Broadcast television media content intelligent arrangement and distribution system

Abstract

The application discloses a broadcast television fusion media content intelligent arrangement and distribution system, which comprises a content collection and labeling module, a deep portrait model based on user time sequence behavior, an intelligent arrangement module, an intelligent distribution module and a feedback optimization module.The content collection and labeling module acquires multi-source fusion media content and carries out multi-dimensional semantic labeling to generate a content label vector.The deep portrait model based on user time sequence behavior predicts a content preference change trend.The intelligent arrangement module comprehensively considers user portraits, platform distribution strategies and content copyright states, constructs a multi-target optimization arrangement model, generates a candidate scheme set, eliminates conflicts through a multi-dimensional conflict detection mechanism, selects a feasible arrangement scheme, and dynamically optimizes a distribution path of the feasible scheme according to network quality states and edge node loads of real-time distribution links, selects an optimal path to execute content distribution, and feeds back optimization of the model parameters through online learning of real-time feedback data of users.

Description

Technical Field

[0001] This invention relates to the field of television converged media technology, and in particular to an intelligent arrangement and distribution system for broadcast television converged media content. Background Technology

[0002] Against the backdrop of the nation's in-depth implementation of the "AI+" action and the accelerated transformation of the broadcasting and television industry towards a comprehensive intelligent audiovisual ecosystem, the intelligent content arrangement and distribution system for converged media in broadcasting and television aims to deeply integrate AI big data models with the existing intelligent platform capabilities of broadcasting and television. This is achieved through the construction of a human-machine collaborative intelligent content production pipeline, a multi-terminal dynamic arrangement mechanism based on refined user profiles, and a distribution strategy that balances mainstream value orientation with personalized recommendations. This system not only efficiently revitalizes broadcasting and television media assets and improves the quality and efficiency of the entire content production process, but also connects large and small screens, online and offline content flow links. While consolidating the advantages of mainstream public opinion dissemination, it provides users in different scenarios with high-quality audiovisual services tailored to each individual. Ultimately, it drives the core value leap of broadcasting and television media from traditional channel broadcasting to comprehensive intelligent services, helping mainstream media reshape their content dissemination competitiveness in the era of intelligent media and better fulfill their mission of cultural dissemination and public cultural services in the new era.

[0003] In existing technologies, in converged media scenarios, a conflict often triggers a chain reaction. For example, moving programs to resolve time overlap may cause new copyright term conflicts, and switching nodes to resolve resource overload may cause content format incompatibility. Therefore, a smart arrangement and distribution system for converged media content of broadcast television is proposed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent arrangement and distribution system for converged media content in broadcasting and television.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A smart content arrangement and distribution system for broadcast television converged media includes: Content Acquisition and Tagging Module: Acquires multi-source converged media content resources, performs multi-dimensional semantic tagging processing on the content resources, and generates content tag vectors; User profile building module: Constructs a deep user profile model based on user time-series behavior data, and predicts the trend of user content preference changes through the deep user profile model; Intelligent orchestration module: Constructs a multi-objective optimization orchestration model based on user profiles, platform distribution strategies, and content copyright status; generates a set of candidate orchestration schemes based on the multi-objective optimization orchestration model; resolves conflicts in the set of candidate orchestration schemes through a multi-dimensional conflict detection mechanism; and selects feasible orchestration schemes from the set of candidate orchestration schemes. Intelligent distribution module: Based on the network quality status of the real-time distribution link and the load of edge nodes, dynamically optimize the distribution path of the feasible orchestration scheme, and select the optimal distribution path to perform content distribution; Feedback optimization module: Based on real-time user feedback data, it performs online learning and performs closed-loop optimization of the parameters of the multi-objective optimization orchestration model.

[0006] The above technical solution further includes: Furthermore, the deep user profile model is built based on a Transformer temporal encoder, and the user profile construction module is specifically used for: Obtain a sequence of user content consumption behaviors within a preset time window, the sequence of content consumption behaviors including at least one of browsing behavior, clicking behavior, playback duration, interactive behavior and cross-platform jump behavior; The content consumption behavior sequence is input into the Transformer temporal encoder to extract the temporal features of user behavior. Based on the aforementioned temporal characteristics of user behavior, a multi-head attention mechanism is used to predict the changing trends of user content preferences in future time periods.

[0007] Furthermore, the content acquisition and tagging module includes: Modality recognition unit: performs modality recognition on the content resource to determine the content modality type of the content resource, wherein the content modality type includes at least one of video modality, audio modality, graphic modality, and interactive modality; Feature extraction unit: Selects the corresponding feature extractor according to the content modality type and extracts the modality feature vector of the content resource; Cross-modal fusion unit: Based on the cross-modal attention mechanism, the modal feature vectors are fused to generate the content tag vector.

[0008] Furthermore, the multi-objective optimization orchestration model includes multiple optimization objectives, including user satisfaction objectives, platform viewership objectives, advertising revenue objectives, copyright utilization objectives, content diversity objectives, and emergency response time objectives. The intelligent orchestration module is specifically used for: Obtain the current weight parameters for each optimization objective; Based on the current weight parameters, a multi-objective evolutionary algorithm is used to perform Pareto optimization on the candidate program sequences to generate the candidate scheduling scheme set. Each candidate scheduling scheme in the candidate scheduling scheme set corresponds to a set of Pareto non-dominated solutions.

[0009] Furthermore, the multi-dimensional conflict detection mechanism includes time-dimensional conflict detection, copyright-dimensional conflict detection, content adaptation-dimensional conflict detection, and resource-dimensional conflict detection; The intelligent orchestration module is specifically used for: For each candidate orchestration scheme, the time dimension conflict detection, the copyright dimension conflict detection, the content adaptation dimension conflict detection, and the resource dimension conflict detection are performed sequentially. When a conflict is detected, the candidate orchestration scheme is adjusted according to a preset conflict resolution strategy, which includes one of time offset resolution, content replacement resolution, resource reallocation resolution, and priority sorting resolution.

[0010] Furthermore, the intelligent distribution module is specifically used for: Real-time collection of current load status, available bandwidth, and transmission latency of each edge node; Obtain the geographic location information of the target user terminal and determine the set of candidate edge nodes that are closest to the target user terminal; Based on the current load status, available bandwidth, and transmission delay of each edge node in the candidate edge node set, the weighted shortest path algorithm is used to calculate the comprehensive transmission cost of each candidate distribution path. The distribution path with the lowest overall transmission cost is selected as the optimal distribution path.

[0011] Furthermore, it includes an emergency broadcast processing module, which is used to respond to an emergency broadcast trigger command, obtain the priority level and content attributes of the emergency broadcast content, determine the insertion position of the emergency broadcast content in the feasible scheduling scheme according to the priority level and generate a program schedule reconstruction strategy, and re-arrange the replaced content in the feasible scheduling scheme according to the program schedule reconstruction strategy to generate a reconstructed scheduling scheme.

[0012] Furthermore, it includes a blockchain copyright management module, which is used to store and manage copyright registration information, including the copyright holder, scope of authorization, authorization period, and authorization region; The content acquisition and tagging module sends a query request carrying the content identifier to the blockchain copyright management module, receives the returned copyright certificate information, and matches and verifies the copyright certificate information with the arrangement target and distribution scope. After the verification is passed, the content to be arranged is included in the multi-source converged media content resources.

[0013] Furthermore, the feedback optimization module is specifically used for: Collect real-time feedback data from users within a preset feedback time window after the feasible programming scheme is executed. The real-time feedback data includes click-through rate, completion rate, share rate, skip rate, and dwell time. The orchestration effect evaluation index is calculated based on the real-time feedback data; The choreography performance evaluation index is used as a reward signal, and a reinforcement learning algorithm is used to update the target weight parameters in the multi-objective optimization choreography model online.

[0014] A smart arrangement and distribution system for broadcast television converged media content also includes an adaptive format conversion module, which is used to adaptively convert the format and adjust the bit rate of the content to be distributed in the feasible arrangement scheme according to the device capability parameters of the target terminal device and the current network bandwidth, and generate a distribution content version adapted to the target terminal device. The device capability parameters include screen resolution, supported decoding formats, maximum frame rate, and color space. The adaptive format conversion includes real-time transcoding to convert the format when the target terminal device does not support the original encoding format of the content to be distributed, and downsampling when the screen resolution is lower than the resolution of the content to be distributed.

[0015] The present invention has the following beneficial effects: In this invention, a multi-dimensional conflict detection mechanism is employed to sequentially perform time-dimension conflict detection, copyright-dimension conflict detection, content adaptation-dimension conflict detection, and resource-dimension conflict detection on each candidate scheduling scheme. This is combined with conflict resolution strategies such as time offsetting, content replacement, and resource reallocation for automatic adjustments. This effectively avoids broadcast accidents and legal risks caused by time overlap, expired copyrights, format incompatibility, or insufficient resources at edge nodes, ensuring the reliable implementation of the scheduling scheme in actual operation. Attached Figure Description

[0016] Figure 1 This is a system block diagram of an intelligent arrangement and distribution system for converged media content in broadcasting and television proposed in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 As shown, this invention is an intelligent arrangement and distribution system for converged media content in broadcasting and television, comprising: Content Acquisition and Tagging Module: Acquires multi-source converged media content resources, performs multi-dimensional semantic tagging processing on the content resources, and generates content tag vectors; User profile building module: Constructs a deep user profile model based on user time-series behavior data, and predicts the trend of user content preference changes through the deep user profile model; Intelligent orchestration module: Constructs a multi-objective optimization orchestration model based on user profiles, platform distribution strategies, and content copyright status; generates a set of candidate orchestration schemes based on the multi-objective optimization orchestration model; resolves conflicts in the set of candidate orchestration schemes through a multi-dimensional conflict detection mechanism; and selects feasible orchestration schemes from the set of candidate orchestration schemes. Intelligent distribution module: Based on the network quality status of the real-time distribution link and the load of edge nodes, dynamically optimize the distribution path of the feasible orchestration scheme, and select the optimal distribution path to perform content distribution; Feedback optimization module: Based on real-time user feedback data, it performs online learning and performs closed-loop optimization of the parameters of the multi-objective optimization orchestration model.

[0019] The specific implementation process of a smart content arrangement and distribution system for broadcast television convergence is as follows: Acquire multi-source converged media content resources, perform multi-dimensional semantic tagging processing on the content resources, and generate content tag vectors: First, it is necessary to connect with television station media asset systems, OTT internet television platforms, news apps, and social media to aggregate multi-source converged media content resources. For the aggregated content, multi-dimensional semantic tagging processing is performed to unify its semantic representation, specifically including: First, the modality recognition unit performs modality recognition on the content resources to determine their content modality type, such as video modality, audio modality, and image-text modality. Recognition can be achieved by parsing file header information, MIME type, etc. Secondly, based on the determined modality type, the feature extraction unit calls the corresponding feature extractor to extract the modality feature vector. For example, for video modalities, a 3D convolutional neural network or video Transformer model is used to extract spatiotemporal features, for image-text modalities, a pre-trained language model (such as BERT) is used to extract text semantic features, and a visual Transformer is used to extract image features. Finally, the cross-modal fusion unit fuses the modal feature vectors of different modalities based on a cross-modal attention mechanism to generate the final content tag vector. This mechanism dynamically adjusts the contribution of each modal feature to the final vector through attention weights. For content containing both video and text modalities, the fusion process can be formally expressed as follows: First, calculate the attention weights between modalities. Let the video modality feature vector be V and the text modality feature vector be T. Map them to a query, key, and value space through a linear transformation, and calculate the attention weights. For example, using video features as the query and text features as the key and value, calculate the attention weight of text on video. ,in, , The weight matrix is ​​a learnable matrix. is the dimension of the key vector, used to scale the dot product; Then, using the calculated attention weights, the value vectors are weighted and summed to obtain the cross-modal fused feature vector. : in, The feature vector representing the text modality (t) or audio modality (a), It is the weight matrix of the value space. The representation layer normalization operation, this content tag vector This provides a high-quality, unified semantic foundation for subsequent intelligent orchestration; A deep user profile model is constructed based on user time-series behavior data. This model is then used to predict trends in user content preferences. Specifically, a deep user profile model based on a Transformer time-series encoder is built to capture dynamically changing user interests. The specific process is as follows: First, obtain the user's content consumption behavior sequence within a preset time window (e.g., the past 30 days). This sequence is a record of user behavior arranged in chronological order, including behaviors such as browsing, clicking, playback duration, interaction, and cross-platform navigation. Next, the content consumption behavior sequence is input into the Transformer temporal encoder. This encoder uses a self-attention mechanism to extract long-distance dependencies in the behavior sequence. The core multi-head attention mechanism's computation process is as follows: , among which, each attention head The calculation is as follows: , Here, Q, K, and V are all obtained by linear transformation of the input behavior sequence embedding vector. , , and It is a learnable parameter matrix, a mechanism that allows the model to understand user behavior patterns from different representation subspaces. For example, one head focuses on weekday viewing habits, while another head focuses on cross-platform interest migration. Finally, based on the temporal features of user behavior output by the encoder, a prediction head is used to predict the trend of user content preference changes in future time periods. ,in, and These are the weights and biases of the prediction layer. It is a sigmoid activation function that outputs the probability change of a user's preference for a specific content type; A multi-objective optimization orchestration model is constructed based on user profiles, platform distribution strategies, and content copyright status. A set of candidate orchestration schemes is then generated based on this multi-objective optimization orchestration model. The intelligent orchestration module constructs a multi-objective optimization orchestration model containing multiple optimization objectives, including at least user satisfaction, platform viewership, and advertising revenue objectives. The model is solved using a multi-objective evolutionary algorithm (preferably NSGA-II) to generate a set of Pareto-optimal candidate orchestration schemes. First, the programming scheme needs to be parameterized, that is, encoded as a chromosome x, whose gene sequence represents the broadcast order and time slot of the program; Then, define functions for each optimization objective, for example: User satisfaction objective function : Maximize the predicted weighted user click-through rate and completion rate; Platform viewership objective function Maximize the expected viewership of the programming during prime time; Advertising revenue objective function Maximize the product of the ad placement's expected exposure and click-through rate; Given a set of current weight parameters, the NSGA-II algorithm searches for the Pareto front through iterative evolution. Its core components are fast non-dominated sorting and crowding distance calculation. Used to measure the density of other solutions around a solution i on the Pareto front, in order to maintain solution diversity, it is calculated as follows: Where M is the total number of optimization objectives. and These are the function values ​​of solution i and its two adjacent solutions on the m-th objective, respectively. and These are the maximum and minimum values ​​of the objective at the same non-dominated level. By selecting solutions with large crowding distances, a set of candidate orchestration schemes with uniform distribution can be generated, where each scheme is a Pareto non-dominated solution. A multi-dimensional conflict detection mechanism is used to resolve conflicts in the candidate orchestration scheme set, and feasible orchestration schemes are selected from the candidate orchestration scheme set. The intelligent orchestration module executes the multi-dimensional conflict detection mechanism sequentially for each generated candidate orchestration scheme. This mechanism includes time-dimensional conflict detection, copyright-dimensional conflict detection, content adaptation-dimensional conflict detection, and resource-dimensional conflict detection. When a conflict is detected, adjustments are made according to a preset conflict resolution strategy. This resolution process can be defined as a constrained adjustment optimization problem. For example, for a conflict scheme S, the goal is to find an adjusted scheme S' that minimizes the difference from the original scheme. in, It is the space of feasible solutions that satisfy all dimensional constraints, including: Time constraints: or To ensure that programs on the same channel do not overlap; Copyright restrictions: and ; Resource constraints: ; Conflict resolution strategies include time offset resolution (adjusting Start(Si)), content replacement resolution (replacing Si), and resource reallocation resolution (changing the edge node of execution distribution), etc. Ultimately, the schemes that passed all tests and resolutions were marked as feasible orchestration schemes; Based on the network quality status of the real-time distribution link and the load of edge nodes, the feasible orchestration scheme is dynamically optimized for distribution paths, and the optimal distribution path is selected for content distribution. The intelligent distribution module is responsible for performing this step. Just before content distribution, it monitors the network status of the distribution link in real time and dynamically selects the optimal distribution path for each user request. This problem can be modeled as a weighted shortest path problem, with the goal of finding a path that minimizes overall transmission cost. Specifically, the overall transmission cost C is calculated using the following formula: ,in, This represents transmission latency, specifically the network round-trip time from the source node to the candidate edge node. and These represent the available bandwidth and total bandwidth of the candidate edge nodes, respectively. The node load rate, representing the candidate edge node, is typically a weighted average of metrics such as CPU utilization, memory utilization, and concurrent connections. , , It is a configurable weighting coefficient used to adjust the optimization focus according to the actual scenario (such as latency-sensitive business); The system collects L and L values ​​from each edge node in real time. In addition to metrics such as U, and combined with the geographical location of the user terminal, the C value is calculated for a group of candidate edge nodes, and finally the distribution path with the smallest C value is selected as the optimal distribution path for content delivery. Online learning is performed based on real-time user feedback data to optimize the parameters of the multi-objective optimization orchestration model in a closed loop: The feedback optimization module is responsible for collecting real-time user feedback data after the orchestration scheme is executed, and using reinforcement learning algorithms to update the weight parameters of the multi-objective optimization orchestration model online to form a closed loop optimization. Specifically, the weight parameters w of the multi-objective model are used as policy parameters θ in reinforcement learning. Real-time feedback data is collected within a preset feedback time window, and a comprehensive orchestration effect evaluation index R is calculated as the reward signal. in, For click-through rate, To achieve a high completion rate, For skip rate, To standardize the duration of stay, These are the hyperparameters that constitute the overall score. The Proximal Policy Optimization (PPO) algorithm is preferred for updating the policy parameter θ, and its core pruning objective function is: in, It is the probability ratio of the new strategy to the old strategy. It is an estimate of the dominance function. It is a hyperparameter used to limit the update magnitude. By maximizing this objective function and updating the parameter θ using the gradient ascent method, the orchestration strategy (i.e., the weight parameter w) gradually approaches the optimum in the continuous "orchestration-feedback" loop, thus achieving rapid adaptation to changes in user interests.

[0020] In one embodiment, the deep user profile model is built based on a Transformer temporal encoder, and the user profile building module is specifically used for: Obtain a sequence of user content consumption behaviors within a preset time window, the sequence of content consumption behaviors including at least one of browsing behavior, clicking behavior, playback duration, interactive behavior and cross-platform jump behavior; The content consumption behavior sequence is input into the Transformer temporal encoder to extract the temporal features of user behavior. Based on the aforementioned temporal characteristics of user behavior, a multi-head attention mechanism is used to predict the changing trends of user content preferences in future time periods.

[0021] In one embodiment, the content acquisition and tagging module includes: Modality recognition unit: performs modality recognition on the content resource to determine the content modality type of the content resource, wherein the content modality type includes at least one of video modality, audio modality, graphic modality, and interactive modality; Feature extraction unit: Selects the corresponding feature extractor according to the content modality type and extracts the modality feature vector of the content resource; Cross-modal fusion unit: Based on the cross-modal attention mechanism, the modal feature vectors are fused to generate the content tag vector.

[0022] In one embodiment, the multi-objective optimization orchestration model includes multiple optimization objectives, including user satisfaction objective, platform viewership objective, advertising revenue objective, copyright utilization objective, content diversity objective, and emergency response time objective. The intelligent orchestration module is specifically used for: Obtain the current weight parameters for each optimization objective; Based on the current weight parameters, a multi-objective evolutionary algorithm is used to perform Pareto optimization on the candidate program sequences to generate the candidate scheduling scheme set. Each candidate scheduling scheme in the candidate scheduling scheme set corresponds to a set of Pareto non-dominated solutions.

[0023] In one embodiment, the multi-dimensional conflict detection mechanism includes time-dimensional conflict detection, copyright-dimensional conflict detection, content adaptation-dimensional conflict detection, and resource-dimensional conflict detection. The intelligent orchestration module is specifically used for: For each candidate orchestration scheme, the time dimension conflict detection, the copyright dimension conflict detection, the content adaptation dimension conflict detection, and the resource dimension conflict detection are performed sequentially. When a conflict is detected, the candidate orchestration scheme is adjusted according to a preset conflict resolution strategy, which includes one of time offset resolution, content replacement resolution, resource reallocation resolution, and priority sorting resolution.

[0024] In one embodiment, the intelligent distribution module is specifically used for: Real-time collection of current load status, available bandwidth, and transmission latency of each edge node; Obtain the geographic location information of the target user terminal and determine the set of candidate edge nodes that are closest to the target user terminal; Based on the current load status, available bandwidth, and transmission delay of each edge node in the candidate edge node set, the weighted shortest path algorithm is used to calculate the comprehensive transmission cost of each candidate distribution path. The distribution path with the lowest overall transmission cost is selected as the optimal distribution path.

[0025] In one embodiment, an emergency broadcast processing module is included. The emergency broadcast processing module is used to respond to an emergency broadcast trigger command, obtain the priority level and content attributes of the emergency broadcast content, determine the insertion position of the emergency broadcast content in the feasible scheduling scheme according to the priority level and generate a program schedule reconstruction strategy, and re-arrange the replaced content in the feasible scheduling scheme according to the program schedule reconstruction strategy to generate a reconstructed scheduling scheme.

[0026] In one embodiment, a blockchain copyright management module is included, which is used to store and manage copyright registration information, including copyright holder, scope of authorization, authorization period and authorization region. The content acquisition and tagging module sends a query request carrying the content identifier to the blockchain copyright management module, receives the returned copyright certificate information, and matches and verifies the copyright certificate information with the arrangement target and distribution scope. After the verification is passed, the content to be arranged is included in the multi-source converged media content resources.

[0027] In one embodiment, the feedback optimization module is specifically used for: Collect real-time feedback data from users within a preset feedback time window after the feasible programming scheme is executed. The real-time feedback data includes click-through rate, completion rate, share rate, skip rate, and dwell time. The orchestration effect evaluation index is calculated based on the real-time feedback data; The choreography performance evaluation index is used as a reward signal, and a reinforcement learning algorithm is used to update the target weight parameters in the multi-objective optimization choreography model online.

[0028] A smart arrangement and distribution system for broadcast television converged media content also includes an adaptive format conversion module, which is used to adaptively convert the format and adjust the bit rate of the content to be distributed in the feasible arrangement scheme according to the device capability parameters of the target terminal device and the current network bandwidth, and generate a distribution content version adapted to the target terminal device. The device capability parameters include screen resolution, supported decoding formats, maximum frame rate, and color space. The adaptive format conversion includes real-time transcoding to convert the format when the target terminal device does not support the original encoding format of the content to be distributed, and downsampling when the screen resolution is lower than the resolution of the content to be distributed.

[0029] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart content arrangement and distribution system for broadcast television converged media, characterized in that, include: Content Acquisition and Tagging Module: Acquires multi-source converged media content resources, performs multi-dimensional semantic tagging processing on the content resources, and generates content tag vectors; User profile building module: Constructs a deep user profile model based on user time-series behavior data, and predicts the trend of user content preference changes through the deep user profile model; Intelligent orchestration module: Constructs a multi-objective optimization orchestration model based on user profiles, platform distribution strategies, and content copyright status; generates a set of candidate orchestration schemes based on the multi-objective optimization orchestration model; resolves conflicts in the set of candidate orchestration schemes through a multi-dimensional conflict detection mechanism; and selects feasible orchestration schemes from the set of candidate orchestration schemes. Intelligent distribution module: Based on the network quality status of the real-time distribution link and the load of edge nodes, dynamically optimize the distribution path of the feasible orchestration scheme, and select the optimal distribution path to perform content distribution; Feedback optimization module: Based on real-time user feedback data, it performs online learning and performs closed-loop optimization of the parameters of the multi-objective optimization orchestration model.

2. The intelligent arrangement and distribution system for converged media content in broadcasting and television according to claim 1, characterized in that, The deep user profile model is built based on the Transformer temporal encoder, and the user profile construction module is specifically used for: Obtain the sequence of user content consumption behavior within a preset time window; The content consumption behavior sequence is input into the Transformer temporal encoder to extract the temporal features of user behavior. Based on the aforementioned temporal characteristics of user behavior, a multi-head attention mechanism is used to predict the changing trends of user content preferences in future time periods.

3. The intelligent arrangement and distribution system for converged media content in broadcasting and television according to claim 1, characterized in that, The content acquisition and tagging module includes: Modality recognition unit: performs modality recognition on the content resource to determine the content modality type of the content resource; Feature extraction unit: Selects the corresponding feature extractor according to the content modality type and extracts the modality feature vector of the content resource; Cross-modal fusion unit: Based on the cross-modal attention mechanism, the modal feature vectors are fused to generate the content tag vector.

4. The intelligent arrangement and distribution system for converged media content in broadcasting and television according to claim 1, characterized in that, The multi-objective optimization orchestration model includes multiple optimization objectives, including user satisfaction objectives, platform viewership objectives, advertising revenue objectives, copyright utilization objectives, content diversity objectives, and emergency response time objectives. The intelligent orchestration module is specifically used for: Obtain the current weight parameters for each optimization objective; Based on the current weight parameters, a multi-objective evolutionary algorithm is used to perform Pareto optimization on the candidate program sequences to generate the candidate scheduling scheme set. Each candidate scheduling scheme in the candidate scheduling scheme set corresponds to a set of Pareto non-dominated solutions.

5. The intelligent arrangement and distribution system for converged media content in broadcasting and television according to claim 1, characterized in that, The multi-dimensional conflict detection mechanism includes time-dimensional conflict detection, copyright-dimensional conflict detection, content adaptation-dimensional conflict detection, and resource-dimensional conflict detection. The intelligent orchestration module is specifically used for: For each candidate orchestration scheme, the time dimension conflict detection, the copyright dimension conflict detection, the content adaptation dimension conflict detection, and the resource dimension conflict detection are performed sequentially. When a conflict is detected, the candidate orchestration scheme is adjusted according to a preset conflict resolution strategy, which includes one of time offset resolution, content replacement resolution, resource reallocation resolution, and priority sorting resolution.

6. The intelligent arrangement and distribution system for converged media content in broadcasting and television according to claim 1, characterized in that, The intelligent distribution module is specifically used for: Real-time collection of current load status, available bandwidth, and transmission latency of each edge node; Obtain the geographic location information of the target user terminal and determine the set of candidate edge nodes that are closest to the target user terminal; Based on the current load status, available bandwidth, and transmission delay of each edge node in the candidate edge node set, the weighted shortest path algorithm is used to calculate the comprehensive transmission cost of each candidate distribution path. The distribution path with the lowest overall transmission cost is selected as the optimal distribution path.

7. The intelligent arrangement and distribution system for converged media content in broadcasting and television according to claim 1, characterized in that, The system includes an emergency broadcast processing module, which is used to respond to an emergency broadcast trigger command, obtain the priority level and content attributes of the emergency broadcast content, determine the insertion position of the emergency broadcast content in the feasible scheduling scheme according to the priority level and generate a program schedule reconstruction strategy, and re-arrange the replaced content in the feasible scheduling scheme according to the program schedule reconstruction strategy to generate a reconstructed scheduling scheme.

8. The intelligent arrangement and distribution system for converged media content in broadcasting and television according to claim 1, characterized in that, It includes a blockchain copyright management module, which is used to store and manage copyright registration information, including the copyright holder, scope of authorization, authorization period and authorization region; The content acquisition and tagging module sends a query request carrying the content identifier to the blockchain copyright management module, receives the returned copyright certificate information, and matches and verifies the copyright certificate information with the arrangement target and distribution scope. After the verification is passed, the content to be arranged is included in the multi-source converged media content resources.

9. A smart arrangement and distribution system for converged media content in broadcasting and television according to claim 1, characterized in that, The feedback optimization module is specifically used for: Collect real-time feedback data from users within a preset feedback time window after the feasible programming scheme is executed. The real-time feedback data includes click-through rate, completion rate, share rate, skip rate, and dwell time. The orchestration effect evaluation index is calculated based on the real-time feedback data; The choreography performance evaluation index is used as a reward signal, and a reinforcement learning algorithm is used to update the target weight parameters in the multi-objective optimization choreography model online.

Images