An educational intelligent learning path recommendation system

Abstract

The present invention belongs to the technical field of path optimization, and discloses an educational intelligent learning path recommendation system; including: establishing an efficient user learning willingness evaluation system and dynamically optimizing the learning path through multimodal data fusion. Collect the user's emotional text data and voice data, convert the voice into text and perform unified processing, including removing neutral vocabulary, punctuation marks and calculating the emotional score; establish a positive / negative vocabulary, combine the emotion intensity label, train a multi-layer perceptron model, predict the emotion intensity and calculate the willingness evaluation value in combination with the score. At the same time, the comprehensive willingness evaluation value is further calculated in combination with the user's screen switching frequency and heart rate mutation data. Finally, based on the evaluation value and stress status, dynamically recommend a learning path suitable for the user, and flexibly adjust the content difficulty according to the user status.

Description

Technical Field

[0001] The present invention relates to the technical field of path optimization, and more specifically, to an educational intelligent learning path recommendation system. Background Art

[0002] Intelligent education refers to the use of advanced technologies such as artificial intelligence, big data, and cloud computing to optimize and improve the education process, and to realize functions such as personalized teaching, intelligent tutoring, and automated assessment to improve the quality and efficiency of education.

[0003] The Chinese patent with authorization announcement number CN114419711B discloses a concentration recognition method based on an AI education system. It captures input video frames of multiple users listening to lectures, detects face areas to build a face appearance model, estimates face direction to calculate concentration, detects frontal posture and gaze time to evaluate user attention, associates body behavior with emotion type labels, and uses features extracted from video frame data to train a classifier to detect users' emotional feedback from lectures.

[0004] The Chinese patent application publication number CN111046852A discloses a method, device and readable storage medium for generating a personal learning path. Based on the analysis results of students' actions and learning feedback during the learning process and a preset learning mode effect response library, as well as the educational knowledge structure specified by the syllabus, the method plans the students' learning paths and generates a personal learning path exclusive to each student, thereby improving the adaptability of the personal learning path to the corresponding students and improving the students' learning efficiency.

[0005] The Chinese patent with authorization announcement number CN118586611B discloses an AI intelligent education method and system based on big data. By analyzing teaching materials and building a knowledge association network, identifying the interdependence between knowledge points, and identifying knowledge points that affect learning progress, teachers can adjust teaching content and strategies in a targeted manner and optimize students' learning paths.

[0006] However, with the popularization of lifelong learning, the above patent fails to effectively provide personalized learning paths for self-learning users, and fails to recommend learning paths on demand based on the user's real-time complex situation.

[0007] In view of this, the present invention proposes an educational intelligent learning path recommendation system to solve the above problems. Summary of the invention

[0008] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an educational intelligent learning path recommendation system.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A method for recommending an intelligent learning path for education, the method comprising:

[0011] Collect emotional text data, emotional voice data and user learning sign data; user learning sign data includes user screen switching frequency and heart rate mutation value;

[0012] Establish a neutral word library, a positive word library and a negative word library; pre-process the emotional text data; compare the heart rate mutation value with the preset first-level heart rate mutation threshold, and when the heart rate mutation value is greater than or equal to the first-level heart rate mutation threshold, mark the user as a high-stress state; set the second-level heart rate mutation threshold to be less than the first-level heart rate mutation threshold, and when the heart rate mutation value is less than or equal to the preset second-level heart rate mutation threshold, mark the user as a low-stress state;

[0013] Calculate sentiment scores based on preprocessed sentiment text data and train a machine learning model to identify sentiment intensity;

[0014] Based on the conversion of emotional speech data into text data and marking it as second-class emotional text data; the second-class emotional text data is processed in the same way as the emotional text data;

[0015] Calculate the willingness evaluation value based on the output of the machine learning model and the sentiment score; calculate the comprehensive willingness evaluation value based on the willingness evaluation value and the user's screen switching frequency;

[0016] The comprehensive willingness evaluation value is compared with the comprehensive willingness evaluation threshold. When the comprehensive willingness evaluation value is less than or equal to the comprehensive willingness evaluation threshold, the learning path recommendation plan is started. The learning path recommendation plan includes: when the user is in a low-stress state, priority is given to recommending learning content that is more difficult and has deeper content than the current learning content; when the user is in a high-stress state, priority is given to recommending learning content that is less difficult and has shallower content than the current learning content.

[0017] Preferably, the sentiment text data includes discussion forum speeches, homework feedback and comments;

[0018] The preprocessing process includes: removing words and punctuation marks belonging to the neutral vocabulary in the sentiment text data;

[0019] The neutral word library is a collection of words with unclear emotional states; the positive word library contains words with positive emotional states; the negative word library contains words with negative emotional states; the emotional intensity refers to the intensity or intensity of the emotions shown by the user when expressing emotions; the emotional speech data includes the language spoken by the user when the user is learning autonomously on the user side;

[0020] The user screen switching frequency is the number of times the user switches to other apps or software on the user end within one minute; the heart rate mutation value is the difference between the sum of the user's heart rate values ​​in the last 30 seconds of each minute and the sum of the heart rate values ​​in the first 30 seconds.

[0021] Preferably, the process of calculating the sentiment score includes:

[0022] The preprocessed sentiment text data is traversed in sequence; when the preprocessed sentiment text data is included in the positive vocabulary, the sentiment score is +1 point; when the preprocessed sentiment text data is included in the negative vocabulary, the sentiment score is -2 points; when it is not included in the positive vocabulary or the negative vocabulary, the sentiment score is +0.1 points.

[0023] Preferably, the training process of the machine learning model includes:

[0024] Jieba is used to segment the preprocessed emotional text data into phrases, and each phrase is converted into a feature vector. The intensity of the emotion is used as the label corresponding to the feature vector. Each set of feature vectors and the label corresponding to each set of feature vectors are constructed as a sample, and multiple samples are collected to construct a data set; the data set is divided into a training set, a validation set, and a test set, where the training set accounts for 80% of the data set, and the validation set and the test set each account for 10% of the data set;

[0025] The training set is used as the input of the machine learning model, and the machine learning model uses the intensity of emotion corresponding to the real-time preprocessed emotional text data as output; the intensity of emotion corresponding to a set of real-time feature vectors is used as the prediction target, and minimizing the loss function value of the machine learning model is used as the training target; when the loss function value of the machine learning model is less than or equal to the preset target loss value, the training is stopped;

[0026] The machine learning model loss function is the mean square error; the mean square error is calculated by Minimize to train the model; in the loss function is the loss function value, i is the feature vector group number; is the number of eigenvector groups; is the label corresponding to the i-th group of feature vectors, is the intensity of emotion predicted by the i-th group of feature vectors; the machine learning model is a multi-layer perceptron;

[0027] Multilayer Perceptron Hidden Layers The calculation methods include:

[0028] G[ ; In the formula, is the output of the hidden layer; G is the activation function; is the bias term of the hidden layer; is the weight of the hidden layer; is the activation function; is the bias term of the previous layer; is the weight of the previous layer; x is the input training set.

[0029] The multilayer perceptron consists of 4 fully connected layers. The first 3 hidden layers contain 32, 64 and 64 neurons respectively. There is 1 SoftMax classification layer containing 3 neurons. The neurons are connected to each other through weights and activated by the ReLU activation function.

[0030] Preferably, the intensity of emotion includes mild, medium and strong, represented by one-hot encoding; the one-hot encoding creates a binary vector for each intensity of emotion:

[0031] Mild: [1, 0, 0];

[0032] Medium: [0, 1, 0];

[0033] Strong: [0, 0, 1];

[0034] Preferably, willingness evaluation value ; Where: is the sentiment score of the ath sentiment text data; is the emotional intensity evaluation value obtained by identifying the a-th emotional text data; is the total number of sentiment text data; is 0.8; The value of the emotional intensity assessment is 0.45. The values ​​of mild, moderate and strong are 1, 2 and 3 respectively.

[0035] Preferably, the same processing process includes: preprocessing, calculating sentiment scores, using them as input to a machine learning model, and setting labels;

[0036] Comprehensive willingness evaluation value ; Where: is a standardization function, which is used to standardize the comprehensive willingness evaluation value to interval; The screen switching frequency for users.

[0037] An educational intelligent learning path recommendation system, the system comprising:

[0038] The data collection module collects emotional text data, emotional voice data and user learning vital signs data; the user learning vital signs data includes the user's screen switching frequency and heart rate mutation value;

[0039] The preprocessing module sets up a neutral word library, a positive word library and a negative word library; preprocesses the emotional text data; compares the heart rate mutation value with the preset first-level heart rate mutation threshold, and when the heart rate mutation value is greater than or equal to the first-level heart rate mutation threshold, marks the user as a high-stress state; sets the second-level heart rate mutation threshold to be less than the first-level heart rate mutation threshold, and when the heart rate mutation value is less than or equal to the preset second-level heart rate mutation threshold, marks the user as a low-stress state;

[0040] Model training module, which calculates sentiment scores based on preprocessed sentiment text data and trains a machine learning model to identify the intensity of sentiment;

[0041] A multimodal processing module, based on converting the emotional speech data into text data and marking it as second-class emotional text data; performing the same processing on the second-class emotional text data as the emotional text data;

[0042] The calculation module calculates the willingness evaluation value based on the output of the machine learning model and the sentiment score; and calculates the comprehensive willingness evaluation value based on the willingness evaluation value and the user's screen switching frequency;

[0043] The decision-making module compares the comprehensive willingness evaluation value with the comprehensive willingness evaluation threshold. When the comprehensive willingness evaluation value is less than or equal to the comprehensive willingness evaluation threshold, the learning path recommendation plan is initiated; the learning path recommendation plan includes: when the user is in a low-stress state, priority is given to recommending learning content that is more difficult and has deeper content than the current learning content; when the user is in a high-stress state, priority is given to recommending learning content that is less difficult and has shallower content than the current learning content.

[0044] An electronic device comprises: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the above-mentioned method for recommending an intelligent educational learning path by calling the computer program stored in the memory.

[0045] A computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned method for recommending an intelligent educational learning path.

[0046] The technical effects and advantages of the educational intelligent learning path recommendation system of the present invention are as follows: through multimodal data fusion, an efficient user learning willingness evaluation system is established and the learning path is dynamically optimized. The user's emotional text data and voice data are collected, and the voice is converted into text and then processed uniformly, including removing neutral vocabulary, punctuation marks and calculating emotional scores; a positive / negative vocabulary is established, combined with emotion intensity labels, through multi-layer perceptron model training, the emotion intensity is predicted and the willingness evaluation value is calculated in combination with the score. At the same time, the comprehensive willingness evaluation value is further calculated in combination with the user's screen switching frequency and heart rate mutation data. Finally, based on the evaluation value and stress status, a learning path suitable for the user is dynamically recommended, and the content difficulty is flexibly adjusted according to the user's status.

[0047] This application significantly improves the emotion recognition accuracy and recommendation precision of self-learning users on the learning path through the fusion of text, voice and vital sign multimodal data, combined with nonlinear weight adjustment and real-time response capabilities. It uses lightweight models to reduce computational complexity and dynamically optimizes the learning path to meet users' personalized needs, thereby enhancing users' learning experience and sustainable learning status. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of an educational intelligent learning path recommendation system method of the present invention;

[0049] Figure 2 A schematic diagram of an educational intelligent learning path recommendation system of the present invention;

[0050] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application;

[0051] Figure 4 It is a schematic diagram of the structure of a computer-readable storage medium provided by an embodiment of the present application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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. Example 1

[0053] See also Figure 1 As shown, the method for recommending an intelligent learning path for education described in this embodiment includes:

[0054] Collect sentiment text data.

[0055] Emotional text data includes discussions in the discussion area, homework feedback, and comments; specifically, it is obtained in real time by calling the database through the user-side API. The acquisition methods of emotional text data also include: real-time capture of the text window where the user inputs in real time.

[0056] The multi-data source method covers various behaviors of users, which can ensure data integrity and timeliness.

[0057] Preprocess the emotional text data;

[0058] Set up a neutral word bank; the neutral word bank is a set of words with unclear emotional states. Exemplarily, words such as "de", "li", and "le" that have nothing to do with emotional states are included in the neutral word bank.

[0059] The preprocessing process includes: removing the words and punctuation marks belonging to the neutral word bank from the emotional text data;

[0060] Traditional preprocessing usually includes removing punctuation marks. In this application, while removing punctuation marks, the preset neutral word bank is removed, which can greatly reduce the real-time data processing volume, improve the response speed of real-time recommended learning paths, and improve the learning experience of users.

[0061] Set up a positive word bank and a negative word bank; the positive word bank contains words with positive emotional states; the negative word bank contains words with negative emotional states. Exemplarily, the positive word bank includes: "like learning", "have improvement", and "not bad", etc., which are words that are willing to continue learning, indicating good learning acceptance and good learning status; the negative word bank includes: "don't want to learn", "too difficult", and "can't", etc., which indicate poor learning status, weak acceptance, and low sustainability;

[0062] Calculate the emotional score based on the preprocessed emotional text data; traverse the preprocessed emotional text data in sequence; when the preprocessed emotional text data is included in the positive word bank, the emotional score is increased by 1 point; when the preprocessed emotional text data is included in the negative word bank, the emotional score is decreased by 2 points; when it is not included in the positive word bank or the negative word bank, the emotional score is increased by 0.1 point;

[0063] The advantage of such a setting is that it is more sensitive to negative words as a whole. When negative words appear in the user, it means that a strong resistance psychology has occurred at present, and the influence on the score is relatively greater at this time. On the contrary, even if positive words appear in the user, they may still terminate learning due to unexpected situations such as temporary business. Therefore, the positive word bank has a limited increase in the emotional score; at the same time, when the vocabulary does not belong to negative or positive words and is the vocabulary that the neutral word bank fails to exclude, it represents the normal level of learning and does not resist, so the emotional score increases slowly at this time.

[0064] Train a machine learning model to identify the intensity of emotions based on the preprocessed emotional text data;

[0065] The training process of the machine learning model includes:

[0066] Jieba is used to segment the preprocessed emotional text data into phrases, and each phrase is converted into a feature vector. The intensity of the emotion is used as the label corresponding to the feature vector. Each set of feature vectors and the label corresponding to each set of feature vectors are constructed as a sample, and multiple samples are collected to construct a data set; the data set is divided into a training set, a validation set, and a test set, where the training set accounts for 80% of the data set, and the validation set and the test set each account for 10% of the data set;

[0067] Jieba is a popular Chinese phrase segmentation tool; it is used in the field of Chinese natural language processing, such as text mining and sentiment analysis information retrieval;

[0068] Emotional intensity refers to the intensity or intensity of the emotion displayed by the user when expressing emotions. It reflects the strength of a person's emotional response and can be understood as the depth, strength or intensity of the emotion. The more intense the emotional intensity, the more obvious the emotion, such as extreme interest in the learning content;

[0069] The intensity of emotion includes mild, moderate and strong levels, which are represented by one-hot encoding;

[0070] One-hot encoding creates a binary vector for each sentiment intensity:

[0071] Mild: [1, 0, 0];

[0072] Medium: [0, 1, 0];

[0073] Strong: [0, 0, 1];

[0074] The training set is used as the input of the machine learning model, and the machine learning model uses the intensity of emotion corresponding to the real-time preprocessed emotional text data as output; the intensity of emotion corresponding to a set of real-time feature vectors is used as the prediction target, and minimizing the loss function value of the machine learning model is used as the training target; when the loss function value of the machine learning model is less than or equal to the preset target loss value, the training is stopped;

[0075] The machine learning model loss function is the mean square error; the mean square error is calculated by Minimize to train the model; in the loss function is the loss function value, i is the feature vector group number; is the number of eigenvector groups; is the label corresponding to the i-th group of feature vectors, is the intensity of emotion predicted by the i-th group of feature vectors; the machine learning model is a multi-layer perceptron;

[0076] Multilayer Perceptron Hidden Layers The calculation methods include:

[0077] G[ ; In the formula, is the output of the hidden layer; G is the activation function; is the bias term of the hidden layer; is the weight of the hidden layer; is the activation function; is the bias term of the previous layer; is the weight of the previous layer; x is the input training set.

[0078] The multilayer perceptron consists of 4 fully connected layers. The first 3 hidden layers contain 32, 64 and 64 neurons respectively. There is 1 SoftMax classification layer containing 3 neurons. The neurons are connected to each other through weights and activated by the ReLU activation function.

[0079] Calculate the willingness evaluation value based on the output of the machine learning model and the sentiment score; ; Where: is the sentiment score of the ath sentiment text data; is the emotional intensity evaluation value obtained by identifying the a-th emotional text data; is the total number of sentiment text data; is 0.8; The willingness evaluation value can more accurately reflect the user's self-study status, ensuring the comprehensiveness of the evaluation; the higher the emotional score, the greater the emotional intensity, and the greater the willingness evaluation value, which reflects the user's stronger willingness to learn; on the contrary, the emotional score is negative and the lower the emotional score, the greater the emotional intensity, and the smaller the willingness evaluation value; the user is less willing to continue learning, and it is urgent to change the learning path. The three levels of mild, moderate and strong in the emotional intensity evaluation value correspond to 1, 2 and 3 respectively.

[0080] Collect emotional speech data; emotional speech data includes the language spoken by the user when the user is learning autonomously on the user end; specifically, it is acquired in real time by the microphone on the user end;

[0081] Based on the conversion of emotional speech data into text data and marking it as second-class emotional text data; the conversion process is speech-to-text technology; the speech-to-text technology is a conventional technology in the field and will not be repeated here; an exemplary process includes: sound capture, signal preprocessing, feature extraction, decoding through acoustic models and language models, and finally post-processing to generate accurate text data.

[0082] It can be understood that there is no essential difference between the second-class emotional text data and the emotional text data. The second-class emotional text data is converted from the voice data, so the emotional score can also be calculated and the willingness evaluation value can be further calculated;

[0083] The second-class sentiment text data is processed in the same way as the sentiment text data; the same processing process includes: preprocessing, calculating sentiment scores, and also serving as input to the machine learning model, setting labels, and so on; at this time, the data source of the machine learning model includes sentiment text data and second-class sentiment text data; the output of the machine learning model is therefore more accurate; the intention at this time integrates the user's real-time voice and text information, and further improves the user's emotional expression level of the current learning progress.

[0084] This multimodal machine learning model reduces the difficulty of training and improves real-time performance based on the original model, and can accurately recommend learning paths based on the user's current status.

[0085] Collect user learning sign data; user learning sign data includes user screen switching frequency and heart rate mutation value; user screen switching frequency is the number of times the user switches to other apps or software on the user side within one minute; it is obtained by the user side screen time management system through mathematical counting of the number of screen switching; frequent content switching may indicate that the content does not meet the needs or interests of the user; the heart rate mutation value is the difference between the sum of the user's heart rate values ​​in the last 30 seconds of each minute and the sum of the heart rate values ​​in the first 30 seconds; a heart rate mutation value that is too high may indicate that the user feels anxious or stressed, and a heart rate mutation value that is too low may indicate that the user feels bored or distracted;

[0086] Compare the heart rate mutation value with the preset first-level heart rate mutation threshold. When the heart rate mutation value is greater than or equal to the first-level heart rate mutation threshold, mark the user as a high-stress state; set the second-level heart rate mutation threshold to be less than the first-level heart rate mutation threshold. When the heart rate mutation value is less than or equal to the preset second-level heart rate mutation threshold, mark the user as a low-stress state; high-stress state and low-stress state are uniformly classified as stress state;

[0087] The comprehensive willingness evaluation value is calculated based on the willingness evaluation value and the user's screen switching frequency; the comprehensive willingness evaluation value ; Where: is a standardization function, which is used to standardize the comprehensive willingness evaluation value to interval, enhancing sensitivity to extreme values; The frequency of users switching screens; the lower the frequency of users switching screens, the higher the willingness evaluation value, and the higher the comprehensive willingness evaluation value; at this time, the user is more receptive to the current learning content; otherwise, the opposite is true.

[0088] Compare the comprehensive willingness evaluation value with the comprehensive willingness evaluation threshold. When the comprehensive willingness evaluation value is less than or equal to the comprehensive willingness evaluation threshold, start the learning path recommendation plan.

[0089] The learning path recommendation plan includes: when the user is in a low-stress state, priority is given to recommending learning content that is more difficult and has deeper content than the current learning content; when the user is in a high-stress state, priority is given to recommending learning content that is less difficult and has shallower content than the current learning content.

[0090] It is understandable that there are two possible reasons why users are unwilling to continue learning: too difficult or too simple. Therefore, adapting to the user's actual complexity in this way helps to better fit the user's personal habits and interests.

[0091] This application establishes an efficient user learning willingness evaluation system and dynamically optimizes the learning path through multimodal data fusion. Collect the user's emotional text data and voice data, convert the voice into text and process it uniformly, including removing neutral vocabulary, punctuation and calculating the emotional score; establish a positive / negative vocabulary, combine the emotion intensity label, train the multi-layer perceptron model, predict the emotion intensity and calculate the willingness evaluation value based on the score. At the same time, the comprehensive willingness evaluation value is further calculated based on the user's screen switching frequency and heart rate mutation data. Finally, based on the evaluation value and stress status, dynamically recommend a learning path suitable for the user, and the content difficulty is flexibly adjusted according to the user's status.

[0092] This application significantly improves the emotion recognition accuracy and recommendation precision of self-learning users on the learning path through the fusion of text, voice and vital sign multimodal data, combined with nonlinear weight adjustment and real-time response capabilities. It uses lightweight models to reduce computational complexity and dynamically optimizes the learning path to meet users' personalized needs and enhance users' learning experience and sustainable learning status. Example 2

[0093] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides an educational intelligent learning path recommendation system, the system comprising:

[0094] The data collection module collects emotional text data, emotional voice data and user learning vital signs data; the user learning vital signs data includes the user's screen switching frequency and heart rate mutation value;

[0095] The preprocessing module sets up a neutral word library, a positive word library and a negative word library; preprocesses the emotional text data; compares the heart rate mutation value with the preset first-level heart rate mutation threshold, and when the heart rate mutation value is greater than or equal to the first-level heart rate mutation threshold, marks the user as a high-stress state; sets the second-level heart rate mutation threshold to be less than the first-level heart rate mutation threshold, and when the heart rate mutation value is less than or equal to the preset second-level heart rate mutation threshold, marks the user as a low-stress state;

[0096] Model training module, which calculates sentiment scores based on preprocessed sentiment text data and trains a machine learning model to identify the intensity of sentiment;

[0097] A multimodal processing module, based on converting the emotional speech data into text data and marking it as second-class emotional text data; performing the same processing on the second-class emotional text data as the emotional text data;

[0098] The calculation module calculates the willingness evaluation value based on the output of the machine learning model and the sentiment score; and calculates the comprehensive willingness evaluation value based on the willingness evaluation value and the user's screen switching frequency;

[0099] The decision-making module compares the comprehensive willingness evaluation value with the comprehensive willingness evaluation threshold. When the comprehensive willingness evaluation value is less than or equal to the comprehensive willingness evaluation threshold, the learning path recommendation plan is initiated; the learning path recommendation plan includes: when the user is in a low-stress state, priority is given to recommending learning content that is more difficult and has deeper content than the current learning content; when the user is in a high-stress state, priority is given to recommending learning content that is less difficult and has shallower content than the current learning content.

[0100] The modules are connected via wired and / or wireless networks. Example 3

[0101] See also Figure 3 As shown, according to another aspect of the present application, an electronic device 500 for an educational intelligent learning path recommendation system is also provided. The electronic device 500 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, an educational intelligent learning path recommendation system as described above may be executed.

[0102] The method or system according to the embodiment of the present application can also be Figure 3 The electronic device shown in the figure is implemented by the electronic device architecture. Figure 3As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store an educational intelligent learning path recommendation system provided by the present application. Furthermore, the electronic device 500 may also include a user interface 508. Of course, Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown. Example 4

[0103] See also Figure 4 As shown, a computer-readable storage medium 600 according to an embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, an educational intelligent learning path recommendation system according to an embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0104] In addition, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided by the present application. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0105] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired network or a wireless network. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0106] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0107] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0108] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only for an educational intelligent learning path recommendation system. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0109] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0110] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0111] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0112] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for recommending intelligent learning paths for education, characterized in that: include: Collect emotional text data, emotional speech data and user learning sign data; User learning sign data include user screen switching frequency and heart rate mutation value; Establish a neutral word library, a positive word library and a negative word library; pre-process the emotional text data; compare the heart rate mutation value with the preset first-level heart rate mutation threshold, and when the heart rate mutation value is greater than or equal to the first-level heart rate mutation threshold, mark the user as being in a high-stress state; Set the secondary heart rate mutation threshold to be lower than the primary heart rate mutation threshold. When the heart rate mutation value is lower than or equal to the preset secondary heart rate mutation threshold, the user is marked as being in a low stress state. Calculate sentiment scores based on preprocessed sentiment text data and train a machine learning model to identify the intensity of sentiment; Based on the conversion of emotional speech data into text data and marking it as second-class emotional text data; The second-class sentiment text data is processed in the same way as the sentiment text data; The same processing process includes: preprocessing, calculating sentiment scores, using them as input to the machine learning model, and setting labels; Calculate the willingness evaluation value based on the output of the machine learning model and the sentiment score; ; Where: is the sentiment score of the ath sentiment text data; is the emotional intensity evaluation value obtained by identifying the a-th emotional text data; is the total number of sentiment text data; is 0.8; The value is 0.45; the three levels of mild, medium and strong in the emotional intensity evaluation value correspond to 1, 2 and 3 respectively; the comprehensive willingness evaluation value is calculated based on the willingness evaluation value and the user's screen switching frequency; the comprehensive willingness evaluation value ; Where: is a standardization function, which is used to standardize the comprehensive willingness evaluation value to the interval [-1,1]; The screen switching frequency for users; The comprehensive willingness evaluation value is compared with the comprehensive willingness evaluation threshold. When the comprehensive willingness evaluation value is less than or equal to the comprehensive willingness evaluation threshold, the learning path recommendation plan is started. The learning path recommendation plan includes: when the user is in a low-stress state, priority is given to recommending learning content that is more difficult and has deeper content than the current learning content; when the user is in a high-stress state, priority is given to recommending learning content that is less difficult and has shallower content than the current learning content.

2. The method for recommending an intelligent learning path for education according to claim 1, characterized in that: Sentimental text data includes discussion forum speeches, homework feedback and comments; The preprocessing process includes: removing words and punctuation marks belonging to the neutral vocabulary in the sentiment text data; The neutral word library is a collection of words with unclear emotional states; the positive word library contains words with positive emotional states; the negative word library contains words with negative emotional states; the emotional intensity refers to the intensity or intensity of the emotions shown by the user when expressing emotions; the emotional speech data includes the language spoken by the user when the user is learning autonomously on the user side; The user screen switching frequency is the number of times the user switches to other apps or software on the user end within one minute; the heart rate mutation value is the difference between the sum of the user's heart rate values ​​in the last 30 seconds of each minute and the sum of the heart rate values ​​in the first 30 seconds.

3. The method for recommending an intelligent educational learning path according to claim 2, characterized in that: The process of calculating sentiment scores includes: The preprocessed sentiment text data is traversed in sequence; when the preprocessed sentiment text data is included in the positive vocabulary, the sentiment score is +1 point; when the preprocessed sentiment text data is included in the negative vocabulary, the sentiment score is -2 points; when it is not included in the positive vocabulary or the negative vocabulary, the sentiment score is +0.1 points.

4. The method for recommending an intelligent educational learning path according to claim 3, characterized in that: The training process of the machine learning model includes: Jieba is used to segment the preprocessed emotional text data into phrases, and each phrase is converted into a feature vector. The intensity of the emotion is used as the label corresponding to the feature vector. Each set of feature vectors and the label corresponding to each set of feature vectors are constructed as a sample, and multiple samples are collected to construct a data set; the data set is divided into a training set, a validation set, and a test set, where the training set accounts for 80% of the data set, and the validation set and the test set each account for 10% of the data set; The training set is used as the input of the machine learning model, and the machine learning model uses the intensity of emotion corresponding to the real-time preprocessed emotional text data as output; the intensity of emotion corresponding to a set of real-time feature vectors is used as the prediction target, and minimizing the loss function value of the machine learning model is used as the training target; when the loss function value of the machine learning model is less than or equal to the preset target loss value, the training is stopped; The machine learning model loss function is the mean square error; the mean square error is calculated by Minimize to train the model; in the loss function is the loss function value, i is the feature vector group number; is the number of eigenvector groups; is the label corresponding to the i-th group of feature vectors, is the intensity of emotion predicted by the i-th group of feature vectors; the machine learning model is a multi-layer perceptron; Multilayer Perceptron Hidden Layers The calculation methods include: ; In the formula, is the output of the hidden layer; G is the activation function; is the bias term of the hidden layer; is the weight of the hidden layer; is the activation function; is the bias term of the previous layer; is the weight of the previous layer; x is the input training set; The multilayer perceptron consists of 4 fully connected layers. The first 3 hidden layers contain 32, 64 and 64 neurons respectively. There is 1 SoftMax classification layer containing 3 neurons. The neurons are connected to each other through weights and activated by the ReLU activation function.

5. The method for recommending an intelligent learning path for education according to claim 4, characterized in that: The intensity of emotion includes mild, moderate and strong levels, which are represented by one-hot encoding; One-hot encoding creates a binary vector for each sentiment intensity: Mild: [1, 0, 0]; Medium: [0, 1, 0]; Strong: [0, 0, 1].

6. An educational intelligent learning path recommendation system, characterized in that: The data collection module collects emotional text data, emotional voice data and user learning vital signs data; the user learning vital signs data includes the user's screen switching frequency and heart rate mutation value; The preprocessing module sets up a neutral word library, a positive word library and a negative word library; preprocesses the emotional text data; compares the heart rate mutation value with the preset first-level heart rate mutation threshold, and when the heart rate mutation value is greater than or equal to the first-level heart rate mutation threshold, marks the user as being in a high-stress state; Set the secondary heart rate mutation threshold to be lower than the primary heart rate mutation threshold. When the heart rate mutation value is lower than or equal to the preset secondary heart rate mutation threshold, the user is marked as being in a low stress state. Model training module, which calculates sentiment scores based on preprocessed sentiment text data and trains a machine learning model to identify the intensity of sentiment; A multimodal processing module, based on converting emotional speech data into text data and marking it as secondary emotional text data; The second-class sentiment text data is processed in the same way as the sentiment text data; The same processing process includes: preprocessing, calculating sentiment scores, using them as input to the machine learning model, and setting labels; The calculation module calculates the willingness evaluation value based on the output of the machine learning model and the sentiment score; the willingness evaluation value ; Where: is the sentiment score of the ath sentiment text data; is the emotional intensity evaluation value obtained by identifying the a-th emotional text data; is the total number of sentiment text data; is 0.8; The value is 0.45; the three levels of mild, medium and strong in the emotional intensity evaluation value correspond to 1, 2 and 3 respectively; the comprehensive willingness evaluation value is calculated based on the willingness evaluation value and the user's screen switching frequency; the comprehensive willingness evaluation value ; Where: is a standardization function, which is used to standardize the comprehensive willingness evaluation value to the interval [-1,1]; The screen switching frequency for users; The decision-making module compares the comprehensive willingness evaluation value with the comprehensive willingness evaluation threshold. When the comprehensive willingness evaluation value is less than or equal to the comprehensive willingness evaluation threshold, the learning path recommendation plan is initiated; the learning path recommendation plan includes: when the user is in a low-stress state, priority is given to recommending learning content that is more difficult and has deeper content than the current learning content; when the user is in a high-stress state, priority is given to recommending learning content that is less difficult and has shallower content than the current learning content.

7. An electronic device, characterized in that include: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the method for recommending an intelligent educational learning path according to any one of claims 1 to 5 by calling the computer program stored in the memory.

8. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are executed on a computer, the computer is caused to execute an educational intelligent learning path recommendation method as described in any one of claims 1 to 5.

Images