Educational cultural relic digital display and ideological and political content fusion system
By integrating data, extracting features, dynamically updating and filtering content, the problems of inaccurate user learning status assessment and poor content recommendation adaptability in the digital display system of educational cultural relics have been solved, achieving accurate matching of personalized ideological and political education content and improving learning participation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGZHOU COLLEGE OF COMMERCE
- Filing Date
- 2026-03-09
- Publication Date
- 2026-06-12
AI Technical Summary
Existing digital display systems for educational cultural relics suffer from inaccurate assessment of user learning status and poor adaptability of content recommendations in terms of integrating ideological and political content, leading to user frustration and low engagement.
The data acquisition module integrates educational and learning data, the feature extraction module generates sets of learned and unlearned content, the set update module dynamically updates unlearned content, the ability matching module calculates learning levels and retains matching content, the content filtering module removes unmatched content, and the ideological and political adjustment module generates personalized educational content.
It achieves a precise match between educational content and user capabilities, improves learning participation and educational efficiency, and ensures the personalization and continuous optimization of ideological and political education content.
Smart Images
Figure CN122196266A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of smart education technology, and more specifically, this application relates to a system for the digital display of educational cultural relics and the integration of ideological and political content. Background Technology
[0002] As an important vehicle for inheriting historical culture and carrying out ideological and political education, the core value of the digital display system for educational cultural relics lies in presenting the connotation of cultural relics through digital means and integrating ideological and political education content to enhance users' ideological and moral qualities. Traditional systems are usually based on fixed sequences of educational content or simple rule-based recommendation mechanisms, such as displaying content according to the type of cultural relic or a preset theme. This static approach can achieve the functions of knowledge dissemination and ideological guidance to a certain extent.
[0003] However, existing digital display systems for educational artifacts suffer from the following problems in integrating ideological and political education content: These systems typically rely on single-dimensional user data, such as the number of times a user studies or basic scores, for content recommendations, making it impossible to accurately assess a user's actual learning status and ability level. Fixed content sequences or thresholds cannot adapt to the dynamic changes in different users' learning: for example, during periods of fluctuating learning ability, static recommendations may result in content that is too difficult or too easy. Overly difficult content can lead to user frustration and abandonment of learning, while overly easy content fails to effectively improve users' ideological level, further weakening the effectiveness of ideological and political education, ultimately leading to low system efficiency and significantly reduced user participation. Therefore, this paper proposes a system for integrating digital display of educational artifacts with ideological and political education content to address these issues. Summary of the Invention
[0004] To address the aforementioned technical issues, this technical solution provides a system for the digital display of educational cultural relics and the integration of ideological and political content. The solution resolves the problems mentioned in the background section.
[0005] To achieve the above objectives, the technical solution of the present invention is as follows:
[0006] Firstly, this application provides a system for the digital display of educational cultural relics and the integration of ideological and political content, including:
[0007] The data acquisition module is used to acquire a set of educational data related to digital education and cultural relics ideological and political education, as well as a set of learning data for users' historical ideological and political education. The educational data includes educational content and learning difficulty, and the learning data includes learning content and learning scores.
[0008] The feature extraction and set construction module is used to extract features from the educational data set and the learning data set to generate a set of learned content and a set of uneducated content.
[0009] The set update module is used to calculate the dispersion index of the learning score corresponding to each learning content in the set of learned content. If the dispersion index is greater than a preset dispersion threshold, the learning content is included in the set of unlearned content and the set of unlearned content is updated.
[0010] The ability matching module is used to calculate the average score of the user's learning score for each learning content within the sliding window and to classify the user's learning level. Based on the learning level, a predetermined number of educational contents from the set of uneducated contents are retained.
[0011] The content filtering module is used to calculate the level difference between the user's current learning level and the difficulty level of each educational content. If the level difference is greater than a preset level difference threshold, the educational content is removed from the set of unlearned educational content.
[0012] The ideological and political adjustment module is used to generate personalized ideological and political education content for users based on the processed set of uneducated content.
[0013] Secondly, this application provides a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the above-described system for the digital display of educational cultural relics and the integration of ideological and political content.
[0014] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0015] This application calculates the dispersion index of learning scores in the set of learned content through the set update module, and updates the set of unlearned content when the index exceeds a preset threshold, thus solving the problem of insufficient content adaptability during user learning fluctuations.
[0016] This application uses an ability matching module to calculate the average score of learning within a sliding window and divide users into learning levels. Based on the level, a predetermined number of educational contents are retained in the set of uneducated content, thus solving the problem that a fixed content sequence cannot adapt to dynamic learning abilities.
[0017] This application uses a content filtering module to calculate the difference between the user's learning level and the difficulty level of the educational content, and removes mismatched content when the difference exceeds a threshold, filtering out educational content that is significantly mismatched with the user's ability, reducing invalid recommendations, and ensuring that ideological and political education content is in line with the user's current ability level. Attached Figure Description
[0018] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Wherein:
[0019] Figure 1This is a structural block diagram of the system for integrating digital display of educational cultural relics with ideological and political content proposed in this invention;
[0020] Figure 2 This is a flowchart of the system implementation method in this invention;
[0021] Figure 3 This is a data flow diagram of the system in this invention. Detailed Implementation
[0022] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.
[0023] Existing digital display systems for educational artifacts suffer from limitations such as a single judgment dimension and a lack of dynamic feedback and adaptive optimization capabilities. They typically rely on fixed content sequences or simple rule-based recommendation mechanisms, pushing content based solely on a single dimension of user data (such as learning frequency or basic scores). This results in an inability to accurately assess a user's actual learning status and ability level. This static approach cannot adapt to the dynamic changes in different users' learning: during periods of fluctuating learning abilities, content that is too difficult can easily lead to frustration, while content that is too easy fails to effectively improve intellectual development, ultimately resulting in low system efficiency and significantly reduced user engagement.
[0024] To address the aforementioned issues, this application integrates educational and learning data sets through a data acquisition module; generates sets of learned and unlearned content through feature extraction and set construction; dynamically updates the unlearned content set based on the dispersion index of learning scores, identifying content with unstable learning; calculates the average score within a sliding window and classifies users' learning levels, retaining matching content; removes content with significant differences between the user's learning level and the difficulty level of the educational content; and finally, generates personalized ideological and political education content through a political and ideological adjustment module. This application achieves dynamic content optimization and iterative adjustment through multi-module collaboration, solving the problem of poor adaptability of static recommendation mechanisms, ensuring accurate matching between ideological and political education content and user abilities, improving learning participation and educational efficiency, and realizing personalized and continuous optimization of ideological and political education.
[0025] Example 1
[0026] like Figure 1-3 As shown, the structural diagram and specific implementation method of the system for integrating digital display of educational cultural relics with ideological and political content are introduced.
[0027] Regarding data acquisition module 100:
[0028] This is used to acquire a set of educational data related to digital education and cultural relics ideological and political education, as well as a set of learning data for users' historical ideological and political education. The educational data includes educational content and learning difficulty, while the learning data includes learning content and learning scores.
[0029] The acquired educational and learning datasets are cleaned and standardized in format, including removing duplicate data, filling in missing values, and standardizing timestamp formats.
[0030] Verify the integrity and validity of the processed educational and learning datasets, including checking whether the learning scores are within the valid score range and whether the learning difficulty level meets the level standard. If the data does not meet the verification conditions, the corresponding data is reacquired.
[0031] Output the validated educational dataset and learning dataset to the feature extraction and dataset construction module. The educational dataset is stored as key-value pairs of <educational content identifier, educational content, learning difficulty>, and the learning dataset is stored as key-value pairs of <learning content identifier, learning content, learning score>.
[0032] Regarding feature extraction module 200:
[0033] This is used to extract features from the educational content of the educational dataset and the learning content of the learning dataset, generating a set of learned content and a set of uneducated content.
[0034] In one optional implementation, the feature extraction process specifically includes:
[0035] Obtain each piece of educational content from the educational dataset and each piece of learning content from the learning dataset;
[0036] The system associates and matches educational content with learning content to identify educational content that the user has already learned.
[0037] Extract the text information vectors of each educational content and each learning content;
[0038] Based on the extracted text information vector, calculate the dot product of the educational content vector and the learning content vector, and calculate the magnitude of the educational content vector and the magnitude of the learning content vector respectively.
[0039] Divide the dot product by the product of the two moduli to obtain the normalized content similarity.
[0040] The formula for calculating content similarity is:
[0041] ;
[0042] in, Indicates educational content, Indicates the learning content, A vector representing textual information about educational content. A vector representing the textual information of the learning content. The modulus of the vector representing the textual information of educational content, and the modulus of the vector representing the textual information of learning content;
[0043] Educational content with a similarity score greater than or equal to a preset similarity threshold is categorized into the "learned content" set, while educational content with a similarity score less than the similarity threshold is categorized into the "unlearned content" set.
[0044] For example, the content similarity can be normalized using the min-max method, and the preset similarity threshold can be set to 0.75.
[0045] Through the above technical solution, this application solves the problem of inaccurate content classification caused by static content recognition in existing ideological and political education systems. It dynamically matches the content that users have learned and the content they have not learned, which not only prevents learning efficiency from being affected by misclassification, but also provides an accurate foundation for the set of unlearned content. At the same time, the vector similarity calculation mechanism ensures the objectivity and accuracy of content classification, and ensures that the ideological and political education content matches the user's learning history.
[0046] Regarding collection update module 300:
[0047] The dispersion index is used to calculate the learning score of each learning content in the set of learned content. If the dispersion index is greater than the preset dispersion threshold, the learning content is included in the set of unlearned content and the set of unlearned content is updated.
[0048] In one optional implementation, the process of calculating the dispersion index specifically includes:
[0049] Get the learning score for each piece of learning content in the set of learned content;
[0050] The learning scores are statistically analyzed for frequency distribution, and a dispersion index is calculated based on the frequency distribution. The dispersion index is the ratio of the arithmetic mean to the standard deviation of the learning scores.
[0051] For example, the dispersion index can be normalized using the min-max method, and the preset dispersion threshold can be set to 0.6.
[0052] Through the above technical solution, this application solves the problem of inaccurate learning status assessment caused by static content management in existing ideological and political education systems. It addresses the fluctuations in the user's mastery of the learned content, preventing the impact on knowledge consolidation due to ignoring learning instability, and providing a quantitative basis for dynamic updates to the system. At the same time, it accurately identifies the learning content that the user has not mastered firmly through the calculation mechanism of the dispersion index, ensuring that the set of unlearned content is updated synchronously with the user's actual mastery status.
[0053] Regarding the capability matching module 400:
[0054] This is used to calculate the average score of each learning content within the sliding window and classify the user's learning level. Based on the learning level, a predetermined number of educational contents from the set of uneducated contents are retained.
[0055] Get the average score of the user's learning score within the number of swipes in the window, and divide the user into three learning levels: high, medium, and low based on the average score.
[0056] For example, the length of the sliding count window can be 10, that is, to select the user's learning scores in the last 10 times; taking a total score of 100 for a single learning score as an example: high learning level is [85,100], medium learning level is (70,84) and low learning level is [0,70];
[0057] Personalized retention based on learning level:
[0058] When the average score is in a high learning level, all educational content in the uneducated content set is retained.
[0059] When the average score is in the middle learning level, a fixed proportion of the educational content is retained.
[0060] When the average score is in the low learning level, half of the educational content is retained at a fixed proportion.
[0061] For example, the fixed percentage can be 70% of the total number of educational content in the uneducated content set.
[0062] In one optional implementation, the ability matching module further includes sorting the set of uneducated content before acquiring it, specifically including:
[0063] For each piece of educational data in the educational dataset, count the number of times its educational content appears in all the learning content associated with it in the user's learning dataset. At the same time, count the number of times the educational content appears and the corresponding learning score is higher than a preset score threshold. For example, if the total score of a single learning session is 100, the preset score threshold can be 70.
[0064] The occurrence rate is the ratio of the number of occurrences to the total amount of content learned by the user, and the accuracy rate is the ratio of the number of correct answers to the number of occurrences.
[0065] The educational content in the uneducated content set is sorted in descending order of occurrence rate. If the accuracy rate of an educational content is greater than a preset accuracy rate threshold, the educational content is removed from the sorted uneducated content set. For example, the accuracy rate threshold can be 0.6.
[0066] Through the above technical solution, this application solves the problem of content redundancy caused by static content management in existing ideological and political education systems. It dynamically sorts and filters the set of untaught content, which not only prevents users from repeatedly accessing content they have already mastered, thus reducing learning efficiency, but also retains educational content with appropriate challenges for the system. At the same time, through a dual-indicator screening mechanism of occurrence rate and accuracy rate, it ensures that the set of untaught content is accurately matched with the user's actual learning progress and mastery level.
[0067] In one optional implementation, the ability matching module further includes candidate processing of the uneducated content set, specifically including:
[0068] Construct a candidate set, which includes the learning content from the top-m learned content set sorted in descending order of occurrence rate. The learning content is arranged in descending order in the candidate set. For example, the number of the first top-m can be 30%-50% of the learning content in the learned content set.
[0069] When the number of educational content in the uneducated content set does not reach the preset upper limit and each learning content does not appear within the specified time window, learning content is selected from the candidate set in sequence and added to the uneducated content set until the number reaches the upper limit. For example, the specified time window can be set to 1 month. Specifically, the time can be increased or decreased according to the importance of the educational content. The preset upper limit can be set to 10.
[0070] Through the above technical solution, this application solves the problem of insufficient content coverage caused by static content management in existing ideological and political education systems. It addresses the need for dynamic supplementation of uneducated content sets, preventing users from being affected by missing content and maintaining their learning continuity and enthusiasm. It also preserves sufficient content diversity for the system to adapt to different learning stages. At the same time, it prioritizes highly relevant content through a frequency ranking mechanism to ensure that the uneducated content set is accurately matched with the user's learning history and actual needs.
[0071] In an optional implementation, the capability matching module further includes anomaly update processing for the retained set of uneducated content, specifically including:
[0072] Get the learning scores of each educational content and related learning content in the set of uneducated content. Related means that the content similarity between the two is greater than or equal to a preset similarity threshold.
[0073] The dispersion or volatility of learning scores corresponding to each educational content is calculated as an anomaly indicator.
[0074] For example, the degree of dispersion or volatility can be taken as the standard deviation of the learning score, which reflects the degree of dispersion of the data distribution, or variance, which measures the magnitude of data fluctuation, or coefficient of variation, which is the ratio of the standard deviation to the mean, and is used to eliminate the influence of dimensions.
[0075] If the abnormal indicator is greater than the preset abnormal threshold, the educational content is determined to be abnormal, and all educational content that is determined to be abnormal is removed from the set of uneducated content. An updated set of uneducated content is generated. For example, the abnormal indicator is normalized using the min-max method, and the preset abnormal threshold can be 0.6.
[0076] Determine whether the number of educational content in the uneducated content set is zero after the abnormal update processing. If so, the top-n abnormal content after sorting the occurrence rate in descending order will be included as educational content in the uneducated content set. For example, the predetermined top-n number can be 10%-20% of the abnormal content.
[0077] Through the above technical solution, this application solves the problem of inaccurate recommended content caused by abnormal data fluctuations in existing ideological and political education systems. It dynamically filters and removes abnormal data in the set of untaught content, which not only prevents abnormal data from affecting the quality of recommendations, but also ensures the continuity of educational content through a backup content supplementation mechanism. At the same time, the dual guarantee of abnormal indicator threshold judgment and high-occurrence content backfilling ensures the stability and robustness of ideological and political education content recommendations.
[0078] Regarding content filtering module 500:
[0079] This is used to calculate the level difference between the user's current learning level and the difficulty level of each educational content. If the level difference is greater than a preset level difference threshold, the educational content will be removed from the set of unlearned educational content.
[0080] Calculate the level difference value between the user's current learning level and the difficulty level of each educational content. The level difference value is calculated as an absolute value. For example, a low learning level can be 1, a medium learning level can be 2, and a high learning level can be 3. Similarly, a low difficulty level can be 1, a medium difficulty level can be 2, and a high difficulty level can be 3. For example, a preset level difference threshold can be set to 1. When the calculated level difference is greater than 1, it indicates that the difficulty level of the educational content is not suitable for the current user's learning ability.
[0081] Through the above process, this module achieves dynamic content filtering based on level differences, ensuring that the educational content in the uneducated content set is accurately matched with the user's learning ability, and avoiding the inefficiency caused by content that is too difficult or too easy.
[0082] In one optional implementation, the content filtering module further includes a secondary filtering process on the set of uneducated content after the initial filtering, specifically including:
[0083] Acquire learning behavior data corresponding to each user's learning content. The learning behavior data includes learning frequency, learning duration, number of interactions, and learning score.
[0084] The learning frequency and learning duration are normalized separately, and the normalized learning frequency and learning duration are weighted and calculated to obtain the learning motivation.
[0085] Learning motivation The calculation formula is:
[0086] ;
[0087] in, The normalized learning frequency, The normalized learning time , For preset weights, and + Exemplary The value is 0.6. The value is 0.4.
[0088] The average learning score is calculated as the level of mastery, and the number of interactions is counted as the level of engagement.
[0089] The learning enthusiasm, mastery level and interactive participation were normalized, and a preset weight was assigned to each data point. The interest level was calculated by weighted summation.
[0090] ;
[0091] in, To normalize learning motivation, The level of mastery after normalization, For normalized interaction engagement, , , For preset weights, and Exemplary The value is 0.6. The value is 0.1. The value is 0.3.
[0092] If the interest level is less than the preset interest level threshold, it is determined that the user lacks interest in the learning content. All educational content related to the learning content that is determined to lack interest is removed from the retained set of uneducated content. Relevance means that the similarity between the two contents is greater than or equal to the preset similarity threshold. For example, the preset interest level threshold can be 0.6.
[0093] All the normalization processes described above use the min-max normalization method.
[0094] Through the above technical solution, this application solves the problem of insufficient interest matching caused by static content screening in existing ideological and political education systems. It dynamically calculates interest based on users' learning behavior data, which not only prevents users from reducing their learning participation and enthusiasm due to exposure to uninteresting content, but also retains recommendation priority for highly interesting content. At the same time, it balances learning enthusiasm, mastery level and interactive participation through a multi-dimensional weighted aggregation mechanism, ensuring that ideological and political education content is accurately matched with users' real-time interests and learning status.
[0095] Regarding the adjustment of module 600 in ideological and political education:
[0096] Used to generate personalized ideological and political education content for users based on the processed set of uneducated content.
[0097] The educational content in the uneducated content set is sorted according to the user's learning level, and educational content whose difficulty level matches the user's learning level is selected first.
[0098] The selected educational content is grouped and sequenced according to the theme of ideological and political education to form a coherent flow of educational content, and interactive elements and case studies are inserted to enhance the educational effect.
[0099] The generated personalized ideological and political education content is output to the user terminal and presented in a multimedia format, including text, images, videos, and interactive Q&A.
[0100] In one optional implementation, the ideological and political adjustment module further includes an effect verification mechanism, specifically including:
[0101] Obtain facial image data and eye movement trajectory data of users after they have studied personalized ideological and political education content;
[0102] For example, facial images are captured by a front-facing camera at a sampling frequency of 30 frames per second, and eye movement trajectories are recorded by an eye tracker at a sampling frequency of 60 times per second.
[0103] Facial feature points are extracted from facial image data to generate facial feature vectors, and gaze point clustering analysis is performed on eye movement trajectory data to generate eye movement feature vectors.
[0104] Specifically, facial feature point extraction includes locating multiple key feature points such as eyebrows, eyes, and lips, calculating their motion displacement and relative positional relationships, and forming a 128-dimensional facial feature vector; eye movement feature vector generation includes performing DBSCAN clustering on fixation points and extracting 12-dimensional eye movement indicators such as fixation duration, saccade speed, and number of regressions.
[0105] The facial feature vector and eye movement feature vector are input into a pre-trained emotion recognition model, which outputs an emotion state index.
[0106] Specifically, the emotion recognition model adopts a dual-channel convolutional neural network architecture, which includes a facial feature encoding branch and an eye-tracking feature encoding branch. It integrates dual-modal information through a feature-level fusion layer and finally outputs an emotion state index containing three dimensions: focus, identification, and confusion. The value range of each index is [0,1].
[0107] Adjusting the content of ideological and political education based on emotional state indicators;
[0108] For example, when the focus level is below 0.6 and the confusion level is above 0.7, it is judged as difficult to understand, and the content difficulty is automatically reduced and example analysis is added; when the agreement level is consistently below 0.5, the value orientation expression is adjusted to enhance the appeal of the case.
[0109] Through the above technical solutions, this application constructs a complete closed-loop optimization system for ideological and political education, which can intelligently adjust educational strategies based on users' real-time emotional feedback: providing timely cognitive support when confused expressions and scattered eye movement patterns are detected, and appropriately increasing the depth of content when an active and focused state is identified, thereby realizing the transformation of ideological and political education from "one-way indoctrination" to "two-way interaction" and improving the accuracy and effectiveness of education.
[0110] Specifically, the emotion recognition model includes a facial feature extraction module, an eye-tracking feature extraction module, a multimodal fusion module, and an emotion state output module.
[0111] The input layer of the facial feature extraction module receives facial image data and extracts facial feature vectors through a pre-trained convolutional neural network. The convolutional neural network adopts the VGG-16 architecture and adds a global average pooling layer before the last fully connected layer to reduce the number of parameters.
[0112] The input layer of the eye-tracking feature extraction module receives eye-tracking trajectory data, extracts time-series features through a bidirectional long short-term memory network, and outputs an eye-tracking feature vector. The hidden layer dimension of the bidirectional long short-term memory network is set to 128, and a Dropout mechanism (dropout rate of 0.3) is introduced to prevent overfitting.
[0113] The multimodal fusion module concatenates facial feature vectors and eye-tracking feature vectors, and performs weighted fusion through a cross-modal attention mechanism. The attention weights are calculated using scaled dot product attention, and the output is a fused feature vector.
[0114] The emotional state output module maps the fused feature vector to emotional state indicators through a three-layer fully connected network. The activation function of the fully connected layer is ReLU, and the output layer uses the Sigmoid function to normalize and generate focus index, confusion index and emotional resonance index, which are used as emotional state indicators.
[0115] The training process of the emotion recognition model includes: using public emotion datasets (such as AffectNet) and a self-built eye-tracking dataset, where the data annotations include multi-label scores of focus, confusion, and emotional resonance; using an end-to-end training framework, the error between the predicted value and the true label is calculated using the mean squared error loss function, the optimizer is Adam, the initial learning rate is set to 0.001, and a learning rate decay strategy is adopted (decaying to 0.5 every 10 epochs); an early stopping mechanism (patience value set to 15) is introduced during training to prevent overfitting, the batch size is set to 32, and the maximum number of training epochs is 100.
[0116] Through the above technical solution, this application solves the problem of insufficient personalization caused by static content adjustment in existing ideological and political education systems. It collects facial images and eye movement trajectory data in real time to address changes in the user's emotional state during the learning process. This not only prevents users from becoming confused or giving up learning due to overly difficult content, but also enhances the depth and interactivity of the content for emotionally positive users. At the same time, it dynamically adjusts the education strategy through a multimodal emotion recognition model to ensure that the ideological and political education content matches the user's real-time emotional feedback.
[0117] In one alternative implementation, the process of generating eye-tracking feature vectors specifically includes:
[0118] Eye-tracking feature vector generation process:
[0119] Extract the gaze point coordinate sequence from eye movement trajectory data and identify the gaze region based on a density clustering algorithm;
[0120] Specifically, the DBSCAN clustering algorithm is used, with a neighborhood radius of 50 pixels and a minimum number of samples of 3, to cluster the gaze points into different regions of interest.
[0121] Calculate the ratio of dwell time to area for each gaze region to generate region attention level;
[0122] For example, for each identified gaze region, the ratio of total gaze duration to the region's bounding box area is calculated to form a 5-dimensional attention index;
[0123] Extract the coordinate data and corresponding timestamp of each gaze point in the gaze point coordinate sequence;
[0124] The Euclidean distance between adjacent fixation points is calculated as the saccade distance based on their coordinate data. The saccade speed is then calculated based on the saccade distance and the time difference between adjacent fixation points.
[0125] The calculated saccade distance sequence and saccade velocity sequence are combined to generate eye movement activity.
[0126] Specifically, statistical features such as mean, peak value, and coefficient of variation are extracted from the saccade distance and velocity sequences to form an 18-dimensional activity index;
[0127] The region attention and eye movement activity are combined to form an eye movement feature vector;
[0128] For example, a 23-dimensional eye-tracking feature vector is ultimately generated;
[0129] The process of generating facial feature vectors specifically includes:
[0130] Extract the coordinate sequence of facial feature points from facial image data, and identify key facial regions based on a facial feature detection model;
[0131] For example, a deep learning-based facial feature detection model is used to identify seven key facial regions, including eyebrows, eyelids, and corners of the mouth, and output a coordinate sequence containing 76 feature points.
[0132] Calculate the amplitude of muscle movement in each key facial region and the offset from the reference position to generate an expression intensity index;
[0133] Specifically, the reference position is the feature point position when the user has a neutral expression. By calculating the average Euclidean distance between the feature points in the current frame and the reference position, a 7-dimensional expression intensity index is generated.
[0134] Extract the motion trajectory data and corresponding timestamp of each feature point in the facial feature point coordinate sequence;
[0135] Based on the facial feature point coordinate data of adjacent time frames, the change in position of its motion trajectory data is calculated as the facial micro-motion distance. Based on the facial micro-motion distance and the time difference between adjacent time frames, the facial micro-motion speed is calculated.
[0136] For example, the time frame interval is 33ms, and the distance of facial micro-motion is calculated using the L2 norm of the feature point coordinates;
[0137] The calculated facial micro-motion distance sequence and facial micro-motion velocity sequence are combined to generate facial dynamic feature indicators;
[0138] Specifically, statistical features such as mean, variance, and extreme values are extracted from the distance and velocity sequences of facial micro-movements to form a 24-dimensional dynamic feature index.
[0139] The facial expression intensity index and facial dynamic feature index are combined to form a facial feature vector;
[0140] For example, a 31-dimensional facial feature vector is ultimately generated;
[0141] Through the above technical solution, this application solves the problem of inaccurate user attention assessment caused by the coarse analysis of eye-tracking data in existing ideological and political education systems. It dynamically generates eye-tracking feature vectors based on the user's gaze trajectory and saccade behavior, which not only prevents excessive cognitive load or decreased participation caused by the mismatch between content and attention patterns, but also provides multi-dimensional eye-tracking indicators to support emotion recognition. At the same time, through the fusion mechanism of regional attention and eye-tracking activity, it ensures that eye-tracking features can comprehensively reflect the user's real-time attention status and cognitive engagement, thereby improving the accuracy and adaptability of ideological and political education content adjustments.
[0142] Example 2
[0143] This application also provides a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the educational cultural relics digital display and ideological and political content integration system as described above.
[0144] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.
Claims
1. A system for integrating digital display of educational cultural relics with ideological and political content, characterized in that: include: The data acquisition module is used to acquire educational data sets related to digital education and cultural relics ideological and political education, as well as learning data sets of users' historical ideological and political education. Among them, the educational data includes educational content and learning difficulty, and the learning data includes learning content and learning scores. The feature extraction module is used to extract features from the educational data set and the learning data set to generate a set of learned content and a set of uneducated content. The set update module is used to calculate the dispersion index of the learning score corresponding to each learning content in the set of learned content. If the dispersion index is greater than a preset dispersion threshold, the learning content is included in the set of unlearned content and the set of unlearned content is updated. The ability matching module is used to calculate the average score of the user's learning score for each learning content within the sliding window and to classify the user's learning level. Based on the learning level, a predetermined number of educational contents from the set of uneducated contents are retained. The content filtering module is used to calculate the level difference between the user's current learning level and the learning difficulty level of each educational content. If the level difference is greater than a preset level difference threshold, the educational content is removed from the set of unlearned educational content. The ideological and political adjustment module is used to generate personalized ideological and political education content for users based on the processed set of uneducated content.
2. The system for digital display of educational cultural relics and integration of ideological and political content according to claim 1, characterized in that, The feature extraction process in the feature extraction module specifically includes: Obtain each piece of educational content from the educational dataset and each piece of learning content from the learning dataset; Extract the text information vectors of each educational content and each learning content; Based on the extracted text information vectors, calculate the dot product of the text information vectors of the educational content and the learning content, and calculate the magnitude of the text information vectors of the educational content and the learning content respectively. Divide the dot product by the product of the moduli to obtain the normalized content similarity. Educational content with a content similarity greater than or equal to a preset similarity threshold is categorized into the set of already learned content, and educational content with a content similarity less than the preset similarity threshold is categorized into the set of unlearned content.
3. The system for digital display of educational cultural relics and integration of ideological and political content according to claim 1, characterized in that, The process of calculating the dispersion index in the set update module specifically includes: Get the learning score for each piece of learning content in the set of learned content; The learning scores are statistically analyzed for frequency distribution, and a dispersion index is calculated based on the frequency distribution results. The dispersion index is the ratio of the arithmetic mean to the standard deviation of the learning scores.
4. The system for digital display of educational cultural relics and integration of ideological and political content according to claim 1, characterized in that, Before acquiring the set of uneducated content, the capability matching module also includes sorting it, specifically including: For each piece of educational data in the educational dataset, count the number of times its educational content appears in all learning content in the user's learning dataset, and also count the number of times that educational content appears and the corresponding learning score is higher than a preset score threshold. The occurrence rate is the ratio of the number of occurrences to the total number of all learning content for the user, and the accuracy rate is the ratio of the number of correct answers to the number of occurrences. The educational content in the uneducated content set is sorted in descending order of occurrence rate. If the accuracy rate of an educational content is greater than a preset accuracy rate threshold, the educational content is removed from the sorted uneducated content set.
5. The system for digital display of educational cultural relics and integration of ideological and political content according to claim 4, characterized in that, The capability matching module also includes candidate processing for the set of untaught content, specifically including: Construct a candidate set, which includes the learning content in the Top-m set of learned content sorted in descending order of occurrence rate, and the learning content in the candidate set is sorted in descending order of occurrence rate; When the number of educational content in the uneducated content set has not reached the preset upper limit and no learning content in the candidate set appears within the specified time window, learning content is selected from the candidate set in sequence and added to the uneducated content set until the upper limit is reached.
6. The system for digital display of educational cultural relics and integration of ideological and political content according to claim 5, characterized in that, The capability matching module also includes anomaly handling for the set of uneducated content after candidate processing, specifically including: Get the learning scores for each educational content and the learning content related to each educational content in the set of uneducated content; The dispersion or volatility of learning scores corresponding to each educational content is calculated as an anomaly indicator. If the abnormal indicator is greater than the preset abnormal threshold, the educational content is determined to be abnormal, and all educational content determined to be abnormal is removed from the set of uneducated content, generating a set of uneducated content after abnormal update processing. Determine whether the number of educational content in the uneducated content set is zero after the anomaly update. If so, add the top-n anomaly content in descending order of occurrence rate as educational content to the uneducated content set.
7. The system for digital display of educational cultural relics and integration of ideological and political content according to claim 1, characterized in that, The content filtering module also includes a secondary filtering process on the set of uneducated content after the initial filtering. Specifically, this includes: Acquire learning behavior data corresponding to each user's learning content, including learning frequency, learning duration, number of interactions, and learning score; The learning frequency and learning duration are normalized respectively, and the normalized learning frequency and learning duration are weighted and calculated to obtain the learning motivation. The average of the learning scores is calculated as the level of mastery, and the number of interactions is counted as the level of interaction participation. The learning enthusiasm, mastery level and interactive participation are normalized, and a preset weight is assigned to each data point. The interest level is obtained by weighted summation. If the interest level is less than a preset interest level threshold, it is determined that the user lacks interest in the learning content, and all educational content related to the learning content that has been determined to lack interest is removed from the set of uneducational content.
8. The system for digital display of educational cultural relics and integration of ideological and political content according to claim 1, characterized in that, The ideological and political adjustment module also includes effect verification, specifically including: Obtain facial image data and eye movement trajectory data of users after they have studied personalized ideological and political education content; Facial feature points are extracted from the facial image data to generate a facial feature vector, and gaze point clustering analysis is performed on the eye movement trajectory data to generate an eye movement feature vector. The facial feature vector and eye movement feature vector are input into a pre-trained emotion recognition model, which outputs an emotion state index. Personalized ideological and political education content is adjusted based on the aforementioned emotional state indicators.
9. The system for digital display of educational cultural relics and integration of ideological and political content according to claim 8, characterized in that, The process of generating eye-tracking feature vectors specifically includes: Extract the gaze point coordinate sequence from eye movement trajectory data and identify the gaze region based on a density clustering algorithm; Calculate the ratio of dwell time to area for each gaze region to generate region attention level; Extract the coordinate data and corresponding timestamp of each gaze point in the gaze point coordinate sequence; Based on the coordinate data of adjacent fixation points, the Euclidean distance is calculated as the saccade distance, and the saccade speed is calculated based on the saccade distance and the time difference between the timestamps of adjacent fixation points. The calculated saccade distance sequence and saccade velocity sequence are combined to generate eye movement activity. The region attention and eye movement activity are combined to form an eye movement feature vector.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the system for digital display of educational cultural relics and integration of ideological and political content as described in any one of claims 1-9.