A popular science quantitative intelligent scoring system

Abstract

The application belongs to the technical field of scientific education, and discloses a popular science quantitative intelligent scoring system; the system comprises a data collection module, which is used for acquiring popular science activity data; an intelligent recommendation module, which is used for recommending different popular science contents to participants according to the popular science activity data; a popular science effect evaluation module, which is used for evaluating whether the participants achieve positive influence after participating in different popular science contents; and a popular science quantitative scoring module, which is used for scoring the positive influence achieved by the participants after participating in different popular science contents, so as to provide an effective basis for the examination of popular science work, help improve the quality and efficiency of the popular science work, and help improve the effect of the popular science activities and the satisfaction of the participants.

Description

Technical Field

[0001] This invention relates to the field of science popularization service technology, and more specifically, to a quantitative intelligent scoring system for science popularization. Background Technology

[0002] With the rapid development of science and technology, popular science work is playing an increasingly important role in disseminating scientific knowledge and improving the public's scientific literacy. However, there is currently no unified quantitative assessment and scoring system for popular science work in China, which to some extent affects the effective implementation and evaluation of popular science work. Popular science work is an important requirement of the Healthy China 2030 Plan. Establishing a fair and just quantitative assessment system, phased policy guidance, and recommendations for individual development are of great significance for improving the quality and efficiency of popular science work.

[0003] Patent application CN111709773A discloses a cloud-based personalized science content recommendation system, including a cloud computing server and an intelligent playback front-end. Through the cooperation of a search playback module, a history browsing information recording module, and an auto-fill module, it intelligently predicts and automatically fills in the keywords searched by the user, achieving intelligent recommendation of science content and facilitating ease of use. The system uses a status parameter detection module and a map module to determine user behavior tags, enabling accurate identification of user behavior. The behavior analysis module, based on search information and user behavior tags, filters out science files that contain search information but have not been viewed, categorizing these files into video, audio, and article types. By coordinating with a noise detection unit, it determines the most suitable science file type for each user behavior tag based on the current user's usage pattern. This ensures a good browsing experience while improving security during use.

[0004] The lack of consideration for quantitative intelligent scoring in science popularization provides an effective basis for evaluating science popularization work, which is detrimental to improving the quality and efficiency of science popularization work; it cannot intelligently recommend suitable science popularization directions based on the individual strengths and backgrounds of participants, which is detrimental to leveraging the expertise of science popularization workers; it is detrimental to enhancing the innovation and effectiveness of science popularization work; and it is detrimental to promoting the healthy development of science popularization work and improving public scientific literacy.

[0005] In view of this, the present invention proposes a quantitative intelligent scoring system for popular science to solve the above problems. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a popular science quantitative intelligent scoring system, characterized in that it comprises:

[0007] The data collection module is used to acquire data from science popularization activities;

[0008] The intelligent recommendation module is used to recommend different science popularization content to participants based on science popularization activity data;

[0009] The science popularization effect evaluation module is used to evaluate whether participants achieve a positive impact after engaging with different science popularization content;

[0010] The science popularization quantitative scoring module is used to score the positive impact achieved by participants after engaging in different science popularization content.

[0011] The modules are connected via wired or wireless means to enable data transmission between them.

[0012] Furthermore, the data on popular science activities include participation in popular science activities, participant feedback satisfaction, participants' knowledge gain rate, and the social impact of popular science activities;

[0013] Furthermore, the method for obtaining the data from the science popularization activities includes:

[0014] Participation in science popularization activities includes the number of participants, the duration of participation, and the number of times participants attend. At the entrance of the science popularization activity venue, participants are given wristbands with RFID tags. Participants participate in the science popularization activities by wearing the wristbands with RFID tags. The number of participants, the duration of participation, and the number of times participants attend are recorded by reading the wristbands with RFID tags using a card reader.

[0015] The participation rate in the aforementioned science popularization activities is: ;in, To increase participation in science popularization activities; The number of people participating in science popularization activities; The duration of participation in science popularization activities; The number of times you participate in interactive science popularization activities; , and A factor influencing participation in science popularization activities;

[0016] Staff distributed questionnaires to participants and recorded their feedback and satisfaction scores, which were the participants' ratings of the science popularization activities.

[0017] The method for obtaining the knowledge gain rate of the participants includes:

[0018] Based on the stored science popularization content database, one or more sets of science popularization knowledge test questions related to the science popularization content are randomly generated. The test scores of participants before and after participating in science popularization activities are recorded through the scoring system.

[0019] The knowledge gain rate of the participants is: ;in, The test score of participants after participating in science popularization activities; The test score of the participants before they participated in the science popularization activities;

[0020] The social impact of science popularization activities includes the scope of dissemination and media attention; the scope of dissemination and media attention of science popularization activities are tracked and recorded through a news monitoring platform;

[0021] The social impact of the aforementioned science popularization activities is as follows: ;in, As an indicator of social influence; To broaden the reach of science popularization activities; For media attention; and These are the influencing factors for the social influence indicator.

[0022] Furthermore, the media attention mentioned includes:

[0023] The media attention mentioned is: ;in, The number of media reports; The duration of media exposure.

[0024] Furthermore, the method of recommending different science popularization content to participants based on science popularization activity data includes:

[0025] The acquired science popularization activity data were normalized to standard deviation and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1 to eliminate the influence of dimensions between data. Finally, the normalized feature dataset was obtained.

[0026] A participant interest prediction model is constructed. The input of the participant interest prediction model is a historical feature dataset; the output label is the participant's preference for different popular science content.

[0027] By combining content recommendation algorithms, different popular science content is recommended to participants based on their preferences for different popular science content and the characteristics of the popular science content.

[0028] Using collaborative filtering algorithms, popular science content that matches the interests of participants is recommended by analyzing the behavioral patterns and interest similarities between participants and other participants.

[0029] By combining machine learning and deep learning technologies, the participant interest prediction model is continuously updated and optimized. This is achieved by constantly collecting and analyzing participants' science popularization activity data.

[0030] Furthermore, the method for pre-building the participant interest prediction model includes:

[0031] Prepare the training dataset, including input data and corresponding labels; the input data is a historical feature dataset, and the output labels are the participants' preferences for different popular science content.

[0032] Define the structure of the multilayer perceptron (MLP) model, including an input layer, hidden layers, and an output layer; the size of the input layer matches the dimension of the feature vector of the science popularization activity; the number and size of the hidden layers are selected according to the task requirements and data complexity; the size of the output layer matches the dimension of the participants' attention preferences for different science popularization content.

[0033] Initialize the weights and biases of the multilayer perceptron (MLP) model using a random initialization method;

[0034] For the weights of each layer: determine the dimensions of the weight matrix, with an input dimension of n and an output dimension of [missing value]; select the range [missing value]. ,in, It is a positive number representing the range of values ​​for the weights; values ​​are randomly generated from a uniform distribution as initial weights, ensuring that these values ​​are within the selected range;

[0035] For the bias at each layer: determine the dimension of the bias vector, and the output dimension is... ; Select range ,in, A positive number representing the range of bias values; values ​​are randomly generated from a uniform distribution as the initial bias, ensuring that these values ​​are within the selected range;

[0036] The feature vectors of the training data are input into the MLP model; the output values ​​of each layer are calculated using the weights and biases of each layer, and this process is continued until the output layer is reached; a non-linear activation function is applied to the output of the hidden layer to transform the output of the hidden layer into a non-linear form;

[0037] The loss function is calculated using the prediction results and label data from the output layer to measure the error of the model's predictions.

[0038] Using mean squared error as the loss function, the formula for calculating the mean squared error loss function is as follows: ;in, This reflects the actual participants' preferences and interests regarding different science popularization content. To predict participants' preferences for different popular science content;

[0039] The backpropagation algorithm is used to calculate the gradient of each parameter with respect to the loss function, thereby updating the parameters to reduce the value of the loss function; the stochastic gradient descent algorithm is used to update the model parameters according to the gradient, so that the loss function gradually decreases.

[0040] The learning rate is adjusted to control the speed and stability of parameter updates, thereby optimizing the multilayer perceptron (MLP) model; the Adam algorithm is selected as the optimizer; the Adam algorithm is based on optimizing stochastic gradient descent and adaptively adjusts the learning rate;

[0041] After the training reaches the predetermined number of iterations, the final trained participant interest prediction model is obtained. The trained participant interest prediction model is then used to predict the current feature dataset to obtain the participants' attention preferences for different popular science content.

[0042] Furthermore, the method of combining content recommendation algorithms to recommend different popular science content to participants based on their preferences for different popular science content and the characteristics of the popular science content includes:

[0043] Collect participants' preferences for different popular science content and the characteristics of popular science content, transform them into computable numerical features, encode the preferences for different popular science content and popular science activity data, and use one-hot encoding to transform them into vector representations;

[0044] Determine the feature columns that need one-hot encoding, collect and deduplicate the values ​​in the feature columns to obtain all possible values; for each feature value of the attention preference of different popular science content, iterate through all possible values; if the feature value of the attention preference of different popular science content is equal to the currently traversed value, then the corresponding position is 1, otherwise it is 0; for each feature value of different popular science content, also iterate through all possible values, if the feature value of different popular science content is equal to the currently traversed value, then the corresponding position is 1, otherwise it is 0;

[0045] For each different feature of popular science content, the results of its one-hot encoding are arranged into a vector in order; for each feature of different popular science content, the results of its one-hot encoding are also arranged into a vector in order; the length of this vector is equal to the number of all possible values, with only one position being 1 and the rest being 0;

[0046] By calculating the similarity between the participant's preference vector and the feature vector of each piece of science content, the most similar science content is found and recommended to the participant; similarity is calculated using cosine similarity. ;in, For participants Attention Preference Vector and popular science content eigenvectors The dot product; For vectors The norm; For vectors The norm;

[0047] Participants focused on the similarity between preference vectors and feature vectors of each piece of science popularization content. Values ​​in Between; if participants focus on the similarity between the preference vector and the feature vector of each popular science content. The closer to 1, the more similar they are; the closer to 0, the more unrelated they are; and the closer to -1, the opposite they are.

[0048] By calculating the similarity between participants' attention preference vectors for different science popularization content and the feature vectors of the science popularization content, different science popularization content is recommended to different participants.

[0049] Furthermore, the method of using collaborative filtering algorithms to recommend popular science content that matches the interests of participants by analyzing their behavioral patterns and interest similarities with other participants includes:

[0050] Acquire participant behavior data, which includes participants' browsing history, favorites, and purchased popular science books; participants' browsing history, favorites, and purchased popular science books constitute projects; for participants and participants' browsing history, favorites, and purchased popular science books, they are represented as a participant-project matrix, where rows represent participants, columns represent projects, and elements in the matrix represent the interaction behavior between participants and projects.

[0051] The Pearson correlation coefficient is used as a similarity measure to calculate the similarity matrix among participants and identify participants with high similarity. For the participant-item matrix, the similarity among participants is calculated. The formula for calculating the Pearson correlation coefficient is as follows: ;in, and The interaction behavior of two participants on different projects; and The average value of the interaction behaviors of the two participants on different projects;

[0052] Based on the similarity matrix, other participants are ranked to find those with high similarity to the target participant. A recommendation list is generated based on the interaction behavior of similar participants. For participants similar to the target participant, the popular science content they are interested in is analyzed and recommended to the target participant. Content that has already been viewed or saved is excluded according to filtering rules, and the recommendation list is further filtered to generate the final recommendation result. The features of the popular science content are extracted and transformed into numerical feature vectors using the bag-of-words model.

[0053] Furthermore, the methods for assessing whether participants achieve a positive impact after engaging with different science popularization content include:

[0054] The scores of participants after engaging in science popularization activities are compared with a preset score threshold for participants who do not participate in science popularization activities; the scores of participants after engaging in science popularization activities are recorded through questionnaires.

[0055] If a participant's score after participating in a science popularization activity is greater than the preset score threshold for participants who do not participate in science popularization activities, then the participant is deemed to have achieved a positive impact.

[0056] If a participant's score after participating in a science popularization activity is less than or equal to the preset score threshold for participants who do not participate in science popularization activities, then the participant is deemed not to have achieved a positive impact.

[0057] Furthermore, methods for scoring based on the positive impact achieved by participants engaging with different science popularization content include:

[0058] The importance of each data point in the science popularization activity data is quantified into a weight, which is expressed as a percentage to ensure that the sum of all weights equals 100%. For each participant, a comprehensive score is calculated based on the score of each data point and its corresponding weight. The score of each data point is obtained by the participant filling out a science popularization activity data question provided by the staff.

[0059] The formula for calculating the comprehensive score is as follows: ;in, This serves as the identifier for each data point in the science popularization activity data; The number of each data point in the science popularization activity data; A score for each data point in the science popularization activity data; The weight assigned to each data point in the science popularization activity data;

[0060] The technical effects and advantages of the present invention, a quantitative intelligent scoring system for popular science:

[0061] 1. The popular science quantitative scoring system provided by this invention can provide an effective basis for the assessment of popular science work, and help improve the quality and efficiency of popular science work;

[0062] 2. The intelligent recommendation module in this invention can recommend different science popularization content to participants based on science popularization activity data, which helps to give full play to the expertise of science popularization workers;

[0063] 3. The science popularization effect evaluation module in this invention is used to evaluate whether participants achieve a positive impact after engaging in different science popularization content; it provides science popularization workers with rich reference resources and helps to improve the innovation and effectiveness of science popularization work;

[0064] 4. The science popularization scoring module in this invention is used to score the positive impact achieved by participants after engaging in different science popularization content. For each participant, a comprehensive score can be calculated based on their scores on various evaluation indicators and their corresponding weights, which helps to improve the effectiveness of science popularization activities and the satisfaction of participants. Attached Figure Description

[0065] Figure 1 This is a module diagram of a popular science quantitative intelligent scoring system according to the present invention;

[0066] Figure 2 This is a flowchart of a popular science quantitative intelligent scoring method according to the present invention; Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Example 1

[0069] Please see Figure 1 As shown in this embodiment, a popular science quantitative intelligent scoring system includes:

[0070] The data collection module is used to acquire data from science popularization activities;

[0071] The intelligent recommendation module is used to recommend different science popularization content to participants based on science popularization activity data;

[0072] The science popularization effect evaluation module is used to evaluate the positive impact achieved by participants after engaging in different science popularization content;

[0073] The science popularization quantitative scoring module is used to score the positive impact achieved by participants after engaging in different science popularization content.

[0074] The modules are connected via wired or wireless means to enable data transmission between them.

[0075] The data on popular science activities include participation in popular science activities, participant feedback satisfaction, participants' knowledge gain rate, and the social impact of popular science activities.

[0076] The methods for obtaining data from the science popularization activities include:

[0077] Participation in science popularization activities includes the number of participants, the duration of participation, and the number of times participants attend. At the entrance of the science popularization activity venue, participants are given wristbands with RFID tags. Participants participate in the science popularization activities by wearing the wristbands with RFID tags. The number of participants, the duration of participation, and the number of times participants attend are recorded by reading the wristbands with RFID tags using a card reader.

[0078] The participation rate in the aforementioned science popularization activities is: ;in, To increase participation in science popularization activities; The number of people participating in science popularization activities; The duration of participation in science popularization activities; The number of times you participate in interactive science popularization activities; , and A factor influencing participation in science popularization activities;

[0079] Staff distributed questionnaires to participants and recorded their feedback satisfaction, which was a score given by participants to the science popularization activity. Participant feedback satisfaction could be set to four levels: excellent, good, average, and poor. The feedback satisfaction level was set by preset first, second, and third feedback satisfaction thresholds: the first feedback satisfaction threshold was 70 points; the second feedback satisfaction threshold was 80 points; and the third feedback satisfaction threshold was 90 points.

[0080] If a participant's feedback satisfaction is less than or equal to a preset first feedback satisfaction threshold, then the participant's feedback satisfaction level will be set to poor.

[0081] If a participant's feedback satisfaction level is greater than the preset first feedback satisfaction level and less than or equal to the second feedback satisfaction level, then the participant's feedback satisfaction level will be set to medium.

[0082] If a participant's feedback satisfaction level is greater than the preset second feedback satisfaction threshold and less than or equal to the third feedback satisfaction threshold, then the participant's feedback satisfaction level will be set to "good".

[0083] If a participant's feedback satisfaction level is greater than the preset third feedback satisfaction threshold, then the participant's feedback satisfaction level will be set to excellent.

[0084] The preset first, second, and third feedback satisfaction thresholds are set by staff. Participants' feedback satisfaction is recorded through questionnaires. Multiple groups of participants' feedback satisfaction are collected, and the average of the feedback satisfaction of multiple participants is taken as the preset first, second, and third feedback satisfaction thresholds. Similarly, a preset score threshold for participants who do not participate in science popularization activities is set.

[0085] The method for obtaining the knowledge gain rate of the participants includes:

[0086] Based on the stored science popularization content database, one or more sets of science popularization knowledge test questions related to the science popularization content are randomly generated. The test scores of participants before and after participating in science popularization activities are recorded through the scoring system.

[0087] The knowledge gain rate of the participants is: ;in, The test score of participants after participating in science popularization activities; The test score of the participants before they participated in the science popularization activities;

[0088] The social impact of science popularization activities includes the scope of dissemination and media attention; the scope of dissemination and media attention of science popularization activities are tracked and recorded through a news monitoring platform;

[0089] The social impact of the aforementioned science popularization activities is as follows: ;in, As an indicator of social influence; To broaden the reach of science popularization activities; For media attention; and Influencing factors for social influence indicators;

[0090] The media attention mentioned includes:

[0091] The media attention mentioned is: ;in, The number of media reports; Duration of media exposure;

[0092] The method of recommending different science popularization content to participants based on science popularization activity data includes:

[0093] The participation rate, feedback satisfaction, knowledge gain rate, and social impact of science popularization activities in the acquired science popularization activity data are normalized by standard deviation, and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1 to eliminate the influence of dimensions between data. Finally, the normalized feature dataset is obtained.

[0094] The feature dataset is as follows:

[0095] ;

[0096] A participant interest prediction model is constructed. The input of the participant interest prediction model is a standardized feature dataset; the output label is the participant's preference for different popular science content.

[0097] By combining content recommendation algorithms, different popular science content is recommended to participants based on their preferences for different popular science content and the characteristics of the popular science content.

[0098] Using collaborative filtering algorithms, the system quickly recommends popular science content that matches the interests of participants by analyzing their behavioral patterns and interest similarities with other participants.

[0099] By combining machine learning and deep learning technologies, the participant interest prediction model is continuously updated and optimized. This is achieved by constantly collecting and analyzing participants' science popularization activity data.

[0100] The method for pre-building the participant interest prediction model includes:

[0101] Prepare the training dataset, including input data and corresponding labels; the input data is a historical feature dataset, and the output labels are the participants' preferences for different popular science content.

[0102] Define the structure of the multilayer perceptron (MLP) model, including an input layer, hidden layers, and an output layer; the size of the input layer matches the dimension of the feature vector of the science popularization activity; the number and size of the hidden layers are selected according to the task requirements and data complexity; the size of the output layer matches the dimension of the participants' attention preferences for different science popularization content.

[0103] Initialize the weights and biases of the multilayer perceptron (MLP) model using a random initialization method;

[0104] For the weights of each layer: determine the dimensions of the weight matrix, with an input dimension of n and an output dimension of [missing value]; select the range [missing value]. ,in, It is a positive number representing the range of values ​​for the weights; values ​​are randomly generated from a uniform distribution as initial weights, ensuring that these values ​​are within the selected range;

[0105] For the bias at each layer: determine the dimension of the bias vector, and the output dimension is... ; Select range ,in, A positive number representing the range of bias values; values ​​are randomly generated from a uniform distribution as the initial bias, ensuring that these values ​​are within the selected range;

[0106] The feature vectors of the training data are input into the MLP model; the output values ​​of each layer are calculated using the weights and biases of each layer, and this process is continued until the output layer is reached; a non-linear activation function is applied to the output of the hidden layer to transform the output of the hidden layer into a non-linear form;

[0107] The loss function is calculated using the prediction results and label data from the output layer to measure the error of the model's predictions.

[0108] Using mean squared error as the loss function, the formula for calculating the mean squared error loss function is as follows: ;in, This reflects the actual participants' preferences and interests regarding different science popularization content. To predict participants' preferences for different popular science content;

[0109] The backpropagation algorithm is used to calculate the gradient of each parameter with respect to the loss function, thereby updating the parameters to reduce the value of the loss function; the stochastic gradient descent algorithm is used to update the model parameters according to the gradient, so that the loss function gradually decreases.

[0110] The learning rate is adjusted to control the speed and stability of parameter updates, thereby optimizing the multilayer perceptron (MLP) model; the Adam algorithm is selected as the optimizer; the Adam algorithm is based on optimizing stochastic gradient descent and adaptively adjusts the learning rate;

[0111] After the training reaches the predetermined number of iterations, the final trained participant interest prediction model is obtained. The trained participant interest prediction model is then used to predict the historical feature dataset to obtain the participants' attention preferences for different popular science content.

[0112] The method of combining content recommendation algorithms to recommend different popular science content to participants based on their preferences for different popular science content and the characteristics of the popular science content includes:

[0113] Collect participants' preferences for different popular science content and the characteristics of popular science content, transform them into computable numerical features, encode the preferences for different popular science content and popular science activity data, and use one-hot encoding to transform them into vector representations;

[0114] Determine the feature columns that need one-hot encoding, collect and deduplicate the values ​​in the feature columns to obtain all possible values; for each feature value of the attention preference of different popular science content, iterate through all possible values; if the feature value of the attention preference of different popular science content is equal to the currently traversed value, then the corresponding position is 1, otherwise it is 0; for each feature value of different popular science content, also iterate through all possible values, if the feature value of different popular science content is equal to the currently traversed value, then the corresponding position is 1, otherwise it is 0;

[0115] For each different feature of popular science content, the results of its one-hot encoding are arranged into a vector in order; for each feature of different popular science content, the results of its one-hot encoding are also arranged into a vector in order; the length of this vector is equal to the number of all possible values, with only one position being 1 and the rest being 0;

[0116] For example, there are two different types of popular science content: "physics" and "biology"; it is necessary to collect participants' attention preferences for these two types of popular science content, and at the same time, it is also necessary to encode the characteristics of the popular science content itself.

[0117] For the one-hot encoding of attention preference features, three values ​​were first collected: "interested", "neutral", and "not interested". For the first type of popular science content, "physics", these three values ​​were iterated through:

[0118] If the focus preference is "interested", then the encoding is [1,0,0];

[0119] If the preference is "neutral", then the encoding is [0,1,0];

[0120] If the preference is "not interested", then the encoding is [0,0,1];

[0121] For the second type of popular science content, "biology", we also went through these three values ​​and got the same encoding result;

[0122] One-hot encoding of content features yielded two types of features for popular science content: "professionalism" and "entertainment." For the popular science content of "physics," these two features were iterated through:

[0123] If the feature is "professionalism", then it is encoded as [1,0];

[0124] If the feature is "fun", then it is encoded as [0,1].

[0125] For the popular science content of "biology", these two features were also examined, and the same coding results were obtained;

[0126] Combining the above steps, we finally obtain: vector representations of attention preference features, such as [1,0,0], [0,1,0], [0,0,1]; and vector representations of content features, such as [1,0], [0,1].

[0127] By calculating the similarity between the participant's preference vector and the feature vector of each piece of science content, the most similar science content is found and recommended to the participant; similarity is calculated using cosine similarity. ;in, For participants Attention Preference Vector and popular science content eigenvectors The dot product; For vectors The norm; For vectors The norm;

[0128] Participants focused on the similarity between preference vectors and feature vectors of each piece of science popularization content. Values ​​in Between; if participants focus on the similarity between the preference vector and the feature vector of each popular science content. The closer to 1, the more similar they are; the closer to 0, the more unrelated they are; and the closer to -1, the opposite they are.

[0129] By calculating the similarity between participants' attention preference vectors for different popular science content and the feature vectors of popular science content, different popular science content is recommended to different participants.

[0130] The method of using collaborative filtering algorithms to quickly recommend popular science content that matches the interests of participants by analyzing their behavioral patterns and interest similarity with other participants includes:

[0131] Acquire participant behavior data, which includes participants' browsing history, favorites, and purchased popular science books; participants' browsing history, favorites, and purchased popular science books constitute projects; for participants and participants' browsing history, favorites, and purchased popular science books, they are represented as a participant-project matrix, where rows represent participants, columns represent projects, and elements in the matrix represent the interaction behavior between participants and projects.

[0132] Using the Pearson correlation coefficient as a similarity measure, a similarity matrix is ​​calculated among participants to identify those with high similarity; for the participant-item matrix, the similarity among participants is calculated.

[0133] The formula for calculating the Pearson correlation coefficient is as follows: ;in, and The interaction behavior of two participants on different projects; and The average value of the interaction behaviors of the two participants on different projects;

[0134] Based on the similarity matrix, other participants are ranked to find those with high similarity to the target participant. A recommendation list is generated based on the interaction behavior of similar participants. For participants similar to the target participant, the popular science content they are interested in is analyzed and recommended to the target participant. Content that has already been viewed or saved is excluded according to filtering rules, and the recommendation list is further filtered to generate the final recommendation result. The features of the popular science content are extracted and transformed into numerical feature vectors using the bag-of-words model.

[0135] The methods used to assess whether participants achieve a positive impact after engaging with different science popularization content include:

[0136] The participants' scores after participating in science popularization activities are compared with a preset score threshold for participants who do not participate in science popularization activities. The participants' scores after participating in science popularization activities are recorded through a questionnaire, which is different from the questionnaire that records the participants' feedback satisfaction.

[0137] If a participant's score after participating in a science popularization activity is greater than the preset score threshold for participants who do not participate in science popularization activities, then the participant is deemed to have achieved a positive impact.

[0138] If a participant's score after participating in a science popularization activity is less than or equal to the preset score threshold for participants who do not participate in science popularization activities, then the participant is deemed not to have achieved a positive impact.

[0139] The method for scoring based on the positive impact achieved by participants after engaging in different science popularization content includes:

[0140] The importance of each data point in the science popularization activity data is quantified into a weight, which is expressed as a percentage to ensure that the sum of all weights equals 100%. For each participant, a comprehensive score is calculated based on the score of each data point and its corresponding weight. The score of each data point is obtained by the participant filling out a science popularization activity data question provided by the staff.

[0141] The formula for calculating the comprehensive score is as follows: ;in, This serves as the identifier for each data point in the science popularization activity data; The number of each data point in the science popularization activity data; A score for each data point in the science popularization activity data; The weight assigned to each data point in the science popularization activity data.

[0142] For example, the participation rate in a science popularization activity has a weight of 20%, and the participant's score on participation is 85. The participant's feedback satisfaction has a weight of 20%, and the participant's score on feedback satisfaction is 80. The participant's knowledge gain rate has a weight of 30%, and the knowledge gain rate is calculated by the difference between the scores of the tests before and after the test. The participant's score on knowledge gain rate is 92. The social impact of the science popularization activity has a weight of 30%, and the participant's score on social impact is 75.

[0143] The overall score is calculated as follows: (85*0.2) + (80*0.2) + (92*0.3) + (75*0.3)

[0144] =17+16+27.6+22.5=83.1 points.

[0145] In this embodiment, the provided quantitative scoring system for popular science provides an effective basis for assessing popular science work, which helps to improve the quality and efficiency of popular science work; the intelligent recommendation module recommends different popular science content to participants based on popular science activity data, which helps to give full play to the expertise of popular science workers;

[0146] The science popularization effect evaluation module assesses whether participants achieve positive impacts after engaging with different science popularization content; it provides rich reference resources for science popularization workers, helping to improve the innovation and effectiveness of science popularization work; the science popularization scoring module scores the positive impacts achieved by participants after engaging with different science popularization content. For each participant, a comprehensive score can be calculated based on their scores on various evaluation indicators and their corresponding weights, which helps to improve the effectiveness of science popularization activities and participant satisfaction.

[0147] Example 2

[0148] Please see Figure 2 As shown, parts not described in detail in this embodiment are described in Embodiment 1. A quantitative intelligent scoring method for popular science is provided, including:

[0149] S1. Obtain data on science popularization activities;

[0150] S2. Based on the data from the science popularization activities, recommend different science popularization content to the participants;

[0151] S3. Evaluate whether participants achieve a positive impact after engaging with different science popularization content;

[0152] S4. Rate the positive impact achieved by participants after engaging in different science popularization content.

[0153] Since the electronic device described in this embodiment is the electronic device used to implement the popular science quantitative intelligent scoring system described in this application embodiment, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the popular science quantitative intelligent scoring system described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. As long as those skilled in the art implement the electronic device used in the popular science quantitative intelligent scoring system described in this application embodiment, it falls within the protection scope of this application.

[0154] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0155] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A quantitative intelligent scoring system for popular science, characterized in that, include: The data collection module is used to acquire data from science popularization activities; The intelligent recommendation module is used to recommend different science popularization content to participants based on science popularization activity data; The science popularization effect evaluation module is used to evaluate whether participants achieve a positive impact after engaging with different science popularization content; The science popularization quantitative scoring module is used to score the positive impact achieved by participants after engaging in different science popularization content. The modules are connected via wired or wireless means to enable data transmission between them.

2. The popular science quantitative intelligent scoring system according to claim 1, characterized in that, The data on popular science activities include participation rates, participant satisfaction, participants' knowledge gain rate, and the social impact of the popular science activities.

3. The popular science quantitative intelligent scoring system according to claim 1, characterized in that, The methods for obtaining data from the science popularization activities include: Participation in science popularization activities includes the number of participants, the duration of participation, and the number of times participants attend. At the entrance of the science popularization activity venue, participants are given wristbands with RFID tags. Participants participate in the science popularization activities by wearing the wristbands with RFID tags. The number of participants, the duration of participation, and the number of times participants attend are recorded by reading the wristbands with RFID tags using a card reader. The participation rate in the aforementioned science popularization activities is: ;in, To increase participation in science popularization activities; The number of people participating in science popularization activities; The duration of participation in science popularization activities; The number of times you participate in interactive science popularization activities; , and A factor influencing participation in science popularization activities; Staff distributed questionnaires to participants and recorded their feedback and satisfaction scores, which were the participants' ratings of the science popularization activities. The method for obtaining the knowledge gain rate of the participants includes: Based on the stored science popularization content database, one or more sets of science popularization knowledge test questions related to the science popularization content are randomly generated. The test scores of participants before and after participating in science popularization activities are recorded through the scoring system. The knowledge gain rate of the participants is: ;in, The test score of participants after participating in science popularization activities; The test score of the participants before they participated in the science popularization activities; The social impact of science popularization activities includes the scope of dissemination and media attention; the scope of dissemination and media attention of science popularization activities are tracked and recorded through a news monitoring platform; The social impact of the aforementioned science popularization activities is as follows: ;in, As an indicator of social influence; To broaden the reach of science popularization activities; For media attention; and These are the influencing factors for the social influence indicator.

4. The popular science quantitative intelligent scoring system according to claim 3, characterized in that, The media attention mentioned includes: The media attention mentioned is: ;in, The number of media reports; The duration of media exposure.

5. A popular science quantitative intelligent scoring system according to claim 3, characterized in that, The method of recommending different science popularization content to participants based on science popularization activity data includes: The acquired science popularization activity data were normalized to standard deviation and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1 to eliminate the influence of dimensions between data. Finally, the normalized feature dataset was obtained. A participant interest prediction model is constructed. The input of the participant interest prediction model is a historical feature dataset; the output label is the participant's preference for different popular science content. By combining content recommendation algorithms, different popular science content is recommended to participants based on their preferences for different popular science content and the characteristics of the popular science content. Using collaborative filtering algorithms, popular science content that matches the interests of participants is recommended by analyzing the behavioral patterns and interest similarities between participants and other participants. By combining machine learning and deep learning technologies, the participant interest prediction model is continuously updated and optimized. This is achieved by constantly collecting and analyzing participants' science popularization activity data.

6. The popular science quantitative intelligent scoring system according to claim 5, characterized in that, The method for pre-building the participant interest prediction model includes: Prepare the training dataset, including input data and corresponding labels; the input data is a historical feature dataset, and the output labels are the participants' preferences for different popular science content. Define the structure of the multilayer perceptron (MLP) model, including an input layer, hidden layers, and an output layer; the size of the input layer matches the dimension of the feature vector of the science popularization activity; the number and size of the hidden layers are selected according to the task requirements and data complexity; the size of the output layer matches the dimension of the participants' attention preferences for different science popularization content. Initialize the weights and biases of the multilayer perceptron (MLP) model using a random initialization method; For the weights of each layer: determine the dimensions of the weight matrix, with an input dimension of n and an output dimension of [missing information]; select the range [missing information]. ,in, It is a positive number representing the range of values ​​for the weights; values ​​are randomly generated from a uniform distribution as initial weights, ensuring that these values ​​are within the selected range. For the bias at each layer: determine the dimension of the bias vector, and the output dimension is... ; Select range ,in, A positive number representing the range of bias values; values ​​are randomly generated from a uniform distribution as the initial bias, ensuring that these values ​​are within the selected range; The feature vectors of the training data are input into the MLP model; the output values ​​of each layer are calculated using the weights and biases of each layer, and this process is continued until the output layer is reached; a non-linear activation function is applied to the output of the hidden layer to transform the output of the hidden layer into a non-linear form; The loss function is calculated using the prediction results and label data from the output layer to measure the error of the model's predictions. Using mean squared error as the loss function, the formula for calculating the mean squared error loss function is as follows: ;in, This reflects the actual participants' preferences and interests regarding different science popularization content. To predict participants' preferences for different popular science content; The backpropagation algorithm is used to calculate the gradient of each parameter with respect to the loss function, thereby updating the parameters to reduce the value of the loss function; the stochastic gradient descent algorithm is used to update the model parameters according to the gradient, so that the loss function gradually decreases. The learning rate is adjusted to control the speed and stability of parameter updates, thereby optimizing the multilayer perceptron (MLP) model; the Adam algorithm is selected as the optimizer; the Adam algorithm is based on optimizing stochastic gradient descent and adaptively adjusts the learning rate; After the training reaches the predetermined number of iterations, the final trained participant interest prediction model is obtained. The trained participant interest prediction model is then used to predict the current feature dataset to obtain the participants' attention preferences for different popular science content.

7. A popular science quantitative intelligent scoring system according to claim 6, characterized in that, The method of combining content recommendation algorithms to recommend different popular science content to participants based on their preferences for different popular science content and the characteristics of the popular science content includes: Collect participants' preferences for different popular science content and the characteristics of popular science content, transform them into computable numerical features, encode the preferences for different popular science content and popular science activity data, and use one-hot encoding to transform them into vector representations; Determine the feature columns that need one-hot encoding, collect and deduplicate the values ​​in the feature columns to obtain all possible values; for each feature value of the attention preference of different popular science content, iterate through all possible values; if the feature value of the attention preference of different popular science content is equal to the value of the current iteration, then the corresponding position is 1, otherwise it is 0; for each feature value of different popular science content, also iterate through all possible values, if the feature value of different popular science content is equal to the value of the current iteration, then the corresponding position is 1, otherwise it is 0; For each different feature of popular science content, the results of its one-hot encoding are arranged into a vector in order; for each feature of different popular science content, the results of its one-hot encoding are also arranged into a vector in order; the length of this vector is equal to the number of all possible values, with only one position being 1 and the rest being 0; By calculating the similarity between the participant's preference vector and the feature vector of each piece of science content, the most similar science content is found and recommended to the participant; similarity is calculated using cosine similarity. ;in, For participants Attention Preference Vector and popular science content eigenvectors The dot product; For vectors The norm; For vectors The norm; Participants focused on the similarity between preference vectors and feature vectors of each piece of science popularization content. Values ​​in Between; if participants focus on the similarity between the preference vector and the feature vector of each popular science content. The closer to 1, the more similar they are; the closer to 0, the more unrelated they are; and the closer to -1, the opposite they are. By calculating the similarity between participants' attention preference vectors for different science popularization content and the feature vectors of the science popularization content, different science popularization content is recommended to different participants.

8. A popular science quantitative intelligent scoring system according to claim 5, characterized in that, The method of using collaborative filtering algorithms to recommend popular science content that matches the interests of participants by analyzing their behavioral patterns and interest similarity with other participants includes: Acquire participant behavior data, which includes participants' browsing history, favorites, and purchased popular science books; participants' browsing history, favorites, and purchased popular science books constitute projects; for participants and participants' browsing history, favorites, and purchased popular science books, they are represented as a participant-project matrix, where rows represent participants, columns represent projects, and elements in the matrix represent the interaction behavior between participants and projects. Using the Pearson correlation coefficient as a similarity measure, a similarity matrix is ​​calculated among participants to identify those with high similarity; for the participant-item matrix, the similarity among participants is calculated. The formula for calculating the Pearson correlation coefficient is as follows: ;in, and The interaction behavior of two participants on different projects; and The average value of the interaction behaviors of the two participants on different projects; Based on the similarity matrix, other participants are ranked to find those with high similarity to the target participant. A recommendation list is generated based on the interaction behavior of similar participants. For participants similar to the target participant, the popular science content they are interested in is analyzed and recommended to the target participant. Content that has already been viewed or saved is excluded according to filtering rules, and the recommendation list is further filtered to generate the final recommendation result. The features of the popular science content are extracted and transformed into numerical feature vectors using the bag-of-words model.

9. A popular science quantitative intelligent scoring system according to claim 7, characterized in that, The methods used to assess whether participants achieve a positive impact after engaging with different science popularization content include: The scores of participants after engaging in science popularization activities are compared with a preset score threshold for participants who do not participate in science popularization activities; the scores of participants after engaging in science popularization activities are recorded through questionnaires. If a participant's score after participating in a science popularization activity is greater than the preset score threshold for a participant who does not participate in a science popularization activity, then the participant is deemed to have achieved a positive impact. If a participant's score after participating in a science popularization activity is less than or equal to the preset score threshold for participants who do not participate in science popularization activities, then the participant is deemed not to have achieved a positive impact.

10. A quantitative intelligent scoring system for popular science according to claim 1, characterized in that, Methods for scoring the positive impact achieved by participants after engaging in different science popularization content include: The importance of each data point in the science popularization activity data is quantified into a weight, which is expressed as a percentage to ensure that the sum of all weights equals 100%. For each participant, a comprehensive score is calculated based on the score of each data point and its corresponding weight. The score of each data point is obtained by the participant filling out a science popularization activity data question distributed by the staff. The formula for calculating the comprehensive score is as follows: ;in, This serves as the identifier for each data point in the science popularization activity data; The number of each data point in the science popularization activity data; A score for each data point in the science popularization activity data; The weight assigned to each data point in the science popularization activity data.

Images