County region non-abandoned AIGC propagation content pushing method and system

Abstract

The invention discloses a method and a system for pushing county non-abandoned AIGC propagation content. The method comprises the following steps of: performing structured extraction on county non-abandoned cultural elements, and constructing a non-abandoned cultural element library; carrying out multi-source user data acquisition, and constructing a user interest image; based on the user interest image and the propagation scene, retrieving matched culture elements from a non-perpetual culture element library, and inputting a non-perpetual exclusive AIGC generation model to generate AIGC personalized propagation content; determining a push channel, and dynamically pushing the AIGC personalized propagation content by adopting a preset hybrid push algorithm; and collecting pushing effect data, and optimizing parameters of the non-exclusive AIGC generation model according to the pushing effect data. The method has the remarkable effects that personalized generation, intelligent matching and dynamic pushing of non-abandoned cultural transmission contents are realized, the transmission breadth and user acceptance of non-abandoned cultural are improved, the problems of narrow coverage and poor adaptation of traditional transmission are solved, and meanwhile, accurate transmission of non-abandoned cultural connotation is ensured.

Description

Methods and Systems for Pushing AIGC Content for the Dissemination of Intangible Cultural Heritage in Counties Technical Field

[0001] This invention relates to the field of artificial intelligence technology for the dissemination of intangible cultural heritage, specifically to a method and system for pushing AIGC (AI-generated content) dissemination of intangible cultural heritage in a county. Background Technology

[0002] County-level intangible cultural heritage carries unique regional cultural genes, but its dissemination currently faces many bottlenecks: traditional dissemination methods such as offline exhibitions and print publicity have limited coverage and low efficiency; existing digital dissemination mostly adopts a "unified content, broad-spectrum push" model, which lacks precise adaptation to user interests, resulting in low user attention and poor dissemination effects; the creation of intangible cultural heritage content relies on professional teams, which is costly and time-consuming, making it difficult to meet the high-frequency dissemination needs; at the same time, county-level intangible cultural heritage has strong regional cultural attributes and many subcategories, so existing technologies cannot achieve deep synergy between "cultural characteristics - user needs - AIGC generation", resulting in a disconnect between the disseminated content and the core cultural connotation of intangible cultural heritage, and failing to effectively convey the cultural value of intangible cultural heritage.

[0003] Furthermore, existing AIGC (Artificial Intelligence Generated Content) content generation and push technologies are mostly applied to general fields such as e-commerce product recommendations and news pushes, and have not been adapted to the special characteristics of county-level intangible cultural heritage. Therefore, they lack the ability to extract structured intangible cultural heritage elements, and the generated content is prone to cultural distortion. In addition, user profiles are only based on superficial behaviors such as browsing and clicking, without combining deeper characteristics such as intangible cultural heritage preferences and regional relevance. The push strategy is static and fixed, and cannot be dynamically optimized according to the dissemination effect, making it difficult to adapt to the diverse scenarios of county-level intangible cultural heritage dissemination, such as festival promotion, cultural tourism attraction, and campus science popularization.

[0004] Therefore, there is an urgent need for a targeted method or system for pushing AIGC (AI-generated content) dissemination of intangible cultural heritage in counties to address the aforementioned core pain points. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for pushing AIGC (AI-Generated Content) dissemination of intangible cultural heritage at the county level. This method and system are applicable to the digital promotion of intangible cultural heritage at the county level. It enables personalized generation, intelligent matching, and dynamic pushing of intangible cultural heritage dissemination content, thereby enhancing the breadth of dissemination and user acceptance of intangible cultural heritage. It solves the problems of narrow coverage and poor adaptability in traditional dissemination methods, while ensuring the accurate transmission of the connotation of intangible cultural heritage.

[0006] To achieve the above objectives, the technical solution adopted by this invention is as follows: Firstly, this invention proposes a method for pushing AIGC (Artificial Intelligence Generic Content) dissemination materials related to intangible cultural heritage (ICH) at the county level. The key aspects include the following steps: Step 1: Structured extraction of ICH cultural elements at the county level to construct an ICH cultural element database; Step 2: Collection of multi-source user data and construction of user interest images; Step 3: Based on user interest images and dissemination scenarios, retrieval of matching cultural elements from the ICH cultural element database and input into an ICH-specific AIGC generation model to generate personalized AIGC dissemination content; Step 4: Determination of push channels and dynamic push of personalized AIGC dissemination content using a preset hybrid push algorithm; Step 5: Collection of push effect data and optimization of the parameters of the ICH-specific AIGC generation model based on the push effect data.

[0007] Furthermore, the steps in step 1 for constructing the intangible cultural heritage element database include: Step 1.1: Collecting multi-dimensional data on county-level intangible cultural heritage and establishing a county-level intangible cultural heritage database; Step 1.2: Preprocessing the data in the county-level intangible cultural heritage database; Step 1.3: Constructing a three-level structured system of intangible cultural heritage core attributes, cultural elements, and dissemination tags, and processing the preprocessed data, wherein the intangible cultural heritage core attributes include intangible cultural heritage categories, regional characteristics, and cultural connotations; Step 1.4: Assigning a unique identifier to each cultural element in the three-level structured system and associating it with the corresponding county-level intangible cultural heritage data to form a structured intangible cultural heritage element database.

[0008] Furthermore, the multi-dimensional data collected in step 1.1 includes: text data, image data, and audio-visual data. The text data includes the historical origins of intangible cultural heritage, the process of craftsmanship, and the folk customs. The image data includes images of intangible cultural heritage works, images of the process of craftsmanship, and images of inheritors. The audio-visual data includes intangible cultural heritage performances, skill demonstrations, and records of folk activities.

[0009] Furthermore, step 2 involves collecting multi-source user data and constructing user interest images, including: Step 2.1: Collecting static user data, dynamic behavior data, and intangible cultural heritage interaction data to generate a user dataset; Step 2.2: Combining the intangible cultural heritage element library, inputting the standardized user dataset into a pre-trained user interest feature extraction model for feature extraction to obtain user interest profile vectors; Step 2.3: Constructing user interest images based on user interest profile vectors.

[0010] Furthermore, the user interest feature extraction model includes: a basic interest layer: extracting user preferences for intangible cultural heritage categories and content formats based on dynamic behavioral data; a deep interest layer: combining an intangible cultural heritage element library to analyze user preferences for specific cultural elements and calculate user interest weights; a regional association layer: associating the user's current location with the regional matching degree of county-level intangible cultural heritage to determine the push weight of regionally related intangible cultural heritage content; and an output layer: outputting a user interest profile vector including multi-dimensional features such as basic attributes, interest preferences, lower relevance, and cultural acceptance.

[0011] Furthermore, step 3, which generates AIGC personalized dissemination content, includes: Step 3.1, based on user interest images and dissemination scenarios, retrieving matching cultural elements from the intangible cultural heritage element library to determine the theme, form, and cultural core of the dissemination content; Step 3.2, constructing an AIGC generation model exclusive to intangible cultural heritage, inputting the matched cultural elements, user interest images, and scenario requirements to generate initial personalized dissemination content; Step 3.3, using cultural verification rules to filter the personalized dissemination content, identifying culturally distorted content, and obtaining the AIGC personalized dissemination content.

[0012] Furthermore, the process of dynamically pushing AIGC personalized content using a preset hybrid push algorithm in step 4 includes: Step 4.1: Filtering AIGC personalized content with similar user preferences based on user interest similarity; Step 4.2: Calculating the similarity between the filtered AIGC personalized content and user interest images, and including content with similarity greater than a threshold in the candidate recommendation list; Step 4.3: Adjusting the weight of the pushed content according to the propagation scenario, and pushing the AIGC personalized content to the user; Step 4.4: Collecting user interaction feedback data in real time, and adjusting the push frequency based on user interaction feedback.

[0013] Furthermore, the push effect data mentioned in step 5 includes: core effect indicators, cultural transmission indicators, and conversion indicators, wherein: the core effect indicators include content open rate, dwell time, collection rate, sharing rate, forwarding rate, and interaction participation rate; the cultural transmission indicators include the correctness rate of knowledge quizzes and the mention rate of intangible cultural heritage elements in user comments; and the conversion indicators include the number of cultural tourism reservations, the purchase volume of intangible cultural heritage products, and the number of participants in offline activities.

[0014] Furthermore, the push effect data mentioned in step 5 is also used to optimize the push strategy in step 4.

[0015] Secondly, this invention proposes a county-level intangible cultural heritage (ICH) AIGC content push system for implementing the method described in the first aspect. The key features include: an element library construction module for structurally extracting county-level ICH cultural elements and constructing an ICH cultural element library; a user interest image construction module for collecting multi-source user data and constructing user interest images; a personalized content generation module for retrieving matching cultural elements from the ICH cultural element library based on user interest images and the dissemination scenario, and inputting them into an ICH-specific AIGC generation model to generate personalized AIGC content; a content push module for determining push channels and dynamically pushing personalized AIGC content using a preset hybrid push algorithm; and a dissemination effect feedback module for collecting push effect data and optimizing the parameters of the ICH-specific AIGC generation model using a reinforcement learning algorithm.

[0016] The significant effects of this invention are: High accuracy in dissemination: User interest matching and content open rate are effectively improved compared to traditional push notifications, the accuracy of intangible cultural heritage knowledge transmission is greatly improved, and user attention is effectively enhanced; Accurate cultural transmission: Through the structured extraction and content verification of intangible cultural heritage elements, the cultural distortion rate is effectively reduced, ensuring the accurate transmission of the core cultural connotations of county-level intangible cultural heritage; Reduced creation cost: AIGC generation replaces traditional professional creation, shortening the content generation cycle from several days to minutes, effectively reducing creation costs; Strong scene adaptability: Supports push notifications in multiple scenarios such as cultural tourism promotion, campus science popularization, and festival promotion, adapting to the dissemination needs of different county-level intangible cultural heritages, and allowing for flexible adjustment of push notification strategies; Expanded dissemination scope: Digital push notifications cover multiple online channels, breaking through geographical limitations, and effectively expanding the user coverage of county-level intangible cultural heritage dissemination compared to traditional methods; Enhanced interactivity and participation: The user participation rate of personalized interactive content can be effectively improved compared to traditional methods, and can effectively drive the increase in offline intangible cultural heritage activity participation and cultural tourism reservations, realizing a closed loop of "online dissemination - offline conversion". Attached Figure Description

[0017] Figure 1 is a flowchart of the method described in this invention.

[0018] Figure 2 is a schematic diagram of the system described in this invention. Detailed Implementation

[0019] The specific embodiments and working principles of the present invention will be further described in detail below with reference to the accompanying drawings. Examples

[0020] As shown in Figure 1, this embodiment of the invention proposes a method for pushing AIGC dissemination content of intangible cultural heritage in counties. The specific steps are as follows: Step 1: Extract structured intangible cultural heritage elements of counties and construct an intangible cultural heritage element database. In the specific implementation process, the specific steps for constructing the intangible cultural heritage element database in this step include: Step 1.1: Collect multi-dimensional data of intangible cultural heritage in counties and establish a county-level intangible cultural heritage database. In some optional embodiments, the collected multi-dimensional data includes: text data, image data, and audio-visual data, wherein: text data includes the historical origin of intangible cultural heritage, the process of craftsmanship, and the folk customs; image data includes images of intangible cultural heritage works, images of the process of craftsmanship, and images of inheritors; audio-visual data includes intangible cultural heritage performances, skill demonstrations, and records of folk activities.

[0021] Step 1.2: Data preprocessing is performed on the data in the county-level intangible cultural heritage database. In some optional implementations, the data preprocessing based on the collected multi-dimensional data is performed as follows: Text data is segmented and stop word removal is performed to extract core cultural keywords, such as "paper cutting - window decoration - auspicious meaning - red paper cutting"; Image data is processed using the ResNet-50 model for feature extraction, labeling intangible cultural heritage visual elements such as color, pattern, and material; ResNet-50 is a deep residual network, belonging to the ResNet series of models, widely used in image recognition and computer vision tasks. It solves the gradient vanishing problem in deep networks by introducing residual connections, thus enabling the training of deeper network structures. The core feature of this model is the use of Bottleneck residual blocks instead of the BasicBlock in ResNet-18. The Bottleneck block contains three convolutional layers: first, dimensionality reduction is achieved through 1×1 convolution, then feature extraction is achieved through 3×3 convolution, and finally, dimensionality increase is achieved through 1×1 convolution. This design reduces computation while maintaining high accuracy. ResNet-50 consists of 50 convolutional layers, with its structure divided into an initial convolutional layer, four residual block groups (each containing multiple Bottleneck blocks), a global average pooling layer, and a fully connected layer. It extracts keyframes / segments from audio and video data to identify intangible cultural heritage auditory elements such as singing styles, musical instruments, skillful manipulations, and sound effects.

[0022] Step 1.3: Construct a three-level structured system of intangible cultural heritage core attributes, cultural elements, and dissemination tags, and process the preprocessed data. The core attributes of intangible cultural heritage include intangible cultural heritage categories, regional characteristics, and cultural connotations. Step 1.4: Assign a unique identifier to each cultural element in the three-level structured system and associate it with the corresponding county-level intangible cultural heritage data to form a structured intangible cultural heritage cultural element database. This database supports multi-dimensional retrieval based on cultural attributes, regional characteristics, dissemination scenarios, etc.

[0023] This step, through the methods described above, constructs a structured element library tailored to the regional characteristics and cultural connotations of county-level intangible cultural heritage, enabling the structured extraction of intangible cultural heritage elements. This effectively reduces cultural distortion rates, ensures the cultural accuracy of AIGC-generated content, and thus helps guarantee the accurate transmission of the core cultural connotations of county-level intangible cultural heritage.

[0024] Step 2: Collect multi-source user data and construct a user interest image; In some implementations, the process of collecting multi-source user data and constructing a user interest image described in this step is as follows: Step 2.1: Collect user static data, dynamic behavior data, and intangible cultural heritage interaction data to generate a user dataset; In specific implementation, the above data specifically includes: User static data includes: age, gender, occupation, profession, and cultural background; Dynamic behavior data includes: browsing time of intangible cultural heritage content, click type, collection behavior, sharing behavior, and comment feedback behavior; Intangible cultural heritage interaction data includes: participation in intangible cultural heritage on-site experience activities The data includes: activity logs, intangible cultural heritage knowledge quizzes, and handicraft tutorial learning records; Step 2.2: Combining the intangible cultural heritage element database, the standardized user dataset is input into a pre-trained user interest feature extraction model for feature extraction to obtain user interest profile vectors; In this example, the user interest feature extraction model includes: Basic interest layer: extracting user preferences for intangible cultural heritage categories and content formats such as text, images, videos, and interactive content based on dynamic behavior data; Deep interest layer: combining the intangible cultural heritage element database to analyze user preferences for specific cultural elements and using the TF-IDF algorithm to calculate user interest weights; The TF-IDF (term frequency–inverse document frequency) algorithm is a commonly used weighting technique for information retrieval and data mining, used to evaluate the importance of a word to a document in a document set or corpus. TF is Term Frequency, and IDF is Inverse Document Frequency. This model posits that the importance of a word increases proportionally to the frequency of its occurrence in a document, and decreases inversely proportionally to the frequency of its occurrence in the corpus. Various weighted forms of this model are frequently used by search engines as a measure of the relevance between a document and a user query, and together with other link analysis-based ranking methods, it determines the order in which documents appear in search results.

[0025] Geographic Relationship Layer: Relates the user's current location to the geographic matching degree of intangible cultural heritage in the county, such as the user's place of origin, the county the user has visited, and the user's interest in regional culture, to determine the push weight of regionally related intangible cultural heritage content; Output Layer: Outputs a user interest profile vector that includes multi-dimensional features such as basic attributes, interest preferences, relevance, and cultural acceptance.

[0026] Step 2.3: Construct user interest images based on user interest profile vectors.

[0027] Based on the above process, it is possible to effectively integrate multi-dimensional features such as user basic behavior, cultural preferences, and regional connections, break through the superficial interest matching, and thus improve the adaptability of intangible cultural heritage content in subsequent processes.

[0028] Step 3: Based on user-interested images and dissemination scenarios, retrieve matching cultural elements from the intangible cultural heritage element database and input them into the intangible cultural heritage-specific AIGC generation model to generate personalized AIGC dissemination content. In some specific implementations, this step can generate personalized AIGC dissemination content in the following way: Step 3.1: Based on user-interested images and dissemination scenarios, retrieve matching cultural elements from the intangible cultural heritage element database to determine the theme, form, and cultural core of the dissemination content. In specific implementations, the dissemination scenario can be automatically identified or selected by the user, such as cultural tourism promotion, campus science popularization, festival promotion, etc. The theme, form, and cultural core of the dissemination content generated based on the dissemination scenario can be: for cultural tourism scenarios, generate a "check-in video of intangible cultural heritage + county scenic spot"; for campus scenarios, generate graphic content such as "simplified tutorial of intangible cultural heritage skills", etc.

[0029] Step 3.2: Construct an AIGC generation model specifically for intangible cultural heritage (ICH). Input the matched cultural elements, user interest images, and scene requirements to generate initial personalized dissemination content. In the specific implementation process, since the dissemination content may take different forms such as text, images, audio, video, or interactive content, the AIGC generation model specifically for ICH in this embodiment adopts different models depending on the generated dissemination content, as follows: Text and image content: Use the improved Stable Diffusion model, incorporating the visual elements of ICH, to generate ICH-themed posters and text and image interpretations that conform to the user's aesthetic preferences (such as generating a "blue calico pattern + modern design" poster for users who prefer traditional patterns). Stable Diffusion is an AI painting generation tool. Users can input whatever content they want, and the system will automatically generate an artistic rendering. The generation system is trained on a huge database of existing artworks and can quickly generate novel images related to the prompt information. Its working principle is as follows: Initialization: Start from a simple distribution (such as a Gaussian distribution) to generate a random noise image.

[0030] Diffusion process: Structure is gradually introduced into the noisy image through multiple steps, each step making the image clearer and eventually approaching the target data distribution.

[0031] Reverse diffusion process: During the training phase, the model learns how to reverse the diffusion process, that is, to recover a meaningful image from noise.

[0032] Conditional generation: When generating images, conditions (such as text descriptions) can be provided, and the model will generate images that match the conditions.

[0033] Video Content: The GPT-4V+VideoLDM model is used to generate short videos demonstrating intangible cultural heritage skills and recreating folk activities, ensuring the accuracy of intangible cultural heritage operation procedures and cultural scenes. GPT-4V (GPT-4 with Vision) is a multimodal language model with visual capabilities developed by OpenAI and officially released in September 2023. This model integrates text and image processing capabilities, enabling it to recognize objects, parse charts, provide step-by-step operation guidance, and is applied in scenarios such as medical image-assisted analysis and assistive tools for the visually impaired (such as the "Be MyAI" function of Be My Eyes). The method and steps by which this model generates videos from input images are existing technologies and will not be elaborated upon here.

[0034] The VideoLDM model is a text-to-video model developed by Nvidia that can automatically generate videos based on user-generated text descriptions. The highest video resolution is 2048*1280, at 24 frames per second, with a maximum duration of 4.7 seconds. The method and steps by which this model generates videos from input text are existing technology and will not be elaborated upon here.

[0035] Interactive content: Generate Q&A on intangible cultural heritage knowledge, H5 interactive experiences for hands-on activities, and dialogue scripts with virtual intangible cultural heritage inheritors to enhance user engagement.

[0036] Step 3.3: Use cultural verification rules to screen personalized communication content, identify culturally distorted content, and obtain the AIGC personalized communication content.

[0037] This step, based on user profiles and scenario needs, customizes the content format and cultural focus, solving the problems of high cost and long cycle in the creation of intangible cultural heritage content.

[0038] Step 4: Determine the push channels and dynamically push AIGC personalized content using a preset hybrid push algorithm. In some optional implementations, the methods for determining push channels and dynamically pushing AIGC personalized content using a preset hybrid push algorithm are as follows: Push channels include short video platforms, social media, cultural tourism apps, campus education platforms, and county-level government official accounts. The methods for determining push channels can be as follows: young people prefer short video platforms, middle-aged and elderly people prefer WeChat official accounts, and student groups prefer educational apps, etc.

[0039] Step 4.1: Based on user interest similarity, filter AIGC personalized content with similar user preferences; Step 4.2: Calculate the cosine similarity between the filtered AIGC personalized content and the user's interest image, and include content with a cosine similarity greater than the threshold of 0.7 in the candidate recommendation list; Step 4.3: Adjust the weight of the pushed content according to the dissemination scenario, and push the AIGC personalized content to the user; In some implementations, the way to adjust the weight of the pushed content according to the dissemination scenario can be: in festival scenarios, the weight of folk intangible cultural heritage content is increased by 0.3; in cultural tourism scenarios, the weight of skill experience content is increased by 0.25.

[0040] Step 4.4: Collect user interaction feedback data in real time, such as open rate, dwell time, likes, comments or shares, and adjust the push frequency based on user interaction feedback. Push 2-5 messages per day to users with high interest and 1-5 messages per week to users with low interest.

[0041] This step adjusts the push strategy based on the dissemination scenario and channel preferences, which can adapt to the diverse needs of county-level intangible cultural heritage dissemination and support push in multiple scenarios such as cultural tourism promotion, campus science popularization, and festival promotion. It can flexibly adjust the push strategy to meet the dissemination needs of different county-level intangible cultural heritage.

[0042] Step 5: Collect push effect data and optimize the parameters of the intangible cultural heritage-specific AIGC generation model based on the push effect data.

[0043] In practical implementation, the push notification performance data includes: core performance indicators, cultural transmission indicators, and conversion indicators. The core performance indicators include content open rate, dwell time, collection rate, sharing rate, forwarding rate, and interactive participation rate. The cultural transmission indicators include the accuracy rate of knowledge quizzes and the mention rate of intangible cultural heritage elements in user comments. The conversion indicators include the number of cultural tourism reservations, the purchase volume of intangible cultural heritage products, and the number of participants in offline activities.

[0044] In practical implementation, reinforcement learning algorithms are used for optimization. The optimization includes: content generation optimization: if the open rate of content of a certain type of intangible cultural heritage element is low, the generation parameters of the AIGC generation model dedicated to intangible cultural heritage are adjusted; if the cultural transmission indicators do not meet the standards, the structured extraction accuracy of intangible cultural heritage elements is enhanced; push strategy optimization: the allocation ratio of push channels is adjusted according to the channel effect, the user profile is updated according to changes in user interests, and the scene weight and push frequency are dynamically adjusted; push cycle optimization: regular optimization is carried out every 24 hours, and real-time optimization is carried out for major dissemination scenarios (such as intangible cultural heritage festivals).

[0045] This step optimizes the generation and push strategies in real time based on the dissemination effect, thereby continuously improving the dissemination effect and the accuracy of cultural transmission.

[0046] In practical implementation, the method described in this embodiment may further include the following steps: regularly collecting new county-level intangible cultural heritage data, such as newly added stories of intangible cultural heritage inheritors, new intangible cultural heritage works, and new forms of folk activities; updating the intangible cultural heritage element database using incremental learning to ensure the timeliness of AIGC-generated content; and revising the structured tags of cultural elements based on user feedback and cultural expert review to avoid cultural distortion and ensure the accurate transmission of the core connotations of intangible cultural heritage. (Embodiment)

[0047] As shown in Figure 2, this embodiment of the invention proposes a county-level intangible cultural heritage AIGC content push system. This system implements the method shown in Embodiment 1, including: an element library construction module for structurally extracting county-level intangible cultural heritage elements and constructing an intangible cultural heritage element library; a user interest image construction module for collecting multi-source user data and constructing user interest images; a personalized content generation module for retrieving matching cultural elements from the intangible cultural heritage element library based on user interest images and the dissemination scenario, and inputting them into an intangible cultural heritage-specific AIGC generation model to generate personalized AIGC dissemination content; a content push module for determining push channels and dynamically pushing personalized AIGC dissemination content using a preset hybrid push algorithm; and a dissemination effect feedback module for collecting push effect data and optimizing the parameters of the intangible cultural heritage-specific AIGC generation model using a reinforcement learning algorithm.

[0048] In summary, this invention achieves precise, personalized, and efficient delivery of intangible cultural heritage dissemination content in counties through a closed-loop design that includes structured extraction of intangible cultural heritage elements, AIGC personalized content generation, deep user profiling, scenario-based dynamic push, and effect feedback optimization. This solves the problems of narrow coverage, poor adaptability, and high cost associated with traditional dissemination methods, while ensuring the accurate transmission of the connotations of intangible cultural heritage.

[0049] The technical solution provided by this invention has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the core idea of ​​the method of this invention. It should be noted that those skilled in the art can make several improvements and modifications to this invention without departing from the principles of this invention, and these improvements and modifications also fall within the protection scope of the claims of this invention.

Claims

1. A method for pushing AIGC (Artificial Intelligence Generated Content) dissemination materials on intangible cultural heritage at the county level, characterized in that, The process includes the following steps: Step 1: Structured extraction of intangible cultural heritage elements from the county level to construct an intangible cultural heritage element database; Step 2: Collection of multi-source user data and construction of user interest images; Step 3: Based on user interest images and dissemination scenarios, retrieval of matching cultural elements from the intangible cultural heritage element database and input into the intangible cultural heritage-specific AIGC generation model to generate personalized AIGC dissemination content; Step 4: Determination of push channels and dynamic push of personalized AIGC dissemination content using a preset hybrid push algorithm; Step 5: Collection of push effect data and optimization of the parameters of the intangible cultural heritage-specific AIGC generation model based on the push effect data.

2. The method for pushing county-level intangible cultural heritage AIGC dissemination content according to claim 1, characterized in that: The steps in step 1 to construct the intangible cultural heritage element database include: Step 1.1: Collect multi-dimensional data on county-level intangible cultural heritage and establish a county-level intangible cultural heritage database; Step 1.2: Preprocess the data in the county-level intangible cultural heritage database; Step 1.3: Construct a three-level structured system of intangible cultural heritage core attributes, cultural elements, and dissemination tags, and process the preprocessed data, wherein the intangible cultural heritage core attributes include intangible cultural heritage categories, regional characteristics, and cultural connotations; Step 1.4: Assign a unique identifier to each cultural element in the three-level structured system and associate it with the corresponding county-level intangible cultural heritage data to form a structured intangible cultural heritage element database.

3. The method for pushing county-level intangible cultural heritage AIGC dissemination content according to claim 2, characterized in that: The multi-dimensional data collected in step 1.1 includes: text data, image data, and audio-visual data. The text data includes the historical origins of intangible cultural heritage, the process of making techniques, and the folk customs. The image data includes images of intangible cultural heritage works, images of the process of making techniques, and images of inheritors. The audio-visual data includes intangible cultural heritage performances, technique demonstrations, and records of folk activities.

4. The method for pushing county-level intangible cultural heritage AIGC dissemination content according to claim 1, characterized in that: Step 2 involves collecting multi-source user data and constructing user interest images, including: Step 2.1: Collecting static user data, dynamic behavior data, and intangible cultural heritage interaction data to generate a user dataset; Step 2.2: Combining the intangible cultural heritage element library, inputting the standardized user dataset into a pre-trained user interest feature extraction model for feature extraction to obtain user interest profile vectors; Step 2.3: Constructing user interest images based on user interest profile vectors.

5. The method for pushing county-level intangible cultural heritage AIGC dissemination content according to claim 4, characterized in that: The user interest feature extraction model includes: a basic interest layer, which extracts user preferences for intangible cultural heritage categories and content formats based on dynamic behavioral data; a deep interest layer, which analyzes user preferences for specific cultural elements by combining an intangible cultural heritage element library and calculates user interest weights; a regional association layer, which associates the user's current location with the regional matching degree of county-level intangible cultural heritage to determine the push weight of regionally related intangible cultural heritage content; and an output layer, which outputs a user interest profile vector including multi-dimensional features such as basic attributes, interest preferences, relevance, and cultural acceptance.

6. The method for pushing county-level intangible cultural heritage AIGC dissemination content according to claim 1, characterized in that: Step 3, which generates AIGC personalized dissemination content, includes the following steps: Step 3.1: Based on user interest images and dissemination scenarios, retrieve matching cultural elements from the intangible cultural heritage element library to determine the theme, form, and cultural core of the dissemination content; Step 3.2: Construct an AIGC generation model exclusive to intangible cultural heritage, input the matched cultural elements, user interest images, and scenario requirements, and generate initial personalized dissemination content; Step 3.3: Use cultural verification rules to filter the personalized dissemination content, identify culturally distorted content, and obtain the AIGC personalized dissemination content.

7. The method for pushing county-level intangible cultural heritage AIGC dissemination content according to claim 1, characterized in that: The process of dynamically pushing AIGC personalized content using a preset hybrid push algorithm in step 4 includes: Step 4.1: Filtering AIGC personalized content with similar user preferences based on user interest similarity; Step 4.2: Calculating the similarity between the filtered AIGC personalized content and user interest images, and including content with similarity greater than a threshold in the candidate recommendation list; Step 4.3: Adjusting the weight of the pushed content according to the propagation scenario, and pushing the AIGC personalized content to the user; Step 4.4: Collecting user interaction feedback data in real time, and adjusting the push frequency based on user interaction feedback.

8. The method for pushing county-level intangible cultural heritage AIGC dissemination content according to claim 1, characterized in that: The push effect data mentioned in step 5 includes: core effect indicators, cultural transmission indicators, and conversion indicators. Among them, the core effect indicators include content open rate, dwell time, collection rate, sharing rate, forwarding rate, and interaction participation rate; the cultural transmission indicators include the correctness rate of knowledge quizzes and the mention rate of intangible cultural heritage elements in user comments; and the conversion indicators include the number of cultural tourism reservations, the purchase volume of intangible cultural heritage products, and the number of participants in offline activities.

9. The method for pushing county-level intangible cultural heritage AIGC dissemination content according to claim 1, characterized in that: The push effect data mentioned in step 5 is also used to optimize the push strategy in step 4.

10. A county-level intangible cultural heritage AIGC dissemination content push system based on the method of any one of claims 1-9, characterized in that, include: The element library construction module is used for the structured extraction of intangible cultural heritage elements in the county and the construction of an intangible cultural heritage element library. The user interest image construction module is used to collect multi-source user data and construct user interest images; the personalized dissemination content generation module is used to retrieve matching cultural elements from the intangible cultural heritage element library based on user interest images and dissemination scenarios, and input them into the intangible cultural heritage exclusive AIGC generation model to generate AIGC personalized dissemination content; the content push module is used to determine the push channel and dynamically push AIGC personalized dissemination content using a preset hybrid push algorithm. The dissemination effect feedback module is used to collect push effect data and optimize the parameters of the intangible cultural heritage-specific AIGC generation model using reinforcement learning algorithms.

Images