Content recommendation method and system based on semantic recognition

Abstract

The invention discloses a content recommendation method and system based on semantic recognition, and the method comprises the steps: analyzing the text, voice or video input of a user through a natural language processing technology, and generating user semantic representation data; the method comprises the following steps: dynamically constructing a user interest graph by combining user historical behavior data, and performing representation learning of nodes and edges through a multilayer graph convolutional network to generate time sequence user interest graph representation data; a self-attention mechanism and a multi-head attention mechanism of a Transform model are utilized, the matching degree of user interest representation and recommended content is optimized in combination with a contrast learning module, and personalized recommended content data are generated; collecting multi-modal information of the user, fusing the multi-modal information with the personalized recommendation content data, and generating optimized multi-modal recommendation content data through a multi-modal fusion algorithm; and collecting user feedback feature data, and dynamically adjusting and optimizing a generation strategy of recommended content through an adaptive learning algorithm. According to the invention, efficient, accurate and personalized recommendation of the recommendation system is realized, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information recommendation, and in particular to a content recommendation method and system based on semantic recognition. Background Art

[0002] In the modern information society, users are faced with a vast amount of information, and how to effectively recommend content that users are interested in has become an important research topic. Semantic recognition technology can understand users' real needs by analyzing their text, voice, and video input, thereby improving the relevance and accuracy of recommended content.

[0003] Current content recommendation technologies mainly rely on traditional methods such as collaborative filtering and content filtering. However, these methods have some inherent problems. Although collaborative filtering methods can capture the relationship between users and content, they perform poorly in data sparsity and cold start problems. Content filtering methods rely on the explicit features of content and have difficulty capturing users' implicit interests. In addition, existing technologies also have obvious deficiencies in processing multimodal information fusion and real-time feedback optimization.

[0004] The existing technology (Chinese invention patent, publication number: CN116340648A, name: A knowledge graph attention network recommendation method based on graph collaborative filtering) mainly adopts a knowledge graph attention network recommendation method based on graph collaborative filtering. The method includes constructing a user-item bipartite graph and using a collaborative filtering layer and a knowledge graph attention embedding layer for recommendation. However, the method has the following defects: Existing recommendation methods based on graph collaborative filtering and knowledge graph attention networks do not adequately address data sparsity and cold start issues; when user or item data is scarce, the accuracy of the recommendation system decreases significantly; The existing graph convolutional network has low training efficiency and slow convergence speed when processing large-scale graph data, which leads to a slow response speed of the recommendation system; Existing recommendation systems tend to generate homogeneous recommendations, lack diversity and personalization, and provide poor user experience. Combining the embedding of the collaborative filtering layer and the knowledge graph attention layer increases the amount of information, but also increases the complexity of the model and the difficulty of optimization. Summary of the invention

[0005] In view of the many problems existing in the above-mentioned prior arts, the present invention provides a content recommendation method and system based on semantic recognition. The present invention analyzes user needs through semantic recognition technology, dynamically constructs a user interest graph, and utilizes a multi-layer graph convolutional network and a Transformer model to perform node and edge representation learning and generate recommended content. By introducing multimodal information fusion and an adaptive learning algorithm, the diversity, personalization and real-time nature of the recommended content are ensured, thereby improving the overall effect of the recommendation system.

[0006] A content recommendation method based on semantic recognition includes the following steps: Analyze the needs of users through text, voice or video input through natural language processing technology to generate user semantic representation data; Based on user semantic representation data and user historical behavior data, dynamically build a user interest graph, and use a multi-layer graph convolutional network to learn the representation of nodes and edges to generate time series user interest graph representation data; The user interest graph of time series is used to represent data. Through the self-attention mechanism and multi-head attention mechanism of the Transformer model, combined with the contrastive learning module, the matching degree between the user interest representation and the recommended content is optimized to generate personalized recommended content data. Collect the user's multimodal information and fuse it with the personalized recommendation content data, generate multimodal recommendation data using the multimodal fusion algorithm, and generate optimized multimodal recommendation content data through the optimization algorithm; Display optimized multimodal recommended content data to users in real time, collect user feedback feature data, and dynamically adjust and optimize the recommended content generation strategy through adaptive learning algorithms to ensure the continued accuracy and personalization of recommended content.

[0007] Preferably, the parsing of the user's needs inputted via text, voice or video by natural language processing technology includes: Perform speech recognition on the speech data and convert it into text data; Perform speech and image analysis on video data, extract speech and image features, and generate video semantic data; Perform word segmentation, part-of-speech tagging and semantic analysis on text data to generate user semantic representation data.

[0008] Preferably, the dynamically constructing user interest graph includes: Extract the user's points of interest and generate point of interest data; combine the user's historical behavior data to build a user interest graph and generate user interest graph data; A multi-layer graph convolutional network is applied on the user interest graph to perform node and edge representation learning and generate user interest graph representation data.

[0009] Preferably, the learning of representation of nodes and edges by using a multi-layer graph convolutional network includes: Initialize the feature vector of each node in the user interest graph; Apply multi-layer graph convolution operations to aggregate the information of neighbor nodes layer by layer and update the feature vector of each node. The convolution operation of each layer is expressed as:

[0010] in, Representation Node In the The feature vector of the layer; For Node The set of neighbor nodes of is the normalization coefficient; For the The weight matrix of the layer; is the activation function; u represents the neighbor node of node v, that is, the node directly connected to node v; Represents the feature vector of node u at the k-1th layer, which is the feature calculated by the previous layer of the graph convolutional network; The attention mechanism is introduced to dynamically adjust the weights of the edges in the user interest graph and generate the adjusted user interest graph representation data. The edge weight calculation formula is:

[0011] in, For edge The weight of and Represents the nodes in the graph and The eigenvector of is the attention weight vector; represents the transpose operation, that is Representation vector The transpose of is a learnable parameter matrix used to linearly transform node feature vectors; ‖ represents a concatenation operation, which will be passed through the weight matrix The transformed node feature vectors are concatenated; It is an activation function used to enhance the nonlinear characteristics of the model; Use the time convolutional network to capture the time series information of user interests and generate time series feature data; The time series feature data is fused with the adjusted user interest graph representation data to generate the time series user interest graph representation data.

[0012] Preferably, the self-attention mechanism and multi-head attention mechanism through the Transformer model include: Convert the time series user interest graph representation data into the input format of the Transformer model; The self-attention mechanism is applied to calculate the importance of each node and weight the encoded user interest data. The calculation formula of the self-attention mechanism is:

[0013] in, represents the query vector; represents the key vector; represents a value vector; is the vector dimension; represents the transposition operation; Softmax represents the normalization function, which converts the input into a probability distribution to ensure that the sum of the output weights is 1; The multi-head attention mechanism is used to capture information of different levels and dimensions in the user interest graph. The calculation formula of the multi-head attention mechanism is:

[0014] Among them, each attention head The calculation method is:

[0015] in, is the weight matrix of the query vector; is the weight matrix of the key vector; is the weight matrix of the value vector; To output the weight matrix, comprehensive user interest data is generated; Concat represents the concatenation operation, which merges multiple vectors or matrices by dimension; Attention represents the self-attention mechanism, which is used to calculate the similarity between the query vector and the key vector and generate a weighted value vector.

[0016] Preferably, the optimizing the matching degree between the user interest representation and the recommended content in combination with the contrastive learning module comprises: Construct positive and negative sample pairs to generate positive and negative sample data; Calculate the distance between positive and negative sample data to generate sample distance data; The loss value is calculated according to the loss function of contrastive learning, which is defined as:

[0017] in, For sample and The distance between is a binary label; is the boundary distance, used to generate loss data; Using the optimization results of the comparative learning module, adjust the comprehensive user interest data to optimize the match between it and the recommended content; Generate personalized recommendation content data based on the optimized comprehensive user interest data.

[0018] Preferably, the collecting of multimodal information of users and fusing it with personalized recommendation content data includes: Collect users’ geographic location, device usage, and social network data as multimodal information; Preprocess the multimodal information to generate standardized multimodal feature data; Use a multimodal fusion algorithm to fuse standardized multimodal feature data and personalized recommendation content data to generate multimodal recommendation data; The multimodal recommendation data is optimized by using an optimization algorithm to ensure the relevance and personalization of the recommended content, and to generate optimized multimodal recommendation content data, wherein the optimization algorithm includes: a gradient descent method, a genetic algorithm, and a simulated annealing algorithm.

[0019] Preferably, the real-time display of optimized multimodal recommended content data to the user and the collection of user feedback feature data include: Real-time display of optimized multi-modal recommended content data; Collect user interaction data on recommended content, including click-through rate, dwell time, number of likes, and comment content; Preprocess user interaction data, extract key indicators and features, and generate user feedback feature data.

[0020] Preferably, the strategy for dynamically adjusting and optimizing the generation of recommended content through an adaptive learning algorithm includes: Based on user feedback feature data, the adaptive learning algorithm is used to dynamically adjust and optimize the generation strategy of recommended content to generate optimized recommended content data; Using the optimized recommended content data, updating the parameters and strategies of the recommended content generation strategy, including the adjustment of model weights, learning rates, and loss functions, to generate updated multimodal recommended content data; Apply the updated multimodal recommendation content data to the real-time recommendation generation module, dynamically adjust the recommendation strategy, and generate optimized real-time recommendation data for displaying and providing recommendation services to users.

[0021] And a system for implementing the content recommendation method based on semantic recognition. Compared with the prior art, the advantages and beneficial effects of the present invention are: The present invention analyzes user needs through semantic recognition technology and realizes more accurate user semantic representation data; The present invention constructs a dynamic user interest graph through a multi-layer graph convolutional network, achieving efficient learning of user interest representation; The present invention combines a contrastive learning module to optimize the matching degree between user interest representation and recommended content, generates personalized recommended content data, and improves the relevance of recommended content; The present invention collects and fuses user multimodal information, and uses a multimodal fusion algorithm to generate optimized multimodal recommendation content data to ensure the diversity and accuracy of the recommended content; The present invention dynamically adjusts and optimizes the generation strategy of recommended content through an adaptive learning algorithm, thereby achieving continuous optimization of the recommendation model and ensuring the real-time and personalization of the recommended content. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 A flowchart of user input parsing in the present invention; Figure 3 Construct a flow chart for the user interest graph in the present invention; Figure 4 This is a flow chart for optimizing the Transformer model in the present invention; Figure 5 This is a flow chart of multimodal information fusion in the present invention; Figure 6 A flowchart of real-time feedback and adaptive learning in the present invention; Figure 7 It is a structural block diagram of the system of the present invention. DETAILED DESCRIPTION

[0023] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0024] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0025] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0026] In the case of using expressions such as "at least one of A, B, and C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.). In the case of using expressions such as "at least one of A, B, or C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, or C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0027] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some boxes or combinations thereof in the block diagrams and / or flow charts can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions can create a device for implementing the functions / operations described in these block diagrams and / or flow charts when executed by the processor. The technology of the present disclosure can be implemented in the form of hardware and / or software (including firmware, microcode, etc.). In addition, the technology of the present disclosure can take the form of a computer program product on a computer-readable storage medium storing instructions, which can be used by an instruction execution system or in combination with an instruction execution system.

[0028] like Figure 1 As shown, a content recommendation method based on semantic recognition includes the following steps: like Figure 2 As shown, natural language processing technology is used to parse the needs of users input through text, voice or video, and generate user semantic representation data; Preferably, the parsing of the user's needs inputted via text, voice or video by natural language processing technology includes: Perform speech recognition on the speech data and convert it into text data; Perform speech and image analysis on video data, extract speech and image features, and generate video semantic data; Perform word segmentation, part-of-speech tagging and semantic analysis on text data to generate user semantic representation data.

[0029] In the present invention, parsing the needs of users input through text, voice or video is a key step, and its principle and effect are described as follows. The method uses natural language processing (NLP) technology to process different forms of data (text, voice, video) input by users to generate semantic representation data of users. This involves the integration and optimization of multiple technologies, aiming to extract semantic information from multimodal data, thereby achieving accurate content recommendation.

[0030] For voice data, the system first performs voice recognition to convert the voice content into processable text data. This step utilizes voice recognition technology to convert voice signals into text representations through acoustic models and language models, such as converting the user's spoken needs into text data that can be understood by the computer. The processing of video data is more complicated, and the system needs to analyze the voice content and image features in the video at the same time. Voice content analysis uses voice recognition technology to extract voice information from the video; image feature analysis uses computer vision technology, such as object recognition and motion detection, to extract visual information from the video. These analysis results are combined to generate the semantic data of the video, that is, the semantic representation of the video content.

[0031] For text data, the system will perform word segmentation, part-of-speech tagging, and semantic analysis. Word segmentation divides the text into meaningful word units; part-of-speech tagging determines the part of speech of each word; and semantic analysis analyzes the semantic relationship between words to generate semantic representation data of the text. This process relies on natural language processing technology, including word embedding models, part-of-speech taggers, and semantic analyzers, to effectively capture the semantic information of the text.

[0032] The user semantic representation data generated through the above steps has the following effects and advantages: it can process multiple user input forms (text, voice, video), thereby achieving unified understanding and processing of multimodal data; it improves the system's understanding of user input requirements and accuracy, avoiding information loss or misunderstanding in a single data form; based on accurate semantic representation data, the system can more accurately build user interest models and behavior patterns to achieve personalized content recommendations; because it can process multiple data forms, the system shows good flexibility and adaptability in real-time recommendation and feedback adjustment. For example, when a user inputs a question through voice, the system will first perform voice recognition and convert the voice into text; then, through word segmentation and semantic analysis, the system can understand and accurately capture the meaning of the question raised by the user; finally, combined with the user's historical behavior data and other contextual information, the system generates personalized recommendation content that suits the user's needs, thereby improving the user experience and recommendation accuracy.

[0033] like Figure 3As shown in the figure, based on user semantic representation data and user historical behavior data, a user interest graph is dynamically constructed, and the representation learning of nodes and edges is performed through a multi-layer graph convolutional network to generate time series user interest graph representation data; First, the system will dynamically build a user interest graph by combining the user's semantic representation data and historical behavior data. User semantic representation data comes from the analysis of user text, voice or video input, reflecting the user's current interests and needs. User historical behavior data records the user's interaction history on the platform, including clicks, browsing, likes, comments and other behaviors. These two types of data together form the basis of the user interest graph. By integrating these data, the system can form a dynamic and comprehensive user interest graph.

[0034] Next, the system learns the representations of nodes and edges in the user interest graph through a multi-layer graph convolutional network (GCN). Graph convolutional network is a deep learning model that can perform convolution operations directly on graph structured data, and is suitable for processing non-Euclidean spatial data such as user interest graphs. Each layer of GCN aggregates the information of the node and its neighboring nodes, and updates the feature representation of the node layer by layer. Specifically, the graph convolution operation captures the relationship between nodes and the complex patterns in the graph structure by weighted summing the feature vector of each node with the feature vector of its neighboring nodes, and then applying an activation function. For example, user A's interest node and its neighboring nodes (such as other users with similar interests or related content) are convolved through a multi-layer GCN, extracting and fusing features layer by layer, and finally generating a high-dimensional node representation.

[0035] In order to capture the time series changes in user interests, the system introduces a time series user interest graph to represent the data. By adding the time dimension to GCN, the system can track the changes in user interests over time. For example, a user may be particularly interested in a topic during a period of time, and pay more attention to other topics during another period of time. Through time series analysis, the system can identify and predict the changing trends of user interests, thereby achieving more accurate and timely content recommendations.

[0036] The effect of this process is that by dynamically building and learning user interest graphs, the system can capture users' multi-dimensional interests and behavior patterns, and update users' interest status in a timely manner. Combined with GCN's powerful representation learning capabilities and time series analysis, the system can not only understand users' current interests, but also predict users' possible changes in interests in the future, thereby providing personalized and forward-looking content recommendations. For example, when a user frequently browses a certain type of content within a certain period of time, the system can identify this change and recommend relevant content in a timely manner, improving user experience and recommendation accuracy.

[0037] In summary, by dynamically constructing a user interest graph based on user semantic representation data and historical behavior data, and using a multi-layer graph convolutional network to learn the representation of nodes and edges, the system can generate time series user interest graph representation data, thereby achieving a deep understanding and accurate prediction of user interests, and ultimately improving the personalization and effectiveness of content recommendations.

[0038] Preferably, the dynamically constructing user interest graph includes: Extract the user's points of interest and generate point of interest data; combine the user's historical behavior data to build a user interest graph and generate user interest graph data; A multi-layer graph convolutional network is applied on the user interest graph to perform node and edge representation learning and generate user interest graph representation data.

[0039] First of all, extracting the user's points of interest is the first step in dynamically building the user's interest graph. Points of interest refer to the attention and preference that users show on certain topics or content within a specific time period. For example, when a user frequently browses articles, videos or pictures related to travel, "travel" can be identified as a point of interest for the user. The method of extracting points of interest includes performing natural language processing (NLP) analysis on the user's text, voice or video input, identifying the user's current topics of interest through keyword extraction, topic modeling and other technologies, and generating point of interest data. These point of interest data constitute the basic nodes of the user's interest graph.

[0040] Next, the system combines the user's historical behavior data to build a user interest graph. Historical behavior data includes all the user's interaction records on the platform, such as clicks, browsing, likes, comments, etc. These data can reflect the user's long-term interest preferences and behavior patterns. By combining the point of interest data with the historical behavior data, the system can build a comprehensive user interest graph and generate user interest graph data. In this graph, nodes represent the user's points of interest, and edges represent the relationship between these points of interest, such as the correlation between multiple points of interest of the same user, or the connection between similar points of interest of different users.

[0041] In order to further improve the expressive power of the user interest graph, the system applies a multi-layer graph convolutional network (GCN) on the user interest graph to learn the representation of nodes and edges. Graph convolutional network is a deep learning model that can process graph structure data. It achieves efficient representation of node features by aggregating and learning the information of nodes and their neighbor nodes in the graph. In the present invention, GCN aggregates the information of nodes and their neighbor nodes in the user interest graph layer by layer through multi-layer convolution operations, and continuously updates the feature representation of nodes. For example, in the first layer of convolution operation, the feature vector of the node is weighted summed with the feature vector of its direct neighbor, and then a new node feature is generated through an activation function; in the second layer of convolution operation, the new node features and the information of its neighbor nodes are aggregated again, and so on, until the multi-layer convolution operation is completed. This process not only captures the local relationship between nodes, but also learns the deep patterns in the graph structure from a global perspective, and finally generates user interest graph representation data.

[0042] Through this process of dynamically constructing a user interest graph, the system is able to achieve comprehensive, dynamic, and accurate modeling of user interests. The user interest graph not only reflects the user's current interests, but also combines the user's historical behavior data to provide insights into changes and development trends in user interests. At the same time, through representation learning of multi-layer graph convolutional networks, the system can extract high-dimensional features from complex graph structures to achieve a deep understanding and representation of user interests. The effect of this process is that through efficient user interest modeling, the system can more accurately predict the user's content needs, thereby providing more personalized and accurate content recommendations.

[0043] For example, if user A has frequently browsed technology-related articles in the past period of time and has recently started to pay attention to content in the field of artificial intelligence, the system uses NLP technology to identify "technology" and "artificial intelligence" as user A's interests, and combines his historical behavior data to build a user interest graph containing these interests. Then, through GCN representation learning of the graph, the system can capture the rising trend of user A's interest in "artificial intelligence" and give priority to recommending more high-quality content related to artificial intelligence to user A in subsequent recommendations, thereby improving user experience and the effectiveness of recommendations.

[0044] Preferably, the learning of representation of nodes and edges by using a multi-layer graph convolutional network includes: Initialize the feature vector of each node in the user interest graph; Apply multi-layer graph convolution operations to aggregate the information of neighbor nodes layer by layer and update the feature vector of each node. The convolution operation of each layer is expressed as:

[0045] in, Representation Node In the The feature vector of the layer; For Node The set of neighbor nodes of is the normalization coefficient; For the The weight matrix of the layer; is the activation function; u represents the neighbor node of node v, that is, the node directly connected to node v; Represents the feature vector of node u at the k-1th layer, which is the feature calculated by the previous layer of the graph convolutional network; The attention mechanism is introduced to dynamically adjust the weights of the edges in the user interest graph and generate the adjusted user interest graph representation data. The edge weight calculation formula is:

[0046] in, For edge The weight of and Represents the nodes in the graph and The eigenvector of is the attention weight vector; represents the transpose operation, that is Representation vector The transpose of is a learnable parameter matrix used to linearly transform node feature vectors; ‖ represents a concatenation operation, which will be passed through the weight matrix The transformed node feature vectors are concatenated; It is an activation function used to enhance the nonlinear characteristics of the model; Use the time convolutional network to capture the time series information of user interests and generate time series feature data; The time series feature data is fused with the adjusted user interest graph representation data to generate the time series user interest graph representation data.

[0047] In this invention, learning the representation of nodes and edges through a multi-layer graph convolutional network (GCN) is a key step in achieving accurate content recommendation. The process includes initializing the feature vector of each node in the user interest graph, and then applying multi-layer graph convolution operations to aggregate the information of neighboring nodes layer by layer and update the feature vector of each node. In each layer, the feature vector of the node The calculation formula is shown above. In this way, GCN can effectively aggregate the information of each node and its neighboring nodes, thereby capturing the complex relationships in the graph structure.

[0048] On the basis of GCN, the attention mechanism is introduced to further dynamically adjust the weight of the edge in the user interest graph and generate the adjusted user interest graph representation data (using the edge weight calculation formula). The attention mechanism enables the model to dynamically adjust the edge weight according to the importance of the node, thereby more accurately reflecting the relationship between the nodes.

[0049] In addition, the temporal convolutional network (TCN) is used to capture the time series information of user interests and generate time series feature data. TCN effectively captures the dynamic pattern of user interests changing over time by performing convolution operations on the time dimension. This process ensures that the system can not only understand the user's current interests, but also predict the user's future interest change trends.

[0050] Finally, the time series feature data is fused with the adjusted user interest graph representation data to generate the time series user interest graph representation data. In this way, the system can integrate the temporal dynamics and structural relationships of user interests, thereby providing more personalized and accurate content recommendations.

[0051] First, GCN combined with the attention mechanism enables the system to extract high-dimensional features from complex graph structures and accurately capture the multi-dimensional relationships of user interests. Second, the introduction of TCN enables the system to dynamically track and predict the changing trends of user interests. Overall, by integrating time series features with graph structure features, the system can achieve comprehensive, dynamic, and in-depth modeling of user interests, thereby providing personalized and accurate content recommendations.

[0052] For example, if user A has shown a high interest in technology and AI-related content in the past few months and has recently started to pay attention to the field of quantum computing, the system will first parse the user's text, voice, and video input through NLP technology to identify user A's current interests. Then, by combining user A's historical behavior data, a user interest graph containing interests such as "technology", "artificial intelligence", and "quantum computing" is constructed. With the help of GCN and attention mechanisms, the system can dynamically adjust the weights of nodes and edges in the graph to accurately reflect the structure and intensity of user A's interests. Finally, by capturing the temporal dynamics of user interests through TCN, the system can predict what content user A may be more interested in in the future, thereby giving priority to high-quality content related to quantum computing in recommendations, improving user experience and recommendation accuracy.

[0053] like Figure 4As shown in the figure, the time series user interest graph is used to represent the data. Through the self-attention mechanism and multi-head attention mechanism of the Transformer model, the matching degree between the user interest representation and the recommended content is optimized in combination with the contrastive learning module to generate personalized recommended content data. The system uses a time series user interest graph to represent data and processes it through the self-attention mechanism of the Transformer model. The core of the self-attention mechanism is that it can calculate the importance of each element (in this case, the node in the user interest graph) to other elements and perform a weighted sum. This mechanism allows the model to focus on the entire time series data while processing the current data, thereby capturing the dynamic changes in user interests. For example, when calculating the user's current interest in a certain content, the self-attention mechanism can refer to the user's related interest changes over the past period of time and comprehensively evaluate the intensity and relevance of the current interest.

[0054] The multi-head attention mechanism further enhances this capability. By dividing the attention mechanism into multiple heads, each head independently learns different relationships and features, and then combining this information, the model can more comprehensively capture the complex patterns in the user's interest graph. Each attention head processes a part of the information independently, and finally the outputs of all heads are spliced ​​together to form the final representation. This approach enables the model to process information at different levels and dimensions in parallel, thereby improving the accuracy and diversity of recommendations.

[0055] In order to optimize the match between user interest representation and recommended content, the system incorporates a contrastive learning module. Contrastive learning learns better representations by constructing positive and negative sample pairs. A positive sample pair refers to the relationship between the content that the user is actually interested in and his interest representation, while a negative sample pair refers to the relationship between the content that the user is not interested in and his interest representation. By calculating the distance between positive and negative sample pairs, the system can optimize the representation so that the distance between positive sample pairs is closer and the distance between negative sample pairs is farther, thereby improving the accuracy of recommendations. Specifically, the loss function is used to measure the performance of the model in distinguishing between positive and negative sample pairs. By minimizing the loss function, the model can continuously optimize its own parameters to achieve better matching results.

[0056] The effect of this process is that through the self-attention mechanism and multi-head attention mechanism of the Transformer model, the system can extract accurate interest representations from the time series user interest graph, and optimize the matching degree of recommended content by combining contrastive learning, and finally generate personalized recommended content data. For example, when user B frequently browses health-related articles in the past period of time and recently begins to pay attention to mental health content, the system captures the changing trend of this interest through the self-attention mechanism, and comprehensively processes the user's multi-dimensional interest data through the multi-head attention mechanism. Combined with the contrastive learning module, the system can optimize the interest representation of user B, so that the recommended mental health content is more in line with the user's needs and preferences, thereby improving the accuracy of the recommendation and user satisfaction.

[0057] In this way, the system can not only dynamically track changes in user interests, but also significantly improve the accuracy and relevance of recommended content, ultimately providing personalized, high-quality recommendation services.

[0058] Preferably, the self-attention mechanism and multi-head attention mechanism through the Transformer model include: Convert the time series user interest graph representation data into the input format of the Transformer model; The self-attention mechanism is applied to calculate the importance of each node and weight the encoded user interest data. The calculation formula of the self-attention mechanism is:

[0059] in, represents the query vector; represents the key vector; represents a value vector; is the vector dimension; represents the transposition operation; Softmax represents the normalization function, which converts the input into a probability distribution to ensure that the sum of the output weights is 1; The multi-head attention mechanism is used to capture information of different levels and dimensions in the user interest graph. The calculation formula of the multi-head attention mechanism is:

[0060] Among them, each attention head The calculation method is:

[0061] in, is the weight matrix of the query vector; is the weight matrix of the key vector; is the weight matrix of the value vector; To output the weight matrix, comprehensive user interest data is generated; Concat represents the concatenation operation, which merges multiple vectors or matrices by dimension; Attention represents the self-attention mechanism, which is used to calculate the similarity between the query vector and the key vector and generate a weighted value vector.

[0062] The time series user interest graph representation data is converted into the input format of the Transformer model. The input of the Transformer model usually includes a query vector (Q), a key vector (K), and a value vector (V), which represent the nodes and their relationships in the user interest graph. The self-attention mechanism weights the encoded user interest data by calculating the importance of each node. The self-attention mechanism can refer to other node information in the entire graph when processing the current node information, thereby capturing global relationships and patterns.

[0063] The multi-head attention mechanism further enhances this capability. By dividing the attention mechanism into multiple heads, each head independently learns different relationships and features, and then combining this information, the model can more comprehensively capture the complex patterns in the user interest graph. In this way, the multi-head attention mechanism can process information at different levels and dimensions in parallel to generate comprehensive user interest data.

[0064] Through the self-attention mechanism and multi-head attention mechanism of the Transformer model, the system can extract accurate interest representations from the time series user interest graph. This process allows the model to comprehensively consider all node information in the entire time series when calculating the representation of the current node, thereby capturing the dynamic changes in user interests. For example, when a user has shown different interests in different categories of content in the past few months, the model can identify the trends of these interest changes through the self-attention mechanism and comprehensively process the user's multi-dimensional interest data through the multi-head attention mechanism.

[0065] Combined with the contrastive learning module, the system can further optimize the match between user interest representation and recommended content. Contrastive learning learns better representations by constructing positive and negative sample pairs. Positive sample pairs refer to the relationship between the content that the user is actually interested in and his interest representation, while negative sample pairs refer to the relationship between the content that the user is not interested in and his interest representation. By calculating the distance between positive and negative sample pairs, the system can optimize the representation so that the distance between positive sample pairs is closer and the distance between negative sample pairs is farther, thereby improving the accuracy of recommendations.

[0066] This method is very effective. Through the self-attention mechanism and the multi-head attention mechanism, the system can extract accurate interest representations from the time series user interest graph, and combine the contrastive learning module to optimize the match between user interest representations and recommended content, thereby generating personalized recommended content data. For example, when a user frequently browses technology and health-related content in the past period of time, the system can identify the user's multi-dimensional interests through the above mechanism, and give priority to displaying high-quality content in these categories in the recommendation, thereby improving the user experience and the effectiveness of the recommendation.

[0067] In summary, through the self-attention mechanism and multi-head attention mechanism of the Transformer model, combined with the contrastive learning module, the system can achieve a deep understanding of user interests and accurate recommendations, and ultimately generate personalized recommended content.

[0068] Preferably, the optimizing the matching degree between the user interest representation and the recommended content in combination with the contrastive learning module comprises: Construct positive and negative sample pairs to generate positive and negative sample data; Calculate the distance between positive and negative sample data to generate sample distance data; The loss value is calculated according to the loss function of contrastive learning, which is defined as:

[0069] in, For sample and The distance between is a binary label; is the boundary distance, used to generate loss data; Using the optimization results of the comparative learning module, adjust the comprehensive user interest data to optimize the match between it and the recommended content; Generate personalized recommendation content data based on the optimized comprehensive user interest data.

[0070] The system generates positive and negative sample data by constructing positive and negative sample pairs. A positive sample pair refers to the relationship between the content that the user is actually interested in and his or her interest expression, while a negative sample pair refers to the relationship between the content that the user is not interested in and his or her interest expression. In this way, the system can learn from the user's actual feedback and distinguish between the content that the user is interested in and the content that he or she is not interested in. For example, user A clicks on and browses an article about health, which forms a positive sample pair with user A's interest expression; while user A shows no interest in another article about sports, which forms a negative sample pair with user A's interest expression.

[0071] The system calculates the distance between positive and negative sample data and generates sample distance data. The distance between samples reflects the degree of match between the user's interest representation and the recommended content. For positive sample pairs, the closer the distance, the higher the match; for negative sample pairs, the farther the distance, the lower the match. This step is achieved by calculating the distance matrix to ensure that the system can quantify the similarity between each pair of samples.

[0072] The system calculates the loss value based on the contrastive learning loss function. This loss function optimizes the model parameters by maximizing the similarity of positive sample pairs and minimizing the similarity of negative sample pairs. Specifically, for positive sample pairs, if the distance is less than the boundary distance , the loss value is zero; if the distance is greater than the boundary distance, the loss value increases. For negative sample pairs, the larger the distance, the smaller the loss value. This design enables the model to effectively learn to distinguish between positive and negative sample pairs, thereby improving the accuracy of recommendations.

[0073] The system uses the optimization results of the contrastive learning module to adjust the user interest data. By minimizing the loss function, the system can continuously optimize the user interest representation to more accurately reflect the user's actual interests. The adjusted user interest data is further used to generate personalized recommendation content, thereby improving the effectiveness of the recommendation system. For example, after contrastive learning optimization, the system can more accurately identify user A's interest in health-related content and recommend more relevant high-quality content.

[0074] Combined with the contrastive learning module to optimize the match between user interest representation and recommended content, the system can achieve a deep understanding of user interests and accurate recommendations by constructing positive and negative sample pairs, calculating sample distances, optimizing loss functions, and adjusting user interest data. This method not only improves the relevance and personalization of recommended content, but also continuously improves the performance of the recommendation system through continuous learning and optimization. In this way, users can obtain personalized recommendations that better suit their interests and needs, improving user experience and satisfaction.

[0075] Collect the user's multimodal information and fuse it with the personalized recommendation content data, generate multimodal recommendation data using the multimodal fusion algorithm, and generate optimized multimodal recommendation content data through the optimization algorithm; Preferably, the collecting of multimodal information of users and fusing it with personalized recommendation content data includes: Collect users’ geographic location, device usage, and social network data as multimodal information; Preprocess the multimodal information to generate standardized multimodal feature data; Use a multimodal fusion algorithm to fuse standardized multimodal feature data and personalized recommendation content data to generate multimodal recommendation data; The multimodal recommendation data is optimized by using an optimization algorithm to ensure the relevance and personalization of the recommended content, and to generate optimized multimodal recommendation content data, wherein the optimization algorithm includes: a gradient descent method, a genetic algorithm, and a simulated annealing algorithm.

[0076] The system collects multimodal information about users, including but not limited to their geographic location, device usage, and social network data. Geographic location data can reflect the user's current activity area and location preferences. For example, a user may be more inclined to accept content recommendations related to his or her current location. Device usage data involves the type of device, operating system, and usage habits used by the user. This information can help the system understand the user's technical preferences and usage environment. Social network data includes the user's activities, interest groups, and interaction records on social platforms. This information can reveal the user's social behavior and interest preferences.

[0077] Preprocess the collected multimodal information to generate standardized multimodal feature data. The preprocessing steps include data cleaning, normalization, and feature extraction, etc. The purpose is to convert multimodal information into a unified feature representation to facilitate subsequent fusion and processing. For example, geographic location data can be represented by coordinate normalization, device usage can be represented by device type encoding, and social network data can be represented by interest tag extraction and user relationship graph construction.

[0078] The standardized multimodal feature data is fused with the personalized recommendation content data using a multimodal fusion algorithm to generate multimodal recommendation data. The core of the multimodal fusion algorithm is the ability to effectively integrate feature information from different sources to capture the user's comprehensive interests and needs. Commonly used multimodal fusion methods include feature-level fusion, decision-level fusion, and hybrid-level fusion. For example, feature-level fusion forms a comprehensive feature vector by splicing feature vectors of different modalities together to represent the user's overall interests.

[0079] Multimodal recommendation data is optimized through optimization algorithms to ensure the relevance and personalization of recommended content. Optimization algorithms can include gradient descent, genetic algorithms, and simulated annealing algorithms. The goal of these algorithms is to find the optimal recommendation parameters and strategies through iterative optimization, thereby generating optimized multimodal recommendation content data. For example, the gradient descent method minimizes the recommendation error function and continuously adjusts the parameters of the recommendation model to make the recommendation results closer to the actual needs of users; the genetic algorithm simulates natural selection and genetic variation to find the global optimal recommendation strategy; the simulated annealing algorithm simulates the physical annealing process to find the optimal solution in the high-dimensional search space.

[0080] By collecting and integrating users' multimodal information, the system can fully understand users' interests and needs, thereby providing more accurate and personalized content recommendations. For example, when a user frequently participates in travel-related discussions on social networks, and their geographic location data shows that the user is planning a trip, and their device usage shows that the user often uses mobile devices to browse travel information, the system can integrate this information, generate personalized travel-related recommendations through multimodal fusion algorithms, and ensure the relevance and personalization of the recommended content through optimization algorithms, thereby improving the user experience and satisfaction.

[0081] In this way, the system can achieve a comprehensive understanding and dynamic adjustment of user interests, thereby providing personalized, high-precision content recommendation services to meet the diverse needs of users in different situations.

[0082] Display optimized multimodal recommended content data to users in real time, collect user feedback feature data, and dynamically adjust and optimize the recommended content generation strategy through adaptive learning algorithms to ensure the continued accuracy and personalization of recommended content.

[0083] like Figure 5 As shown, the system will display the optimized multimodal recommended content data to the user in real time. This step involves the design of the user interface and the presentation of recommended content. The optimized multimodal recommended content data is the result of multimodal information fusion and optimization algorithm processing, which ensures the high relevance and personalization of the recommended content. For example, when a user opens a content recommendation platform, the system will generate and display the content that best suits the user's current interests in real time based on the user's historical behavior, current environment and multimodal information. The advantage of real-time display is that it can respond to changes in user needs immediately and improve user satisfaction and engagement.

[0084] While displaying recommended content, the system will collect user feedback feature data. This data includes the user's click-through rate, dwell time, number of likes, and comment content of the recommended content. Through this data, we can understand the user's actual reaction and satisfaction with the recommended content. For example, if a user frequently clicks and browses a certain type of recommended content, the system will consider that this type of content meets the user's interests and record these behavior data as feedback features. The collected feedback feature data provides valuable reference information for subsequent recommendation optimization.

[0085] Through the adaptive learning algorithm, the system dynamically adjusts and optimizes the strategy for generating recommended content. The core of the adaptive learning algorithm is its ability to adjust model parameters and recommendation strategies in real time based on user feedback feature data to continuously improve the recommendation effect. Specifically, the adaptive learning algorithm analyzes user feedback feature data to identify which recommendation strategies are effective and which need to be adjusted. For example, the system may find that certain content is more popular with users during a specific time period, thereby increasing the weight of such content in future recommendations. Through continuous iteration and optimization, the adaptive learning algorithm enables the recommendation system to dynamically adapt to changes in user interests and generate more accurate and personalized recommended content.

[0086] Through this process, the system can achieve the following effects: First, by real-time display of optimized multimodal recommendation content data, the system can instantly meet the user's personalized needs and improve user engagement and satisfaction; second, by collecting and analyzing user feedback feature data, the system can deeply understand the user's interests and behavior patterns, and provide data support for subsequent recommendation optimization; finally, through the dynamic adjustment and optimization of the adaptive learning algorithm, the system can continuously improve the accuracy and personalization of the recommended content, ensuring the continuous optimization of the recommendation effect.

[0087] For example, when user C opens the content recommendation platform, the system generates and displays a series of content related to user C's interests in real time based on user C's historical behavior and current environment. During the browsing process, user C shows strong interest in some recommended content and interacts with it multiple times (such as likes and comments). The system collects these feedback feature data, analyzes user C's behavior patterns, and adjusts the recommendation strategy through an adaptive learning algorithm. In the next recommendation, the system will prioritize content that is more in line with user C's interests based on the optimized strategy, thereby further improving user experience and satisfaction.

[0088] In summary, by displaying optimized multimodal recommendation content data in real time, collecting user feedback feature data, and dynamically adjusting and optimizing the generation strategy of recommended content through adaptive learning algorithms, the system can achieve continuous, accurate and personalized content recommendations, effectively improving user satisfaction and loyalty.

[0089] Preferably, the real-time display of optimized multimodal recommended content data to the user and the collection of user feedback feature data include: Real-time display of optimized multi-modal recommended content data; Collect user interaction data on recommended content, including click-through rate, dwell time, number of likes, and comment content; Preprocess user interaction data, extract key indicators and features, and generate user feedback feature data.

[0090] The system will display optimized multimodal recommended content data to users in real time. Multimodal recommended content data is generated by integrating users' multimodal information (such as geographic location, device usage, social network data, etc.) and personalized recommended content data, and processed by optimization algorithms. These data reflect users' current interests and needs, ensuring the relevance and personalization of recommended content. For example, when a user opens a content recommendation platform, the system will generate and display content that best suits the user's interests, such as the latest news, articles of interest, or videos, in real time based on the user's historical behavior and current environment.

[0091] When users browse and interact with recommended content, the system will collect user interaction data. Interaction data includes click-through rate, dwell time, number of likes, and comment content. These data can reflect the user's actual reaction and satisfaction with the recommended content. For example, when a user clicks on an article, the system records the click behavior; the time the user stays on the article page reflects the attractiveness of the content; likes and comments indicate the user's recognition and participation in the content. These interaction data provide an important basis for the system to understand the user's interest preferences and behavior patterns.

[0092] Next, the system preprocesses the collected user interaction data, extracts key indicators and features, and generates user feedback feature data. The preprocessing steps include data cleaning, normalization, and feature extraction, with the goal of converting raw interaction data into standardized and structured feature representations. For example, the click-through rate can be obtained by calculating the ratio of the number of times a recommended content is clicked to the number of times it is displayed; the dwell time can be expressed by the number of seconds a user stays on the page; and the number of likes and comments can be directly used as an indicator of user engagement. Through these preprocessing steps, the system can extract the most valuable feature data for recommendation optimization.

[0093] By displaying optimized multimodal recommendation content data in real time, collecting and preprocessing user feedback feature data, the system can dynamically capture and accurately model user interests. The effect of this process is that by displaying the most relevant and personalized content, the system can instantly meet user needs and improve user engagement and satisfaction; by collecting and analyzing user interaction data, the system can continuously learn and optimize recommendation strategies to ensure the continued accuracy and personalization of recommended content.

[0094] For example, after user D opens the content recommendation platform, the system will display a series of personalized recommended content in real time based on user D's historical behavior and multimodal information. User D clicked and browsed an article about artificial intelligence and stayed on the page for a long time, indicating a strong interest. At the same time, user D liked and commented on this article. The system will collect these interaction data and extract key indicators such as click-through rate, dwell time, number of likes, and comment content through preprocessing. By analyzing these user feedback feature data, the system recognizes user D's high interest in artificial intelligence and prioritizes displaying more relevant content in future recommendations, thereby improving the accuracy of recommendations and user satisfaction.

[0095] In summary, by displaying optimized multimodal recommendation content data in real time, collecting and preprocessing user feedback feature data, the system can dynamically track and accurately model user interests, thereby providing continuous, accurate and personalized content recommendation services. This process not only improves the effectiveness of the recommendation system, but also significantly improves user experience and satisfaction.

[0096] like Figure 6 As shown, preferably, the strategy for dynamically adjusting and optimizing the generation of recommended content through an adaptive learning algorithm includes: Based on user feedback feature data, the adaptive learning algorithm is used to dynamically adjust and optimize the generation strategy of recommended content to generate optimized recommended content data; Using the optimized recommended content data, updating the parameters and strategies of the recommended content generation strategy, including the adjustment of model weights, learning rates, and loss functions, to generate updated multimodal recommended content data; Apply the updated multimodal recommendation content data to the real-time recommendation generation module, dynamically adjust the recommendation strategy, and generate optimized real-time recommendation data for displaying and providing recommendation services to users.

[0097] Based on user feedback feature data, the system uses an adaptive learning algorithm to dynamically adjust and optimize the generation strategy of recommended content. User feedback feature data includes the user's click-through rate, dwell time, number of likes, and comment content of the recommended content. Through these data, we can understand the user's actual reaction and satisfaction with the recommended content. The core of the adaptive learning algorithm is its ability to adjust model parameters and recommendation strategies in real time based on these feedback data. For example, if the system detects that the user shows a higher interest in a certain type of content, it will increase the weight of this type of content in the recommendation and generate optimized recommended content data. This step enables the system to dynamically respond to changes in user interests by analyzing real-time feedback from users, thereby improving the relevance and personalization of recommendations.

[0098] Using the optimized recommended content data, the system updates the parameters and strategies of the recommended content generation strategy. Specifically, this includes adjustments to the model weights, learning rate, and loss function. Model weights determine the degree of influence of each feature in the recommendation model; the learning rate controls the pace of model parameter updates; and the loss function is used to measure the accuracy of the recommendation results. With the optimized recommended content data, the system is able to identify and adjust these parameters so that the recommendation model can better adapt to the user's changing interests and needs. For example, the system may increase the weight of certain features to make them occupy a more important position in the recommendation results; or adjust the learning rate to make the model update more stable and efficient.

[0099] Apply the updated multimodal recommendation content data to the real-time recommendation generation module, dynamically adjust the recommendation strategy, and generate optimized real-time recommendation data. The role of the real-time recommendation generation module is to generate and display the content that best suits the user's current interests based on the latest recommendation strategy. By applying the updated multimodal recommendation content data to this module, the system can ensure the continued accuracy and personalization of recommended content. For example, every interaction of the user on the platform will be recorded and analyzed in real time. The system will immediately adjust the recommendation results based on the latest feedback data and optimization strategies to show the most relevant and personalized content to the user.

[0100] The effect of this process is remarkable. Through the adaptive learning algorithm, the system can continuously learn and dynamically adjust user interests and behaviors. First, based on the user feedback feature data, the system can instantly identify and respond to changes in user interests, and improve the relevance and satisfaction of recommended content; second, by optimizing the parameters and strategies of the recommendation content data update generation strategy, the system can continuously improve the performance and accuracy of the recommendation model; finally, by applying the updated recommendation content data to the real-time recommendation generation module, the system can ensure that each recommendation can instantly reflect the user's latest interests and needs, thereby providing continuous accurate and personalized recommendation services.

[0101] For example, when user E browses a series of articles about technological innovation on the content recommendation platform and likes and comments on several of them, the system will collect these feedback feature data, analyze the changes in user E's interests through adaptive learning algorithms, and dynamically adjust the recommendation strategy. The optimized recommended content data will reflect user E's high interest in technological innovation. The system will update the parameters and strategies of the recommendation model to increase the weight of technological innovation content. The next time user E visits the platform, the system will prioritize recommending more high-quality content related to technological innovation to him through the real-time recommendation generation module, thereby improving user experience and satisfaction.

[0102] In general, by dynamically adjusting and optimizing the generation strategy of recommended content through adaptive learning algorithms, the system can achieve continuous learning and dynamic adjustment of user interests, ensure the continuous accuracy and personalization of recommended content, and effectively improve user satisfaction and loyalty.

[0103] like Figure 7 As shown, a system for implementing the content recommendation method based on semantic recognition includes: The natural language processing module is configured to parse the needs of users input through text, voice or video, and generate user semantic representation data; the natural language processing module is configured to parse the needs of users input through text, voice or video, and generate user semantic representation data. This module uses natural language processing technology to parse and process the user's input data. For example, the text data input by the user is segmented, part-of-speech tagged and semantically parsed to generate structured user semantic representation data. For voice input, the module uses speech recognition technology to convert speech into text and further parse its semantic content. For video input, voice and image features are extracted through voice and image analysis to generate video semantic data. In this way, the natural language processing module can accurately capture the user's intentions and improve the recommendation system's ability to understand and respond to user needs.

[0104] The user interest graph construction module dynamically constructs the user interest graph based on the user semantic representation data and the user historical behavior data, and learns the representation of nodes and edges through a multi-layer graph convolutional network to generate the time series user interest graph representation data; The user interest graph construction module dynamically constructs the user interest graph based on the user semantic representation data and the user historical behavior data, and learns the representation of nodes and edges through a multi-layer graph convolutional network to generate the time series user interest graph representation data. This module first extracts the user's interest point data and combines it with the user's historical behavior data to construct the user interest graph. On the user interest graph, a multi-layer graph convolutional network (GCN) is applied to learn the representation of nodes and edges. Through multi-layer convolution operations, the information of neighbor nodes is aggregated layer by layer to update the feature vector of each node. At the same time, the attention mechanism is introduced to dynamically adjust the weight of the edge in the user interest graph to generate the adjusted user interest graph representation data. The time series information of the user's interest is captured by the temporal convolutional network (TCN), and the time series feature data is fused with the adjusted user interest graph representation data to generate the time series user interest graph representation data. This process not only improves the accuracy of user interest representation, but also effectively deals with data sparsity and cold start problems.

[0105] The Transformer model module uses the time series user interest graph to represent data, and optimizes the matching degree between user interest representation and recommended content through self-attention mechanism and multi-head attention mechanism, combined with contrastive learning module, to generate personalized recommended content data; The Transformer model module uses the time series user interest graph to represent data, and optimizes the matching degree between user interest representation and recommended content through self-attention mechanism and multi-head attention mechanism, combined with contrastive learning module, to generate personalized recommended content data. This module first converts the time series user interest graph representation data into the input format of the Transformer model. Through the self-attention mechanism, the importance of each node is calculated, and the encoded user interest data is weighted. The multi-head attention mechanism further captures the information of different levels and dimensions in the user interest graph to generate comprehensive user interest data. During the training process, the contrastive learning module is introduced to optimize the matching degree between user interest representation and recommended content by constructing positive and negative sample pairs and calculating their distance, ensuring that the recommended content matches the user's preferences with high similarity, and generating personalized recommended content data. This process not only improves the relevance and personalization of the recommended content, but also solves the problems of homogeneity and high difficulty of model optimization in the recommendation system.

[0106] The multimodal information collection and fusion module collects the multimodal information of users, fuses it with the personalized recommendation content data, generates multimodal recommendation data using the multimodal fusion algorithm, and generates optimized multimodal recommendation content data through the optimization algorithm; The multimodal information collection and fusion module collects the multimodal information of users, fuses it with the personalized recommendation content data, generates multimodal recommendation data using the multimodal fusion algorithm, and generates optimized multimodal recommendation content data through the optimization algorithm. This module first collects multimodal information such as the user's geographic location, device usage, and social network data, and preprocesses these data to generate standardized multimodal feature data. Next, the standardized multimodal feature data is fused with the personalized recommendation content data using the multimodal fusion algorithm to generate multimodal recommendation data. The multimodal recommendation data is optimized through optimization algorithms (such as gradient descent, genetic algorithm, and simulated annealing algorithm) to ensure the relevance and personalization of the recommended content, and finally generates optimized multimodal recommendation content data. This process can effectively improve the diversity and accuracy of the recommended content and enhance the user experience.

[0107] The real-time display and feedback collection module displays the optimized multimodal recommendation content data to users in real time and collects user feedback feature data; the real-time display and feedback collection module is responsible for displaying the optimized multimodal recommendation content data to users in real time and collecting user feedback feature data. This module first presents the optimized multimodal recommendation content data to users in real time, and collects user feedback feature data through the user's interaction data on the recommended content (such as click-through rate, dwell time, number of likes, comment content, etc.). The user interaction data is preprocessed to extract key indicators and features to generate user feedback feature data. This process can help the system understand the user's satisfaction and preference for the recommended content in a timely manner, and provide basic data support for subsequent adaptive learning and recommendation strategy optimization.

[0108] The adaptive learning module dynamically adjusts and optimizes the generation strategy of recommended content through an adaptive learning algorithm based on user feedback feature data to ensure the continuous accuracy and personalization of recommended content. The adaptive learning module dynamically adjusts and optimizes the generation strategy of recommended content through an adaptive learning algorithm based on user feedback feature data to ensure the continuous accuracy and personalization of recommended content. The module first uses user feedback feature data to dynamically adjust the parameters and strategies of the recommendation content generation strategy, including the adjustment of model weights, learning rate and loss function, to generate optimized recommended content data. Next, the optimized recommended content data is applied to the real-time recommendation generation module, the recommendation strategy is dynamically adjusted, and the optimized real-time recommendation data is generated for displaying and providing recommendation services to users. Through this process, the adaptive learning module can continuously optimize the recommendation model and strategy, ensure the real-time, accuracy and personalization of recommended content, and improve user satisfaction and user experience.

[0109] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0110] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0111] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0113] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0114] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0115] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0116] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0117] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A content recommendation method based on semantic recognition, characterized in that: The following steps are involved: Analyze the needs of users through text, voice or video input through natural language processing technology to generate user semantic representation data; Based on user semantic representation data and user historical behavior data, dynamically build a user interest graph, and use a multi-layer graph convolutional network to learn the representation of nodes and edges to generate time series user interest graph representation data; The user interest graph of time series is used to represent data. Through the self-attention mechanism and multi-head attention mechanism of the Transformer model, combined with the contrastive learning module, the matching degree between the user interest representation and the recommended content is optimized to generate personalized recommended content data. Collect the user's multimodal information and fuse it with the personalized recommendation content data, generate multimodal recommendation data using the multimodal fusion algorithm, and generate optimized multimodal recommendation content data through the optimization algorithm; Display optimized multimodal recommended content data to users in real time, collect user feedback feature data, and dynamically adjust and optimize the recommended content generation strategy through adaptive learning algorithms to ensure the continued accuracy and personalization of recommended content.

2. The method according to claim 1, characterized in that The requirements of analyzing the user's input through text, voice or video by natural language processing technology include: Perform speech recognition on the speech data and convert it into text data; Perform speech and image analysis on video data, extract speech and image features, and generate video semantic data; Perform word segmentation, part-of-speech tagging and semantic analysis on text data to generate user semantic representation data.

3. The method according to claim 1, characterized in that The dynamically constructing user interest graph includes: Extract the user's points of interest and generate point of interest data; combine the user's historical behavior data to build a user interest graph and generate user interest graph data; A multi-layer graph convolutional network is applied on the user interest graph to perform node and edge representation learning and generate user interest graph representation data.

4. The method according to claim 3, characterized in that The representation learning of nodes and edges by using a multi-layer graph convolutional network includes: Initialize the feature vector of each node in the user interest graph; Apply multi-layer graph convolution operations to aggregate the information of neighbor nodes layer by layer and update the feature vector of each node. The convolution operation of each layer is expressed as: in, Representation Node In the The feature vector of the layer; For Node The set of neighbor nodes; is the normalization coefficient; For the The weight matrix of the layer; is the activation function; u represents the neighbor node of node v, that is, the node directly connected to node v; Represents the feature vector of node u at the k-1th layer, which is the feature calculated by the previous layer of the graph convolutional network; The attention mechanism is introduced to dynamically adjust the weights of the edges in the user interest graph and generate the adjusted user interest graph representation data. The edge weight calculation formula is: in, For edge The weight of and Represents the nodes in the graph and The eigenvector of is the attention weight vector; represents the transpose operation, that is Representation vector The transpose of is a learnable parameter matrix used to linearly transform node feature vectors; ‖ represents a concatenation operation, which will be passed through the weight matrix The transformed node feature vectors are concatenated; It is an activation function used to enhance the nonlinear characteristics of the model; Use the time convolutional network to capture the time series information of user interests and generate time series feature data; The time series feature data is fused with the adjusted user interest graph representation data to generate the time series user interest graph representation data.

5. The method according to claim 1, characterized in that The self-attention mechanism and multi-head attention mechanism through the Transformer model include: Convert the time series user interest graph representation data into the input format of the Transformer model; The self-attention mechanism is applied to calculate the importance of each node and weight the encoded user interest data. The calculation formula of the self-attention mechanism is: in, represents the query vector; represents the key vector; represents a value vector; is the vector dimension; represents the transposition operation; Softmax represents the normalization function, which converts the input into a probability distribution to ensure that the sum of the output weights is 1; The multi-head attention mechanism is used to capture information of different levels and dimensions in the user interest graph. The calculation formula of the multi-head attention mechanism is: Among them, each attention head The calculation method is: in, is the weight matrix of the query vector; is the weight matrix of the key vector; is the weight matrix of the value vector; To output the weight matrix, comprehensive user interest data is generated; Concat represents the concatenation operation, which merges multiple vectors or matrices by dimension; Attention represents the self-attention mechanism, which is used to calculate the similarity between the query vector and the key vector and generate a weighted value vector.

6. The method according to claim 5, characterized in that The optimization of the matching degree between the user interest representation and the recommended content by combining the contrastive learning module includes: Construct positive and negative sample pairs to generate positive and negative sample data; Calculate the distance between positive and negative sample data to generate sample distance data; The loss value is calculated according to the loss function of contrastive learning, which is defined as: in, For sample and The distance between is a binary label; is the boundary distance, used to generate loss data; Using the optimization results of the comparative learning module, adjust the comprehensive user interest data to optimize the match between it and the recommended content; Generate personalized recommendation content data based on the optimized comprehensive user interest data.

7. The method according to claim 1, characterized in that The collecting of multimodal information of users and fusing it with personalized recommendation content data includes: Collect users’ geographic location, device usage, and social network data as multimodal information; Preprocess the multimodal information to generate standardized multimodal feature data; Use a multimodal fusion algorithm to fuse standardized multimodal feature data and personalized recommendation content data to generate multimodal recommendation data; The multimodal recommendation data is optimized by using an optimization algorithm to ensure the relevance and personalization of the recommended content, and to generate optimized multimodal recommendation content data, wherein the optimization algorithm includes: a gradient descent method, a genetic algorithm, and a simulated annealing algorithm.

8. The method according to claim 1, characterized in that The real-time display of optimized multi-modal recommended content data to the user and the collection of user feedback feature data include: Real-time display of optimized multi-modal recommended content data; Collect user interaction data on recommended content, including click-through rate, dwell time, number of likes, and comment content; Preprocess user interaction data, extract key indicators and features, and generate user feedback feature data.

9. The method according to claim 8, characterized in that The strategy for dynamically adjusting and optimizing the generation of recommended content through an adaptive learning algorithm includes: Based on user feedback feature data, the adaptive learning algorithm is used to dynamically adjust and optimize the generation strategy of recommended content to generate optimized recommended content data; Using the optimized recommended content data, updating the parameters and strategies of the recommended content generation strategy, including the adjustment of model weights, learning rates, and loss functions, to generate updated multimodal recommended content data; Apply the updated multimodal recommendation content data to the real-time recommendation generation module, dynamically adjust the recommendation strategy, and generate optimized real-time recommendation data for displaying and providing recommendation services to users.

10. A system for implementing the content recommendation method based on semantic recognition according to any one of claims 1 to 9, characterized in that: include: A natural language processing module configured to parse user requirements inputted via text, voice or video and generate user semantic representation data; The user interest graph construction module dynamically constructs the user interest graph based on the user semantic representation data and the user historical behavior data, and learns the representation of nodes and edges through a multi-layer graph convolutional network to generate time series user interest graph representation data; The Transformer model module uses the time series user interest graph to represent data, and optimizes the matching degree between user interest representation and recommended content through self-attention mechanism and multi-head attention mechanism combined with the contrastive learning module to generate personalized recommended content data; The multimodal information collection and fusion module collects the multimodal information of users and fuses it with the personalized recommendation content data, generates multimodal recommendation data using the multimodal fusion algorithm, and generates optimized multimodal recommendation content data through the optimization algorithm; Real-time display and feedback collection module, which displays optimized multi-modal recommended content data to users in real time and collects user feedback feature data; The adaptive learning module dynamically adjusts and optimizes the generation strategy of recommended content based on user feedback feature data through an adaptive learning algorithm to ensure the continued accuracy and personalization of recommended content.

Images