and anywhere.

[0003] In the cloud education platform, the interaction between teachers and students is an important link. One of the current problems is that the accuracy of semantic extraction during the interaction between teachers and students is not high, which makes it impossible for teachers to fully understand students' thinking defects and knowledge blind spots, and therefore cannot make personalized tutoring plans. Summary of the invention

[0004] The purpose of the present invention is to provide a cloud education platform management system based on big data to solve the shortcomings of the background technology.

[0005] In order to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a cloud education platform management system based on big data, including a cloud server platform, which is communicatively connected with a data acquisition module, a data processing module, and a data analysis module.

[0006] The data collection module is used to collect voice information when students interact with teachers to form an audio file; The data processing module is used to process the audio file and convert it into text form to form text information; The data analysis module is used to extract semantic features from text information and obtain semantic expression results. Through the semantic expression results, teachers can analyze logical errors, thinking loopholes, etc., and provide personalized teaching and tutoring suggestions for teachers. At the same time, according to the logical defects and knowledge blind spots of students' answers, teachers can also be provided with corresponding improvement strategies and teaching resources to meet students' personalized needs.

[0007] The personalized tutoring module is used to generate a personalized tutoring plan based on the semantic expression results. In a preferred embodiment, the process of collecting voice information when students interact with teachers is as follows: in video teaching, the real-time voice call function is turned on, the voice interaction information between teachers and students is recorded, and saved in the form of an audio file on the cloud server.

[0008] In a preferred embodiment, the data processing module processes the audio file and forms text information in the following process: the audio file is divided into blocks, the data processing module recognizes each block, and saves the recognition results in the form of text information to the cloud server. For large audio files, they can be divided into smaller blocks for processing, which helps to improve processing efficiency and avoid timeouts or excessive resource consumption caused by too long audio.

[0009] The basis for segmenting the audio file into blocks is as follows: segmenting the audio file into blocks according to the volume; setting a minimum volume threshold, starting from the first audio in the audio file, when the volume value is lower than the minimum volume threshold, segmenting the audio file until the audio ends, thereby forming multiple audio file blocks in sequence.

[0010] If the audio is divided into multiple blocks for processing, the recognition results of each block need to be merged in the end. The results can be sorted and merged according to the timestamp or the order of the blocks to obtain the complete audio recognition text.

[0011] In a preferred embodiment, the process of extracting semantic features from text information by the data analysis module is as follows: 1. Collect interactive voice data between teachers and students during class and build an interactive question-and-answer voice data set; 2. Train a semantic extraction model with attention based on the interactive question-answering speech dataset to obtain a trained semantic extraction model; 3. Use the interactive question-and-answer voice data to be extracted as the input of the semantic extraction model to obtain the semantic extraction result.

[0012] In a preferred embodiment, the semantic extraction model with attention includes: an embedding layer, a semantic encoding layer, and a decoding layer; The embedding layer is a trained word embedding model that converts the input text information into a word embedding vector form for output; The semantic encoding layer is composed of several encoders, which are used to obtain hidden features in each encoder; The decoding layer is used to decode and output the hidden features in different encoders through the attention mechanism, and make predictions to obtain semantic extraction results.

[0013] In a preferred embodiment, the training method of the word embedding model comprises the following steps: 1. Obtain a large number of interactive question-and-answer voice data sets during teaching interaction between teachers and students; the data sets include multiple historical interactive question-and-answer voice data and corresponding real interactive question-and-answer data; 2. Preprocess the interactive question-answering speech data set, including converting the audio file into a text file, removing special characters, punctuation marks, numbers, splitting the text into words, and removing prepositions to form a preprocessed text data; 3. Based on the preprocessed text data, build a vocabulary containing all the words that appear in the text data; 4. Construct a context set based on the vocabulary, taking each word in the vocabulary as the center, selecting three words within the word range as a set, and forming multiple context sets; 5. Use the constructed context set data as input to train the word embedding model. After the training is completed, the word embedding vector is obtained; In a preferred embodiment, the encoding process of the semantic encoder layer is: The word embedding vector is input into the encoder for encoding. At the first time step, the encoder calculates the first hidden feature based on the embedding vector of the first word, and then updates the hidden feature using the embedding vector of the second word at the next time step. In the encoder model, a time step refers to each discrete time point or step when the model processes sequence data.

[0014] In a preferred embodiment, the decoding process of the decoding layer is: Step 1: Calculate the initial hidden features of the decoder based on the final hidden features of the encoder; Step 2: At each time step, the decoder predicts the output of the current time step based on the current hidden features and the previously generated output. The output result can be a word or a character. Step 3: The decoder performs weighted aggregation on the hidden features of the current corresponding coding layer to obtain a weighted aggregation vector; Step 4: According to the weighted pooled vector and the previous hidden features, the current hidden features are obtained, and the word output of the next time step is continued; By repeatedly executing steps 2 to 4, the decoder will continue to generate output until a complete semantic expression result is generated.

[0015] In a preferred embodiment, the personalized tutoring module forms a complete question-and-answer logic for interaction between teachers and students based on the complete semantic expression results. Based on the complete question-and-answer logic, the teacher analyzes the logical flaws and knowledge blind spots in the students' answers to questions, and thus formulates a personalized tutoring plan for the students, which is displayed in the personalized tutoring module of the cloud education platform. Teachers and students can log in to the cloud education platform to access the tutoring plan, which includes question answering, knowledge point reinforcement exercises, reference material recommendations, study plan arrangements, and tutoring feedback and evaluation.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. Provide timely and accurate student evaluation: Through speech-to-text conversion and feature extraction, teachers can obtain students’ answers in class discussions or question-and-answer sessions, which enables teachers to conduct detailed analysis and evaluation of students’ answers, thereby accurately understanding students’ mastery of knowledge, level of understanding, and way of thinking.

[0017] 2. Personalized tutoring and teaching adjustment: By correcting the logical errors and thinking loopholes in students' answers, teachers can provide targeted explanations and guidance to help students better digest and understand knowledge, thereby developing personalized tutoring plans based on different needs and levels. In addition, teachers can adjust teaching strategies in a timely manner based on students' learning performance and feedback to provide more effective teaching.

[0018] 3. Improve student participation and enthusiasm: Through interactive teaching, teachers can better understand students' interests in the course. Based on this information, teachers can design more attractive and targeted teaching content, mobilize students' enthusiasm and initiative in learning, and stimulate students' participation and enthusiasm. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0020] Figure 1 : is a system block diagram of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Example 1, please refer to Figure 1 As shown, the present embodiment is used for a cloud education platform management system based on big data: it includes a cloud server platform, and the cloud server platform is communicatively connected with a data acquisition module, a data processing module, and a data analysis module.

[0023] The data collection module is used to collect voice information when students interact with teachers; before the video teaching begins, make sure to set up the audio and video recording functions so that the sound and picture of the entire teaching process can be recorded. After the recording is completed, their voice answers will be uploaded to the server of the cloud teaching platform. The server transcodes the voice file and converts it into a format that can be used for subsequent processing, such as MP3, WAV, etc.

[0024] The data processing module is used to process the voice information during the interaction between students and teachers, convert it into text form, and form text information; the specific implementation method is: the audio file is divided into blocks, and the data processing module performs voice recognition block by block, and the recognition results are saved in the form of text information on the cloud server. For audio files, they can be divided into smaller blocks for processing, which helps to improve processing efficiency and avoid timeouts or excessive resource consumption caused by too long audio. If the audio is divided into multiple blocks for processing, the recognition results of each block need to be merged at the end. The results can be sorted and merged according to the timestamp or the order of the blocks to obtain a complete audio recognition text; The basis for segmenting the audio file into blocks is as follows: segmenting the audio file into blocks according to the volume; setting a volume threshold, starting from the first audio in the audio file, when the volume is lower than the threshold, segmenting the audio file to form file blocks until the audio ends, thereby forming multiple audio file blocks in sequence.

[0025] The data analysis module is used to extract text features from text information, obtain the complete logic of students' answers to questions, and analyze the complete logic of students' answers to form logical analysis results; by analyzing the logical errors and thinking loopholes in students' answers, it provides teachers with personalized teaching and tutoring suggestions. At the same time, it can also provide teachers with corresponding improvement strategies and teaching resources based on the logical defects and knowledge blind spots in students' answers to questions to meet students' personalized needs.

[0026] The process of data analysis module extracting semantic features from text information is as follows: 1. Collect interactive question-and-answer voice data between teachers and students during class and build an interactive question-and-answer voice data set; 2. Train a semantic extraction model with attention based on the interactive question-answering speech dataset to obtain a trained semantic extraction model; 3. Use the interactive question-and-answer voice data to be extracted as the input of the semantic extraction model to obtain the semantic extraction result.

[0027] The attention mechanism is an important technology in deep learning. It can be used to improve the model's ability to process and understand input data. The introduction of the attention mechanism enables the model to pay more attention to the parts related to the current task when processing the input sequence. Specifically, the attention mechanism can help the model determine which parts of the input data are relevant and need more attention based on the different parts of the input data. By calculating the attention weights, the model can dynamically allocate attention to different positions or elements in the input sequence. In this way, the model can perform targeted weighted focusing based on the contextual information of the input, thereby extracting and utilizing the most relevant information.

[0028] The semantic extraction model with attention includes: an embedding layer, a semantic encoding layer and a decoding layer; The embedding layer is a trained word embedding model that converts the input text information into word embedding vector form for output; The role of the embedding layer is to map the input discrete symbols (such as words and characters) into continuous numerical representations, that is, to convert them into vector form. The purpose of this is to enable the model to better process text data and capture the semantics and associations between words.

[0029] The input of the embedding layer can be a discrete identifier (such as the index of a word). After the embedding operation, each identifier is mapped to a fixed-length vector representation. In this way, the text data can be represented as a matrix consisting of embedded vectors, where each row corresponds to a position or word in the input sequence.

[0030] In the attention mechanism model, this vector matrix transformed by the embedding layer will be used as the input of the encoding layer. The encoding layer is responsible for further processing the input data and extracting features for the final task or prediction.

[0031] The semantic encoding layer is composed of several encoders, which are used to obtain the hidden features in each encoder; the encoding layer is responsible for converting the input sequence into a series of hidden feature representations. The encoding layer can adopt different structures. Here, a recurrent neural network (model) is used. The attention mechanism plays a key role in the encoding layer. By weighting the output of the encoding layer, the most relevant information can be extracted and utilized.

[0032] The decoding layer is used to decode and output the hidden features in different encoders through the attention mechanism, and make predictions to obtain semantic extraction results.

[0033] The role of the decoding layer is to combine the output of the encoding layer with the previous context information to generate the final output sequence.

[0034] The training method of the word embedding model includes: Obtain a large number of interactive question-and-answer voice data sets during teaching interaction between teachers and students; the data sets include multiple historical interactive question-and-answer voice data and corresponding real interactive question-and-answer data; Preprocessing the interactive question-answering speech data set includes converting the audio file into a text file, removing special characters, punctuation marks, and numbers, splitting the text into words, and removing prepositions to form a preprocessed text data; Based on the preprocessed text data, build a vocabulary containing all the words that appear in the text data; Build a context set based on the vocabulary, take each word in the vocabulary as the center, select three words within the word range as a set, and form multiple context sets; When constructing context set data, you need to determine a fixed context set size and the number of context words around each set center word. This set size can be determined according to task requirements. Generally, a window size of 3 to 10 is a common choice.

[0035] Let's take an example to illustrate the process of building context collection data: Suppose there is the following sentence: "I don't understand this concept, can you explain it again?"

[0036] Each word can be used as the central word of the set, and then the context words within the set can be selected. Assuming the set size is set to 3, the following context set data can be obtained: For the central word "I": context set: [no, understand, this concept] For the central word "understanding": context set: [I, no, this concept] For the central word "this concept": context set: [I, don't, understand] In this example, the set size is 3, and each center word contains the three nearest context words. It can be seen that the context sets corresponding to different center words are different, and the order of words in the set maintains the relative position relationship in the original sentence.

[0037] The constructed context set data can be used as training samples and input into the word embedding model for training. By learning the commonality and semantic relevance of words in the set, the model can generate word vector representations that capture the semantic and contextual relationships between words.

[0038] The encoding process of the semantic encoder layer is: First, in the embedding layer, each word is mapped to a low-dimensional continuous vector representation. For example, "I" can be mapped to a 5-dimensional vector and "no" can be mapped to a 4-dimensional vector.

[0039] Next, with the output of the embedding layer as input, the encoder iterates over each word in the input sequence and computes the hidden features of the current time step at each time step. For example, at the first time step, the encoder computes the first hidden feature based on the embedding vector of the first word, and then updates the hidden feature using the embedding vector of the second word at the next time step.

[0040] When the encoder processes the entire input sequence, the concatenation of the forward and backward hidden features of the last time step can be used as the output representation of the encoding layer. This output vector can be used for subsequent classification tasks to help the model understand the contextual information of the entire text.

[0041] In the encoder model, a time step refers to each discrete time point or step when the model processes sequence data. Specifically, for an input sequence of length T, the encoder will process each element in the sequence in turn and update its hidden features at each time step. The hidden features of each time step contain information about the input at the corresponding time point, helping the model establish context and capture correlations in the sequence.

[0042] Let's take a simple example to illustrate the concept of time steps. Suppose there is a sentence: "I don't understand this concept", where each word represents a time step. Now use the encoder model to process this sentence: Time step 1: The input is "i", and the encoder model calculates the hidden features of the first time step and updates.

[0043] Time step 2: The input is “No”, and the encoder model calculates the hidden features of the current time step based on the current input and the hidden features of the previous time step.

[0044] Time step 3: The input is "understanding", and the encoder model continues to calculate the hidden features of the current time step.

[0045] Time step 4: The input is "this concept", and the encoder model calculates the hidden features of the current time step based on the current input and the hidden features of the previous time step.

[0046] Through the above process, the encoder model gradually processes each word in the sentence and updates the hidden features at each time step. In this way, the model can use the information in the hidden features to understand the semantics and context of the sentence after processing the entire input sequence.

[0047] The decoding process of the decoding layer is: 1. First, you need to initialize the decoder's hidden features. This initial hidden feature is usually calculated based on the encoder's final hidden feature.

[0048] 2. Next, it starts generating the output sequence, one time step at a time. At each time step, the decoder predicts the output of the current time step based on the current hidden features and the previously generated partial sequence. This prediction can be a word, character, or other form of identifier.

[0049] 3. To generate a prediction for the current time step, the decoder uses an attention mechanism to compute a weighted pool of the encoder layer outputs. The decoder's current hidden features are treated as a query vector, and attention weights are computed on the encoder layer outputs. These weights measure how relevant different positions in the encoder layer output are to the current hidden features.

[0050] 4. Using the calculated attention weights, the decoder performs weighted aggregation on the encoding layer output. By multiplying the attention weights with the encoding layer output vector, a weighted aggregation vector representation is obtained. This vector contains the encoding layer information that is most relevant to the current time step.

[0051] 5. Next, the decoder updates its own hidden features based on the weighted pooled vector of the current time step and the previous hidden features. This allows the decoder to combine the information and context of the encoding layer to better control the generation of the output of the current time step.

[0052] Repeat steps 2-5: Repeat the above steps until a complete output sequence is generated or a specific stop condition is met. In this way, a complete semantic expression is output.

[0053] The personalized tutoring module forms a complete question-and-answer logic for the interaction between teachers and students based on the complete semantic expression results. Based on the complete question-and-answer logic, the teacher analyzes the logical flaws and knowledge blind spots in the students' answers to questions, and thus develops a personalized tutoring plan for the students, which is displayed in the personalized tutoring module of the cloud education platform. Teachers and students can log in to the cloud education platform to access the tutoring plan, which includes question answering, knowledge point reinforcement exercises, reference material recommendations, study plan arrangements, and tutoring feedback and evaluation. Teachers can also further adjust and optimize the plan based on the students' learning situation.

[0054] Problem Solving: Provide students with detailed answers and solutions to their individual problems. For example, if a student is confused about a math problem, the personalized tutoring solution will provide clear explanations, steps and examples to help students resolve their confusion.

[0055] Knowledge point reinforcement exercises: Based on the student's performance and understanding in the interactive question-and-answer session, the personalized tutoring program can recommend some targeted knowledge point reinforcement exercises. For example, if a student has a weak grasp of a certain knowledge point in English grammar, the personalized tutoring program will provide relevant exercises and answers to consolidate the student's knowledge point.

[0056] Reference material recommendation: The personalized tutoring program can also recommend some suitable reference materials based on the students’ learning needs and characteristics, such as textbook chapters, online documents, learning websites or video resources. For example, if a student is interested in a certain historical event, the personalized tutoring program will provide some related reading materials and website links to help the student gain in-depth understanding and learning.

[0057] Study plan: Personalized tutoring can generate a personalized study plan based on the student's learning progress and goals. For example, if a student is behind in a certain subject, the personalized tutoring plan will provide a detailed study plan, including daily tasks and schedules, to help the student catch up and achieve his or her learning goals.

[0058] Tutoring feedback and evaluation: Personalized tutoring programs can also provide teachers with some feedback and evaluation on students’ learning. For example, personalized tutoring programs may record information such as the accuracy of students’ answers to questions and reaction time in interactive Q&A sessions and present it to teachers so that they can track and evaluate students’ learning.

[0059] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A cloud education platform management system based on big data, including a cloud server platform, characterized in that: The cloud server platform is communicatively connected with a data acquisition module, a data processing module, a data analysis module and a personalized tutoring module; The data collection module is used to collect voice information when students interact with teachers to form an audio file; The data processing module is used to process the audio file and generate corresponding text information; The data analysis module is used to extract semantic features from text information and obtain semantic expression results; The personalized tutoring module is used to generate a personalized tutoring plan according to the semantic expression result.

2. A cloud education platform management system based on big data according to claim 1, characterized in that: The process of collecting voice information when students interact with teachers is as follows: in video teaching, turn on the real-time voice call function, record the voice information when teachers and students interact, and save it in the form of audio files on the cloud server.

3. A cloud education platform management system based on big data according to claim 2, characterized in that: The audio file is divided into blocks according to the volume, a minimum volume threshold is set, and starting from the first audio in the audio file, when the volume value is lower than the minimum volume threshold, the audio file is segmented to form a file block until the audio ends, and multiple audio file blocks are formed in sequence; The data processing module recognizes and processes the audio file blocks one by one to form text information.

4. A cloud education platform management system based on big data according to claim 3, characterized in that: The process of data analysis module extracting semantic features from text information is as follows: Use text information to build an interactive voice dataset; Based on the interactive speech dataset, a semantic extraction model with attention is trained to obtain a trained semantic extraction model; The interactive speech data set to be extracted is input into the semantic extraction model to obtain the semantic expression result.

5. A cloud education platform management system based on big data according to claim 4, characterized in that: The semantic extraction model with attention includes: embedding layer, semantic encoding layer and decoding layer; The embedding layer is a trained word embedding model that converts the input text information into a word embedding vector form for output; The semantic encoding layer is composed of several encoders, which are used to obtain hidden features in each encoder; The decoding layer is used to decode and output the hidden features in different encoders through the attention mechanism, and make predictions to obtain semantic extraction results.

6. A cloud education platform management system based on big data according to claim 5, characterized in that: The training method of the word embedding model includes: Acquire an interactive voice data set when a teacher and a student interact with each other in teaching; the interactive voice data set includes a plurality of historical interactive voice data; Preprocessing the interactive speech data set includes converting the audio file into a text file, removing special characters, punctuation marks, numbers, splitting the text into words, and removing prepositions to form a preprocessed text data; Based on the preprocessed text data, build a vocabulary containing all the words that appear in the text data; Build a context set based on the vocabulary, take each word in the vocabulary as the center, select three words within the word range as a set, and form multiple context sets; Use the constructed context set data as input to train the word embedding model. After the training is completed, the word embedding vector is obtained.

7. A cloud education platform management system based on big data according to claim 5, characterized in that: The encoding process of the semantic encoder layer is: The word embedding vector is input into the encoder for encoding. At the first time step, the encoder calculates the first hidden feature based on the embedding vector of the first word, and then updates the hidden feature using the embedding vector of the second word at the next time step; the time step is the step corresponding to each time point when the encoder processes the encoding.

8. The cloud education platform management system based on big data according to claim 5, characterized in that: The decoding process of the decoding layer is: Step 1: Calculate the initial hidden features of the decoder based on the final hidden features of the encoder; Step 2: At each time step, the decoder predicts the output of the current time step based on the current hidden features and the output generated at the previous time step. The output result is a word or a character. Step 3: The decoder performs weighted aggregation on the hidden features of the current corresponding coding layer to obtain a weighted aggregation vector; Step 4: According to the weighted pooled vector and the hidden features of the previous time step, the hidden features of the current time step are obtained, and the output of the next time step is continued to be generated; By repeatedly executing steps 2 to 4, the decoder will continue to generate output until a complete semantic expression result is generated.

9. A cloud education platform management system based on big data according to claim 8, characterized in that: Based on the complete semantic expression results, teachers analyze the logical flaws and knowledge blind spots in students' answers to questions, and thus develop personalized tutoring plans for students. The plans are displayed in the personalized tutoring module of the cloud education management system. Teachers and students can log in to the cloud education platform to access the tutoring plans.

Citation Information

Patent Citations

  • Distance-teaching interactive method, server, terminal and system

    CN110033659A

  • Online teaching artificial intelligence tutoring method, medium and system

    CN118053331A

  • Internet teaching platform-based accompanying teaching system

    WO2019075826A1