Knowledge content recommendation and push methods and related devices

Abstract

This application provides a knowledge content recommendation and push method and related equipment. The method includes: acquiring user characteristics, including user data characteristics, content data characteristics, and background data characteristics; inputting the user characteristics into a knowledge content recommendation model to obtain a knowledge content recommendation table; and pushing knowledge content to the user based on the knowledge content recommendation table. The technical solution of this application, by inputting user-related data characteristics, content data characteristics, and background data characteristics together into a knowledge content recommendation model for comprehensive analysis, obtains knowledge content that better matches the user's interests and needs, forming a knowledge content recommendation table. Pushing knowledge content to the user based on this table can help users quickly find knowledge content of interest in an environment with a large amount of information, thereby reducing the time cost of information filtering.

Description

Technical Field

[0001] This application relates to the fields of computer and communication technology, and more specifically, to a method and related equipment for recommending and pushing knowledge content. Background Technology

[0002] With the continuous advancement of technology, the types and quantities of knowledge and information are increasing day by day.

[0003] Many users who pursue lifelong learning, hope to expand their knowledge, and are willing to learn through social interaction often face the problem of excessively long search times for useful information. This makes it difficult for them to efficiently utilize fragmented time to learn and explore the knowledge they need. Summary of the Invention

[0004] The embodiments of this application provide a knowledge content recommendation and push method and related equipment, which can at least to some extent overcome the current problem that users find it difficult to efficiently utilize fragmented time to learn and explore the knowledge they need.

[0005] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0006] According to one aspect of the embodiments of this application, a knowledge content recommendation and push method is provided, comprising: obtaining user features, the user features including user data features, content data features and background data features; inputting the user features into a knowledge content recommendation model to obtain a knowledge content recommendation table; and pushing knowledge content to users according to the knowledge content recommendation table.

[0007] In some possible embodiments of this application, the acquisition of user features specifically includes: acquiring target user data, target content data, and target background data; and extracting features from the target user data, target content data, and target background data respectively to obtain the user data features, the content data features, and the background data features.

[0008] In some possible embodiments of this application, the acquisition of target user data, target content data, and target background data specifically includes: acquiring user data, content data, and background data; and filtering the user data, content data, and background data respectively to obtain target user data, target content data, and target background data.

[0009] In some possible embodiments of this application, the step of inputting the user features into a knowledge content recommendation model to obtain a knowledge content recommendation table specifically includes: standardizing and encoding the user features to obtain a user feature matrix; performing dimensionality reduction and depth processing on the user feature matrix to obtain a user feature target representation; sequentially inputting the user feature target representation into multiple neural networks to obtain a user preference feature representation; and inputting the user preference feature representation into a residual network to obtain a knowledge content recommendation table.

[0010] In some possible embodiments of this application, the step of performing dimensionality reduction and depth processing on the user feature matrix to obtain a user feature target representation specifically includes: inputting the user feature matrix into an autoencoder to obtain a low-dimensional representation of the user features; inputting the low-dimensional representation of the user features into a deep belief network to obtain a deep representation of the user features; the user feature target representation includes at least one of the low-dimensional representation of the user features and the deep representation of the user features.

[0011] In some possible embodiments of this application, the step of sequentially inputting the user feature target representation into multiple neural networks to obtain the user preference feature representation specifically includes: inputting the user feature target representation into a recurrent neural network to obtain sequence information; and inputting the sequence information into an attention network to obtain the user preference feature representation.

[0012] In some possible embodiments of this application, the step of inputting sequence information into an attention network to obtain a user preference feature representation specifically includes: integrating sequence information and content features into a three-dimensional array; and inputting the three-dimensional array into an attention network to obtain a user preference feature representation.

[0013] In some possible embodiments of this application, the step of inputting the user feature target representation into a recurrent neural network to obtain sequence information specifically includes: analyzing the user features to obtain a user behavior sequence; and inputting the user behavior sequence and the user feature target representation into a recurrent neural network to obtain sequence information.

[0014] In some possible embodiments of this application, the step of inputting the user features into the knowledge content recommendation model to obtain the knowledge content recommendation table specifically includes: analyzing the user features to obtain a user knowledge graph; and inputting the user knowledge graph into the knowledge content recommendation model to obtain the knowledge content recommendation table.

[0015] In some possible embodiments of this application, the step of inputting the user knowledge graph into a knowledge content recommendation model to obtain a knowledge content recommendation table specifically includes: inputting the knowledge graph into a graph convolutional layer to obtain corresponding embedding vectors; inputting the embedding vectors into a neural collaborative filtering network to obtain the user's preference results for each knowledge content; and generating a knowledge content recommendation table based on the user's preference results for each knowledge content.

[0016] In some possible embodiments of this application, the step of inputting the embedding vector into a neural collaborative filtering network to obtain the user's preference results for each knowledge content specifically includes: inputting the embedding vector into an interaction layer to obtain an interaction feature vector; and inputting the interaction feature vector into multiple hidden layers to obtain the user's preference results for each knowledge content.

[0017] In some possible embodiments of this application, the step of analyzing the user features to obtain a user knowledge graph specifically includes: obtaining the entity names of each entity, wherein the entities include user entities and knowledge entities, and the entity names correspond to user entity names and knowledge entity names; extracting entity relationships from the user features based on the entity names, wherein the entity relationships are the relationships between entities; and establishing a user knowledge graph based on the entity names and the entity relationships.

[0018] According to one aspect of the embodiments of this application, a knowledge content recommendation and push device is provided, the knowledge content recommendation and push device comprising: a feature acquisition module, configured to acquire user features, the user features including user data features, content data features and background data features; a model recommendation module, configured to input the user features into a knowledge content recommendation model to obtain a knowledge content recommendation table; and a content push module, configured to push knowledge content to the user according to the knowledge content recommendation table.

[0019] In some possible embodiments of this application, the feature acquisition module specifically includes: a data acquisition submodule, used to acquire target user data, target content data, and target background data; and a feature extraction submodule, used to extract features from the target user data, target content data, and target background data respectively to obtain the user data features, the content data features, and the background data features.

[0020] In some possible embodiments of this application, the data acquisition submodule specifically includes: a data acquisition unit, used to acquire user data, content data, and background data; and a data filtering unit, used to filter the user data, content data, and background data respectively to obtain target user data, target content data, and target background data.

[0021] In some possible embodiments of this application, the model recommendation module specifically includes: a matrix generation submodule, used to standardize and encode the user features to obtain a user feature matrix; a target representation submodule, used to perform dimensionality reduction and depth processing on the user feature matrix to obtain a user feature target representation; a user preference submodule, used to sequentially input the user feature target representation into multiple neural networks to obtain a user preference feature representation; and a residual network submodule, used to input the user preference feature representation into a residual network to obtain a knowledge content recommendation table. In some possible embodiments of this application, the target representation submodule specifically includes: a dimensionality reduction processing unit, used to input the user feature matrix into an autoencoder to obtain a low-dimensional representation of user features; and a deep processing unit, used to input the low-dimensional representation of user features into a deep belief network to obtain a deep representation of user features; the user feature target representation includes at least one of the low-dimensional representation and the deep representation.

[0022] In some possible embodiments of this application, the user preference submodule specifically includes: a sequence information unit, used to input the user feature target representation into a recurrent neural network to obtain sequence information; and a user preference unit, used to input the sequence information into an attention network to obtain a user preference feature representation.

[0023] In some possible embodiments of this application, the user preference unit is specifically used to perform: integrating sequence information and content features into a three-dimensional array; and inputting the three-dimensional array into an attention network to obtain a user preference feature representation.

[0024] In some possible embodiments of this application, the sequence information unit is specifically used to perform: analyzing the user features to obtain a user behavior sequence; and inputting the user behavior sequence and the user feature target representation into a recurrent neural network to obtain sequence information.

[0025] In some possible embodiments of this application, the model recommendation module specifically includes: a graph generation submodule, used to analyze the user features to obtain a user knowledge graph; and a model recommendation submodule, used to input the user knowledge graph into a knowledge content recommendation model to obtain a knowledge content recommendation table.

[0026] In some possible embodiments of this application, the model recommendation submodule specifically includes: a vector embedding unit, used to input the knowledge graph into a graph convolutional layer to obtain corresponding embedding vectors; a collaborative filtering unit, used to input the embedding vectors into a neural collaborative filtering network to obtain the user's preference results for each knowledge content; and a table generation unit, used to generate a knowledge content recommendation table based on the user's preference results for each knowledge content.

[0027] In some possible embodiments of this application, the collaborative filtering unit is specifically used to perform: inputting the embedding vector into the interaction layer to obtain the interaction feature vector; inputting the interaction feature vector into multiple hidden layers to obtain the user's preference results for each knowledge content.

[0028] In some possible embodiments of this application, the graph generation submodule specifically includes: an entity name extraction unit, used to obtain the entity name of each entity, wherein the entity includes user entities and knowledge entities, and the entity name corresponds to the user entity name and the knowledge entity name; an entity relationship extraction unit, used to extract entity relationships from the user features based on the entity names, wherein the entity relationships are the relationships between entities; and a knowledge graph generation unit, used to establish a user knowledge graph based on the entity names and the entity relationships.

[0029] According to one aspect of the embodiments of this application, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the knowledge content recommendation and push method as described in the above embodiments.

[0030] According to one aspect of the embodiments of this application, an electronic device is provided, including: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the knowledge content recommendation and push method as described in the above embodiments.

[0031] A computer program product includes one or more computer programs, characterized in that, when the one or more computer programs are executed by one or more processors, they implement the steps of the knowledge content recommendation and push method as described in the above embodiments.

[0032] In some embodiments of this application, the technical solutions provide that by inputting user-related data features, content data features, and background data features into a knowledge content recommendation model for comprehensive analysis, knowledge content that better matches user interests and needs is obtained, forming a knowledge content recommendation table. Based on this table, knowledge content is pushed to the user. This helps users quickly find interesting knowledge content in an environment with a large amount of information, thereby reducing the time cost of information filtering. Simultaneously, the recommended content received by the user is closer to their interests, which can improve user satisfaction and user experience, increase user activity, and enable users to utilize fragmented time to learn and explore the knowledge they need.

[0033] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0034] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings:

[0035] Figure 1 A schematic diagram of an exemplary system architecture to which the technical solutions of the embodiments of this application can be applied is shown.

[0036] Figure 2 The illustration shows a flowchart of a knowledge content recommendation and push method provided in an embodiment of this application.

[0037] Figure 3 It shows according to Figure 2 A flowchart illustrating a specific implementation of step S100 in the knowledge content recommendation and push method shown in the corresponding embodiment.

[0038] Figure 4 It shows according to Figure 2 A flowchart illustrating a specific implementation of step S200 in the knowledge content recommendation and push method shown in the corresponding embodiment.

[0039] Figure 5 It shows according to Figure 4 A flowchart illustrating a specific implementation of step S230 in the knowledge content recommendation and push method shown in the corresponding embodiment.

[0040] Figure 6 It shows according to Figure 2 The flowchart shown is a specific implementation of step S200 in the knowledge content recommendation and push method.

[0041] Figure 7 A schematic diagram of the structure of a knowledge content recommendation and push device provided in an embodiment of this application is shown.

[0042] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0043] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.

[0044] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0045] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0046] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily need to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0047] Figure 1 A schematic diagram of an exemplary system architecture to which the technical solutions of the embodiments of this application can be applied is shown.

[0048] like Figure 1 As shown, the system architecture may include terminal devices (such as...) Figure 1 The device shown includes one or more of a smartphone 101, tablet 102, and portable computer 103 (which could also be a desktop computer, etc.), a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the terminal device and the server 105. The network 104 can include various connection types, such as wired communication links, wireless communication links, etc.

[0049] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, there can be any number of terminal devices, networks, and servers. For example, server 105 could be a server cluster composed of multiple servers.

[0050] Users can interact with server 105 via network 104 using terminal devices to receive or send messages, etc. Server 105 can be a server that provides various services. For example, a user can upload user features to server 105 using terminal device 103 (or terminal device 101 or 102). Server 105 can input the user features into a knowledge content recommendation model to obtain a knowledge content recommendation table; and push knowledge content to the user based on the knowledge content recommendation table.

[0051] It should be noted that the knowledge content recommendation and push method provided in this application embodiment is generally executed by server 105, and correspondingly, the knowledge content recommendation and push device is generally set in server 105. However, in other embodiments of this application, the terminal device may also have similar functions to the server, thereby executing the knowledge content recommendation and push scheme provided in this application embodiment.

[0052] The implementation details of the technical solutions in the embodiments of this application are described in detail below:

[0053] Figure 2 A flowchart of a knowledge content recommendation and push method according to an embodiment of this application is shown. This knowledge content recommendation and push method can be executed by a server, which can be... Figure 1 The server shown. (Refer to...) Figure 2 As shown, the methods for recommending and pushing this knowledge content include at least the following:

[0054] S100, Obtain user characteristics, which include user data characteristics, content data characteristics, and background data characteristics.

[0055] S200, input the user features into the knowledge content recommendation model to obtain the knowledge content recommendation table.

[0056] S300: Based on the knowledge content recommendation table, push knowledge content to the user.

[0057] In this embodiment, by inputting user-related data features, content data features, and background data features into the knowledge content recommendation model for comprehensive analysis, a knowledge content recommendation table is generated that better matches the user's interests and needs. Based on this table, knowledge content is pushed to the user. In an environment with a large amount of information, this helps users quickly find knowledge content of interest, thereby reducing the time cost of information filtering. At the same time, the recommended content received by the user is closer to their interests, which can improve user satisfaction and user experience, increase user activity, and enable users to utilize fragmented time to learn and explore the knowledge they need.

[0058] In S100, multi-dimensional user features, including user data features, content data features, and background data features, integrate data from multiple dimensions, making user feature data more comprehensive. Through the comprehensive analysis of user data features, content data features, and background data features, a comprehensive user profile is constructed. This helps to understand the user's personalized interests and needs regarding knowledge content, solving the problem that a single feature may not be able to fully reflect user needs, improving the accuracy of the recommendation model, and making the recommendation results more precise.

[0059] Specifically, in some embodiments, the specific implementation of step S100 can be found in [reference needed]. Figure 3 . Figure 3 It is based on Figure 2 The detailed description of step S100 in the knowledge content recommendation and push method shown in the corresponding embodiment includes the following steps:

[0060] S110, acquire target user data, target content data, and target background data.

[0061] S120, feature extraction is performed on the target user data, target content data, and target background data respectively to obtain the user data features, the content data features, and the background data features.

[0062] In this embodiment, target raw data such as target user data, target content data, and target background data are first collected, and then features are extracted and constructed from the above target raw data. These features can represent the user's interests and preferences, as well as the characteristics of the content.

[0063] In S110, target user data includes basic user information (such as age, gender, occupation, etc.) and user behavior data (such as browsing history, learning path, completed courses, learning preferences, etc.). Target content data involves metadata of learning materials, such as course titles, descriptions, categories, difficulty levels, related tags, user ratings, and feedback. Target background data may include time (such as the time the user studies), location (such as the user's geographical location), and device type (such as the learning device the user uses).

[0064] Specifically, in some embodiments, the specific implementation of step S110 can be found in the following embodiments. This embodiment is based on... Figure 3 According to the detailed description of step S110 in the knowledge content recommendation and push method shown in the corresponding embodiment, step S110 in the knowledge content recommendation and push method may include the following steps:

[0065] Acquire user data, content data, and background data.

[0066] The user data, content data, and background data are filtered separately to obtain target user data, target content data, and target background data.

[0067] In this embodiment, before feature engineering, data needs to be collected and preprocessed, including filtering. Data preprocessing includes cleaning the data, handling missing and outlier values, and normalizing the data. This involves removing invalid data, filling in missing data, correcting outlier data, and normalizing the data. The resulting data are the target user data, target content data, and target background data.

[0068] In S120, key features are effectively identified and core information is obtained by extracting features from user, content and background data, avoiding interference from noisy data and improving the effectiveness of the analysis.

[0069] Specifically, when performing feature engineering, features can be constructed using traditional feature extraction methods or through convolution of multiple convolutional layers; this application does not impose any limitations on this.

[0070] In S200, the aforementioned knowledge content recommendation model is a neural network model. It continuously updates user characteristics based on user behavior and feedback, optimizing recommended content in real time and maintaining the dynamics and adaptability of the recommendation system. The user characteristics input into the knowledge content recommendation model are multi-dimensional, which helps improve the accuracy of the recommendation model and makes the recommendation results more precise.

[0071] At the same time, this knowledge content recommendation model is also a customized recommendation model, capable of generating recommendation results that are highly relevant to user characteristics.

[0072] Specifically, in some embodiments, the specific implementation of step S200 can be found in [reference needed]. Figure 4 . Figure 4 It is based on Figure 2 The detailed description of step S200 in the knowledge content recommendation and push method shown in the corresponding embodiment includes the following steps:

[0073] S210, The user features are standardized and encoded to obtain a user feature matrix.

[0074] S220, the user feature matrix is ​​subjected to dimensionality reduction and depth processing to obtain the user feature target representation.

[0075] S230, the user feature target representation is sequentially input into multiple neural networks to obtain the user preference feature representation.

[0076] S240, the user preference feature representation is input into the residual network to obtain the knowledge content recommendation table.

[0077] In this embodiment, by applying standardization, dimensionality reduction, deep processing, and residual networks, technical challenges such as data consistency, computational complexity, high-dimensional data issues, and recommendation accuracy are effectively addressed. Ultimately, efficient knowledge content recommendation is achieved, improving the performance of the recommendation system and significantly enhancing the user experience.

[0078] In S210, on the one hand, standardization eliminates the influence of different units and distributions on model training, ensuring the comparability of input data and improving data consistency and model learning efficiency; on the other hand, through encoding (e.g., One-Hot encoding), categorical data is converted into numerical form, making it suitable for model input and reducing errors or biases caused by data format diversity.

[0079] In S220, on the one hand, dimensionality reduction reduces computational complexity and avoids overfitting while retaining important information. On the other hand, deep processing removes redundant features and focuses on features that actually contribute to the recommendation results.

[0080] Specifically, in some embodiments, the specific implementation of step S220 can be found in the following embodiments. This embodiment is based on... Figure 4 The detailed description of step S220 in the knowledge content recommendation and push method shown in the corresponding embodiment includes the following steps:

[0081] The user feature matrix is ​​input into the autoencoder to obtain a low-dimensional representation of the user features.

[0082] The low-dimensional representation of the user features is input into a deep belief network to obtain a deep representation of the user features.

[0083] The user feature target representation includes at least one of the following: a low-dimensional representation of user features and a deep representation of user features.

[0084] In this embodiment, on the one hand, dimensionality reduction is performed using an autoencoder to compress high-dimensional data into a low-dimensional representation. This process retains the most important features while removing irrelevant information and noise, significantly reducing data dimensionality and computational burden, thus making subsequent model training more efficient. On the other hand, a deep belief network is used to perform deep processing on the low-dimensional representation of user features, uncovering deeper feature relationships and patterns, enhancing the expressive power of features, and better capturing users' potential preferences and behavioral patterns. The resulting user feature target representation combines both low-dimensional and deep representations, providing a multi-layered, multi-faceted user profile. This allows the recommendation system to more comprehensively understand user needs, improving the accuracy and personalization of the recommendation system, enhancing the model's adaptability to unknown data, and ultimately improving user experience and satisfaction. In other words, this embodiment effectively solves the problems of computational complexity, feature redundancy, and noise caused by high-dimensional data by inputting the user feature matrix into the autoencoder and deep belief network, while simultaneously achieving a multi-layered deep representation of user features. The structure combining low-dimensional and deep representations not only improves the quality of feature extraction but also lays the foundation for building a more powerful recommendation system.

[0085] In S230, user preferences are mined from different angles through diverse neural network structures, providing a more comprehensive understanding of user preferences. The generated user preference feature representation can more accurately reflect users' interests and behavioral patterns, enhancing the model's expressive power and improving the generalization ability and accuracy of the recommendation system.

[0086] Specifically, in some embodiments, the specific implementation of step S230 can be found in [reference needed]. Figure 5 . Figure 5 It is based on Figure 4 A detailed description of step S230 in the knowledge content recommendation and push method shown in the corresponding embodiment: Step S230 in the knowledge content recommendation and push method may include the following steps:

[0087] S232, The user feature target representation is input into a recurrent neural network to obtain sequence information.

[0088] S234. Input the sequence information into the attention network to obtain the user preference feature representation.

[0089] In this embodiment, the combination of recurrent neural networks and attention networks enhances the correlation between features. Through sequence modeling and weight allocation, the mutual influence of features is improved, resulting in a more comprehensive representation of user preferences. This representation comprehensively considers time factors and the relationships between features, providing a richer user profile for the recommendation system. Simultaneously, in-depth analysis of user behavior sequences and dynamic weighting of important features enable the recommendation system to more accurately match user interests, thereby improving recommendation relevance and user satisfaction. In other words, this embodiment solves technical problems such as sequence information capture, important information extraction, and insufficient feature correlation by sequentially inputting user features into a recurrent neural network and an attention network. Ultimately, it achieves a more accurate and personalized representation of user preferences.

[0090] In S232, the user feature target representation is processed by a recurrent neural network, which can effectively capture and retain the temporal features of user behavior, enabling the model to understand the changes in user preferences at different time periods, thereby more accurately reflecting the user's overall preferences.

[0091] Specifically, in some embodiments, the specific implementation of step S232 can be found in the following embodiments. This embodiment is based on... Figure 4 The detailed description of step S232 in the knowledge content recommendation and push method shown in the corresponding embodiment includes the following steps:

[0092] The user characteristics are analyzed to obtain a user behavior sequence.

[0093] The user behavior sequence and the user feature target representation are input into a recurrent neural network to obtain sequence information.

[0094] In this embodiment, on the one hand, by combining user features and behavior sequences into a recurrent neural network, the model can understand changes in user features over time, effectively capture the time dependence in user behavior sequences, and simultaneously focus on both short-term and long-term behavioral features. This results in a more dynamic representation of user features, more accurately reflecting the user's current preferences and needs, and achieving more personalized recommendations. On the other hand, by comprehensively learning and capturing complete user behavior sequences, the recurrent neural network can retain more contextual information, effectively reducing information loss during feature extraction and enhancing the model's robustness to noise. In other words, this embodiment solves the problems of sequence data processing, static feature representation, information loss, and long short-term memory by inputting user features and behavior sequences into a recurrent neural network. Ultimately, it achieves the technical effects of dynamic user feature representation, enhanced sequence information capture, improved recommendation personalization, and reduced information loss.

[0095] In S234, an attention mechanism is introduced, which can dynamically assign different weights to the input at different time points, helping the model focus on the behaviors or features that have the greatest impact on user preferences, improving the accuracy of information extraction and reducing noise interference.

[0096] Specifically, in some embodiments, the specific implementation of step S230 can be found in [reference needed]. Figure 5 . Figure 5 It is based on Figure 4 A detailed description of step S230 in the knowledge content recommendation and push method shown in the corresponding embodiment: Step S230 in the knowledge content recommendation and push method may include the following steps:

[0097] The sequence information and content features are integrated into a three-dimensional array.

[0098] The three-dimensional array is input into the attention network to obtain a representation of user preference features.

[0099] This embodiment integrates sequence information and content features into a three-dimensional array, effectively capturing time, content features, and their interactions. It also processes multi-dimensional information within a unified framework, enhancing the model's expressive power. The generated user preference feature representation is more comprehensive, considering the complex relationships between time and features, and better understanding changes in user interests, thereby improving the personalization and user satisfaction of the recommendation system. Furthermore, this embodiment uses an attention network to dynamically adjust the focus and assign different weights to features at different time points. This filters out information less relevant to user preferences and more accurately focuses on information with a greater impact on user preferences, improving the model's robustness and accuracy. Specifically, through dynamic weighting of the attention mechanism, it can more accurately extract the features with the greatest impact on user preferences, reduce redundant information, and improve the relevance of recommendations.

[0100] In this embodiment, by integrating sequence information and content features into a three-dimensional array and inputting it into an attention network, technical challenges such as sequence information capture, feature fusion, dynamic adjustment of importance, and noise sensitivity are addressed. Ultimately, a more accurate and comprehensive representation of user preference features is obtained.

[0101] In S240, the residual network alleviates the vanishing gradient problem in deep networks by introducing shortcut connections, allowing deeper networks to still be trained effectively. Simultaneously, residual learning helps capture more complex user-content interactions, resulting in a knowledge content recommendation table that provides more personalized and accurate content recommendations, better meeting user needs and thus improving user satisfaction and engagement.

[0102] During model training, the knowledge content recommendation model can be treated as a whole, with all neural networks trained together on the same sample set to obtain a well-trained model. In other embodiments, each neural network can use a pre-trained neural network model; that is, before training the knowledge content recommendation model, each neural network can be pre-trained separately, and then all neural networks can be incorporated into the knowledge content recommendation model for joint training.

[0103] Specifically, the training methods for the aforementioned knowledge content recommendation model include:

[0104] Obtain the user feature sample set, which contains multiple user feature samples, each of which is labeled with a corresponding knowledge content recommendation tag.

[0105] The user feature samples are input one by one into the knowledge content recommendation model to obtain the knowledge content recommendation results.

[0106] Based on the output knowledge content recommendation results and the knowledge content recommendation tags, the parameters of the knowledge content recommendation model are updated until a predetermined termination condition is met, at which point training ends, and a trained knowledge content recommendation model is obtained.

[0107] In the embodiments of this application, during training, a user feature sample set containing multiple user feature samples can be obtained first, and each user feature sample is labeled with a corresponding knowledge content recommendation tag; then, the multiple user feature samples are divided into a training set, a validation set, and a test set according to a predetermined ratio; then, the parameters of the encoder and decoder in the knowledge content recommendation model are adjusted and determined according to the user feature samples included in the training set, validation set, and test set, and the trained knowledge content recommendation model is obtained.

[0108] When training the model, the user feature sample set can be divided into a training set, a validation set, and a test set. Then, the model is trained based on the training set, validated based on the validation set, and tested based on the test set to obtain a well-trained knowledge content recommendation model.

[0109] Before training on the training set, the user feature samples in the training set can be preprocessed. Preprocessing includes vector adjustment, normalization, data augmentation, and class encoding.

[0110] After obtaining the enhanced training set, the knowledge content recommendation model can be trained based on the enhanced training set, and the parameters and weights in the network can be updated.

[0111] Specifically, user feature samples from the training set are input into the knowledge content recommendation model to obtain the knowledge content recommendation results output by the model. The knowledge content recommendation results are compared with the recommended knowledge content tags, and a loss function is calculated. Then, the stochastic gradient descent method is used to minimize the loss function. The parameters and weights in the knowledge content recommendation model are updated via backpropagation until the loss function meets predetermined conditions, such as convergence or falling below a predetermined threshold.

[0112] In some embodiments, the knowledge content recommendation result includes a knowledge content recommendation table and user preference parameters. The knowledge content recommendation tags may include at least one of the knowledge content recommendation table tags and user preference parameter tags. The loss function is calculated by comparing the knowledge content recommendation table with the knowledge content recommendation table tags and / or by comparing the user preference parameters with the user preference parameter tags.

[0113] After training, the image knowledge content recommendation network with updated parameters can be validated using a validation set. Specifically, the network is tuned based on the validation set data. When the loss function meets predetermined conditions, the model parameters for that stage are output. If the loss function does not meet the predetermined conditions, hyperparameters such as the learning rate are automatically adjusted, and the next round of network model training begins.

[0114] When the loss function computed on the validation set meets predetermined conditions, the parameters and weights can be retained, and then the retained parameters are tested based on the test set. Specifically, the input test set data enables the image knowledge content recommendation network with retained parameters and weights to output knowledge content recommendation results and model weights. By comparing the model loss and corresponding weights across multiple rounds, the model weights with the minimum loss are output, thus determining the trained knowledge content recommendation model.

[0115] After obtaining the trained knowledge content recommendation model, knowledge content recommendation can be completed based on the knowledge content recommendation model.

[0116] Furthermore, data augmentation is not required for the input data in the validation and test sets.

[0117] Specifically, in some embodiments, the specific implementation of step S200 can be found in [reference needed]. Figure 6 . Figure 6 It is based on Figure 2 The detailed description of step S200 in the knowledge content recommendation and push method shown in the corresponding embodiment includes the following steps:

[0118] S280, Analyze the user characteristics to obtain a user knowledge graph.

[0119] S290, input the user knowledge graph into the knowledge content recommendation model to obtain the knowledge content recommendation table.

[0120] In this embodiment, the user knowledge graph can reveal the connections between different knowledge domains, enabling the system to make cross-domain knowledge recommendations and expand the user's knowledge base. By constructing and analyzing changes in the user knowledge graph, the understanding of user characteristics can be further deepened, improving the intelligence and personalization of knowledge content recommendations, realizing a data-driven feedback mechanism, continuously optimizing recommendation strategies, and enhancing user experience.

[0121] Meanwhile, the constructed user knowledge graph can also be used for display to increase users' sense of participation and identification with the recommendation system, and promote interaction between users and the platform.

[0122] Specifically, in some embodiments, the training method for the above-mentioned knowledge content recommendation model includes:

[0123] Obtain a user knowledge graph sample set, which contains multiple user knowledge graph samples, each of which is pre-labeled with a corresponding knowledge content recommendation result;

[0124] Input each user knowledge graph sample in the user knowledge graph sample set into the knowledge content recommendation model to obtain the knowledge content recommendation result.

[0125] Based on the knowledge content recommendation results and knowledge content recommendation tags, the parameters of the knowledge content recommendation model are updated until a predetermined termination condition is met, at which point the training ends and a well-trained knowledge content recommendation model is obtained.

[0126] Specifically, in some embodiments, updating the parameters of the knowledge content recommendation model based on the knowledge content recommendation results and knowledge content recommendation tags until a predetermined termination condition is met, thus ending training and obtaining a trained knowledge content recommendation model, may specifically include:

[0127] If, in the user knowledge graph sample set, less than a predetermined number of user knowledge graph samples are input into the knowledge content recommendation model, and the knowledge content recommendation result is consistent with the knowledge content recommendation label, then the parameters of the knowledge content recommendation model are updated.

[0128] If, in the user knowledge graph sample set, more than a predetermined number of user knowledge graph samples are input into the knowledge content recommendation model, and the knowledge content recommendation result is consistent with the knowledge content recommendation label, then the predetermined termination condition is met, the training ends, and the trained knowledge content recommendation model is obtained.

[0129] Specifically, in some other embodiments, updating the parameters of the knowledge content recommendation model based on the knowledge content recommendation results and knowledge content recommendation tags until a predetermined termination condition is met, thus ending the training and obtaining a trained knowledge content recommendation model, may specifically include:

[0130] Based on the knowledge content recommendation results and knowledge content recommendation tags, determine the loss function;

[0131] The parameters of the knowledge content recommendation model are updated according to the loss function until a predetermined termination condition is met, at which point the training ends and a well-trained knowledge content recommendation model is obtained.

[0132] In this embodiment, the predetermined termination condition for the loss function can be that the loss function converges or the loss function is less than a predetermined loss (e.g., 0.001).

[0133] In the above embodiments, the corresponding knowledge content recommendation results are obtained by correcting the neural network model obtained through multiple trainings. The neural network model is obtained through multiple trainings. The more training samples it has, the more accurate the results are. It basically does not require maintenance during operation, which also reduces maintenance costs and improves the efficiency and accuracy of knowledge content recommendation.

[0134] In S280, by establishing a user knowledge graph, technical problems such as user feature modeling, relationship mining, and information integration can be effectively solved, and positive effects such as improved recommendation accuracy, enhanced user insights, and improved content relevance can be achieved.

[0135] Specifically, in some embodiments, the specific implementation of step S280 can be found in the following embodiments. This embodiment is based on... Figure 6 The detailed description of step S280 in the knowledge content recommendation and push method shown in the corresponding embodiment includes the following steps:

[0136] Obtain the entity name of each entity, which includes user entities and knowledge entities, and the entity name includes the user entity name and the knowledge entity name respectively.

[0137] Based on the entity name, entity relationships are extracted from the user characteristics, where the entity relationship is the relationship between the entities.

[0138] A user knowledge graph is established based on the entity names and entity relationships.

[0139] In this embodiment, by modeling the relationship between user entities and knowledge entities, the relationship between them is clarified. Through relationship mining between entities in the knowledge graph, potential knowledge associations and new points of interest can be discovered. The recommendation system can then provide more relevant content, achieve more accurate personalized recommendations, improve the effectiveness of the recommendation system, promote user learning and exploration, and thus increase user satisfaction and usage. At the same time, the structured characteristics of the user knowledge graph can also support more complex queries and analyses, enhancing data mining capabilities.

[0140] In some embodiments, knowledge graphs can be visualized to help users understand the knowledge domains and personal interests involved more intuitively. Users can also understand the source and basis of recommendation results through knowledge graphs, thereby increasing their trust in the recommendation system and improving the user experience.

[0141] In S290, the user knowledge graph is input into the knowledge content recommendation model. The user knowledge graph contains rich feature information and retains the correlation between feature information, which more accurately captures the user's true preferences for knowledge content, helps to improve the accuracy of the recommendation model, and makes the recommendation results more accurate.

[0142] Specifically, in some embodiments, the specific implementation of step S290 can be found in the following embodiments. This embodiment is based on... Figure 6 A detailed description of step S290 in the knowledge content recommendation and push method shown in the corresponding embodiment: Step S290 in the knowledge content recommendation and push method may include the following steps:

[0143] The knowledge graph is input into the graph convolutional layer to obtain the corresponding embedding vector.

[0144] The embedded vectors are input into a neural collaborative filtering network to obtain the user's preference results for each knowledge content.

[0145] Based on the user's preferences for each knowledge content, a knowledge content recommendation table is generated.

[0146] In this embodiment, by embedding the user's knowledge graph into the recommendation model, the model, combined with the user's knowledge structure, can more accurately capture the user's true preferences for knowledge content, provide diversified recommendation results, expand the user's knowledge base, and improve recommendation quality. Simultaneously, during use, the model can also adjust the recommendation strategy in a timely manner based on the dynamic changes in the user's knowledge graph, maintaining efficiency and adaptability. In other words, this embodiment utilizes the embedding vectors extracted by graph convolutional layers and neural collaborative filtering methods to make the recommended content more relevant and personalized, meeting the user's needs.

[0147] Specifically, the neural collaborative filtering network contains an interaction layer and multiple hidden layers. The neural collaborative filtering network performs the following operations:

[0148] The embedded vector is input into the interaction layer to obtain the interaction feature vector;

[0149] The interactive feature vectors are input into multiple hidden layers to obtain the user's preference results for each knowledge content.

[0150] In this embodiment, the introduction of an interaction layer allows the model to consider the complexity of various factors influencing user preferences, thereby improving the quality and relevance of recommendation results. This embodiment, through the combination of interaction feature vectors and multiple hidden layers, learns the user's latent interests and preferences, enabling more accurate prediction of user preferences for different knowledge content. This leads to more diverse and novel recommended content, not only improving recommendation accuracy but also expanding the user's knowledge base.

[0151] In S300, content pushed to users based on the recommendation table is more in line with their interests, enabling users to quickly obtain the knowledge they need, thereby enhancing overall satisfaction and willingness to use, and helping to increase click-through rate and engagement.

[0152] Pushing knowledge content to users can be done by recommending relevant learning content, such as courses, articles, and videos, based on user characteristics and preferences, and recommending learning communities with similar levels of expertise in the same supported field.

[0153] Specifically, in some embodiments, knowledge content ranked higher in the knowledge content recommendation table can be prioritized for push to users. At the same time, user feedback and interaction data (such as click-through rate and reading time) can be analyzed in real time to dynamically adjust the recommendation list and ensure that the content is always relevant. In other words, user interactions with the recommended content (such as clicks, views, completions, and ratings) are used as inputs to further optimize the recommendation algorithm and update the model.

[0154] In other embodiments, the corresponding knowledge content can be determined from the knowledge content recommendation table by combining the user's current activity or state (e.g., login time, content being read) and context-related content can be pushed.

[0155] It should be noted that in this embodiment, all knowledge content pushed to users is collected from all file information across multiple internal knowledge platforms using methods such as web scraping and importing. For example, to find an employee handbook, there's no need to search back and forth between cloud drives and internal knowledge platforms; this system acts as a central manager, allowing users to search all internal platform materials. Furthermore, obtaining push content using the above method eliminates the need for migration between old and new knowledge systems.

[0156] The knowledge content pushed to users can be obtained through the knowledge collection module, which supports various knowledge collection methods for both structured and unstructured data, including importing and web scraping. Users can quickly collect various internal knowledge resources, such as documents, images, videos, audio, and links, into the knowledge base.

[0157] After acquiring various content data, it can also be organized through the knowledge organization module. This module provides powerful knowledge organization functions. On the one hand, there is passive organization, where users can classify, tag, and summarize the collected knowledge for convenient knowledge management and retrieval. On the other hand, there is proactive organization functionality recommended by the knowledge management system, such as categorizing knowledge into different categories based on different page access levels. Simultaneously, it also supports knowledge association and recommendation, helping users discover knowledge they might need.

[0158] The organized knowledge content can be stored in the cloud via a knowledge storage module, allowing users to access and use the knowledge base anytime, anywhere, without network or environmental limitations. It supports multiple knowledge sharing methods, including internal sharing, external sharing, and access control. Users can share knowledge from the knowledge base with employees, partners, or customers within the enterprise, maximizing the value of that knowledge.

[0159] In some feasible implementations, features such as intelligent question answering, intelligent search, and intelligent recommendation can also be added, implemented through a knowledge-based question-and-answer module. Users can quickly acquire the knowledge they need through these application scenarios, improving work efficiency and innovation capabilities.

[0160] The modules described above employ a micro-frontend architecture, a software development methodology that breaks down a large frontend application into multiple smaller, independent frontend modules. Each module can be developed and maintained independently by different teams. These modules can be considered "micro-frontends" because they can be deployed independently, updated and iterated separately without interfering with other modules, making them more flexible and easier to maintain.

[0161] Therefore, in this embodiment, the content pushed to the user can also be based on the knowledge description input by the user. This knowledge description can be a file name, knowledge description, or image description. The knowledge can be located in different platforms through keyword matching, fuzzy search, etc., so as to avoid a cumbersome search process.

[0162] In the above embodiments, data is encrypted during storage and transmission to ensure that even if data is intercepted, it cannot be read by unauthorized individuals. Strong encryption standards such as AES or RSA can be used to protect the data. Sensitive information, such as user passwords or personal information, should be processed using hash algorithms to ensure security. Role-based access control (RBAC) is implemented to ensure that users can only access the information required for their roles. Sensitive operations are subject to permission auditing to ensure that only authorized users can execute them. Servers and data centers are protected from physical intrusion, ensuring that all physical access is authorized and logged.

[0163] The following describes an apparatus embodiment of this application, which can be used to execute the knowledge content recommendation and push method described in the above embodiments of this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the knowledge content recommendation and push method described above.

[0164] Figure 7 A block diagram of a knowledge content recommendation push device according to an embodiment of this application is shown.

[0165] Reference Figure 7 As shown, a knowledge content recommendation and push device 700 according to an embodiment of this application includes: a feature acquisition module 710, a model recommendation module 720, and a content push module 730.

[0166] The feature acquisition module 710 is used to acquire user features, which include user data features, content data features, and background data features; the model recommendation module 720 is used to input the user features into a knowledge content recommendation model to obtain a knowledge content recommendation table; and the content push module 730 is used to push knowledge content to users based on the knowledge content recommendation table.

[0167] In some possible embodiments of this application, the feature acquisition module specifically includes: a data acquisition submodule, used to acquire target user data, target content data, and target background data; and a feature extraction submodule, used to extract features from the target user data, target content data, and target background data respectively to obtain the user data features, the content data features, and the background data features.

[0168] In some possible embodiments of this application, the data acquisition submodule specifically includes: a data acquisition unit, used to acquire user data, content data, and background data; and a data filtering unit, used to filter the user data, content data, and background data respectively to obtain target user data, target content data, and target background data.

[0169] In some possible embodiments of this application, the model recommendation module specifically includes: a matrix generation submodule, used to standardize and encode the user features to obtain a user feature matrix; a target representation submodule, used to perform dimensionality reduction and depth processing on the user feature matrix to obtain a user feature target representation; a user preference submodule, used to sequentially input the user feature target representation into multiple neural networks to obtain a user preference feature representation; and a residual network submodule, used to input the user preference feature representation into a residual network to obtain a knowledge content recommendation table. In some possible embodiments of this application, the target representation submodule specifically includes: a dimensionality reduction processing unit, used to input the user feature matrix into an autoencoder to obtain a low-dimensional representation of user features; and a deep processing unit, used to input the low-dimensional representation of user features into a deep belief network to obtain a deep representation of user features; the user feature target representation includes at least one of the low-dimensional representation and the deep representation.

[0170] In some possible embodiments of this application, the user preference submodule specifically includes: a sequence information unit, used to input the user feature target representation into a recurrent neural network to obtain sequence information; and a user preference unit, used to input the sequence information into an attention network to obtain a user preference feature representation.

[0171] In some possible embodiments of this application, the user preference unit is specifically used to perform: integrating sequence information and content features into a three-dimensional array; and inputting the three-dimensional array into an attention network to obtain a user preference feature representation.

[0172] In some possible embodiments of this application, the sequence information unit is specifically used to perform: analyzing the user features to obtain a user behavior sequence; and inputting the user behavior sequence and the user feature target representation into a recurrent neural network to obtain sequence information.

[0173] In some possible embodiments of this application, the model recommendation module specifically includes: a graph generation submodule, used to analyze the user features to obtain a user knowledge graph; and a model recommendation submodule, used to input the user knowledge graph into a knowledge content recommendation model to obtain a knowledge content recommendation table.

[0174] In some possible embodiments of this application, the model recommendation submodule specifically includes: a vector embedding unit, used to input the knowledge graph into a graph convolutional layer to obtain corresponding embedding vectors; a collaborative filtering unit, used to input the embedding vectors into a neural collaborative filtering network to obtain the user's preference results for each knowledge content; and a table generation unit, used to generate a knowledge content recommendation table based on the user's preference results for each knowledge content.

[0175] In some possible embodiments of this application, the collaborative filtering unit is specifically used to perform: inputting the embedding vector into the interaction layer to obtain the interaction feature vector; inputting the interaction feature vector into multiple hidden layers to obtain the user's preference results for each knowledge content.

[0176] In some possible embodiments of this application, the graph generation submodule specifically includes: an entity name extraction unit, used to obtain the entity name of each entity, wherein the entity includes user entities and knowledge entities, and the entity name corresponds to the user entity name and the knowledge entity name; an entity relationship extraction unit, used to extract entity relationships from the user features based on the entity names, wherein the entity relationships are the relationships between entities; and a knowledge graph generation unit, used to establish a user knowledge graph based on the entity names and the entity relationships.

[0177] In this embodiment, by inputting user-related data features, content data features, and background data features into the knowledge content recommendation model for comprehensive analysis, a knowledge content recommendation table is generated that better matches the user's interests and needs. Based on this table, knowledge content is pushed to the user. In an environment with a large amount of information, this helps users quickly find knowledge content of interest, thereby reducing the time cost of information filtering. At the same time, the recommended content received by the user is closer to their interests, which can improve user satisfaction and user experience, increase user activity, and enable users to utilize fragmented time to learn and explore the knowledge they need.

[0178] Figure 8 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown.

[0179] It should be noted that, Figure 8 The computer system of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0180] like Figure 8As shown, the computer system includes a Central Processing Unit (CPU) 1801, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 1802 or programs loaded from storage portion 1808 into Random Access Memory (RAM) 1803, such as performing the methods described in the above embodiments. The RAM 1803 also stores various programs and data required for system operation. The CPU 1801, ROM 1802, and RAM 1803 are interconnected via a bus 1804. An Input / Output (I / O) interface 1805 is also connected to the bus 1804.

[0181] The following components are connected to I / O interface 1805: an input section 1806 including a keyboard, mouse, etc.; an output section 1807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1808 including a hard disk, etc.; and a communication section 1809 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 1809 performs communication processing via a network such as the Internet. A drive 1810 is also connected to I / O interface 1805 as needed. Removable media 1811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 1810 as needed so that computer programs read from them can be installed into storage section 1808 as needed.

[0182] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 1809, and / or installed from removable medium 1811. When the computer program is executed by central processing unit (CPU) 1801, it performs various functions defined in the system of this application.

[0183] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. The transmitted data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0184] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0185] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0186] In another aspect, this application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the methods described in the above embodiments.

[0187] This specification also provides a computer program product that stores at least one instruction, said at least one instruction being loaded and executed by the processor as described above. Figures 1-6 The method described in the illustrated embodiment can be found in the following document for a detailed execution process. Figures 1-6 The specific details of the illustrated embodiments will not be elaborated here.

[0188] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0189] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this application.

[0190] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0191] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for recommending and pushing knowledge content, characterized in that, include: Obtain user characteristics, which include user data characteristics, content data characteristics, and background data characteristics; The user features are input into the knowledge content recommendation model to obtain the knowledge content recommendation table; Based on the knowledge content recommendation table, knowledge content is pushed to users.

2. The knowledge content recommendation and push method as described in claim 1, characterized in that, The acquisition of user characteristics specifically includes: Acquire target user data, target content data, and target background data; Feature extraction is performed on the target user data, target content data, and target background data respectively to obtain the user data features, the content data features, and the background data features.

3. The knowledge content recommendation and push method as described in claim 2, characterized in that, The acquisition of target user data, target content data, and target background data specifically includes: Acquire user data, content data, and background data; The user data, content data, and background data are filtered separately to obtain target user data, target content data, and target background data.

4. The knowledge content recommendation and push method as described in claim 1, characterized in that, The step of inputting the user features into the knowledge content recommendation model to obtain the knowledge content recommendation table specifically includes: The user features are standardized and encoded to obtain a user feature matrix; The user feature matrix is ​​subjected to dimensionality reduction and depth processing to obtain the user feature target representation; The user feature target representation is sequentially input into multiple neural networks to obtain the user preference feature representation; The user preference features are input into the residual network to obtain a knowledge content recommendation table.

5. The knowledge content recommendation and push method as described in claim 4, characterized in that, The step of performing dimensionality reduction and depth processing on the user feature matrix to obtain the user feature target representation specifically includes: The user feature matrix is ​​input into the autoencoder to obtain a low-dimensional representation of the user features; The low-dimensional representation of the user features is input into a deep belief network to obtain a deep representation of the user features; The user feature target representation includes at least one of the following: a low-dimensional representation of user features and a deep representation of user features.

6. The knowledge content recommendation and push method as described in claim 4, characterized in that, The step of sequentially inputting the user feature target representation into multiple neural networks to obtain the user preference feature representation specifically includes: The user feature target representation is input into a recurrent neural network to obtain sequence information; The sequence information is input into the attention network to obtain the user preference feature representation.

7. A knowledge content recommendation and push device, characterized in that, The knowledge content recommendation and push device includes: The feature acquisition module is used to acquire user features, which include user data features, content data features, and background data features. The model recommendation module is used to input the user features into the knowledge content recommendation model to obtain a knowledge content recommendation table; The content push module is used to push knowledge content to users based on the knowledge content recommendation table.

8. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the knowledge content recommendation and push method as described in any one of claims 1 to 6.

9. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the knowledge content recommendation and push method as described in any one of claims 1 to 6.

10. A computer program product comprising one or more computer programs, characterized in that, When the one or more computer programs are executed by one or more processors, they implement the steps of the knowledge content recommendation and push method according to any one of claims 1 to 6.

Images