Intelligent display method and system for cultural and creative products

Abstract

The application discloses an intelligent display method and system for cultural and creative products, and relates to the technical field of digital display. The disclosed intelligent display method and system for cultural and creative products can realize personalized display, dynamically adjust display content according to user interests, and improve user participation and cultural communication depth by acquiring user interest data, generating a cognitive vector, searching for a cultural node, matching a similarity, generating a narrative path, dynamically rendering elements, and adjusting display content in real time.

Description

Technical Field

[0001] This application relates to the field of digital display technology, and in particular to intelligent display methods and systems for cultural and creative products. Background Technology

[0002] In the digital transformation of the cultural and creative industries, online display technologies have evolved from basic text and image descriptions to advanced forms such as 3D model interaction and augmented reality. However, current mainstream display systems still adhere to a one-way information output model, with their core mechanisms relying on preset templates or fixed logic, such as standardized product model rotation demonstrations, scripted video narratives, or content pushes based on simple category tags. This design paradigm suffers from two structural defects: First, the system lacks the ability to dynamically perceive the individual cognitive state of users. It cannot effectively integrate explicit interests and preferences expressed by users, nor can it analyze implicit emotional feedback and attention distribution conveyed by users through physiological signals such as facial expressions and eye movements during previews. This results in highly homogenized display content, making it difficult to match users' unique cultural knowledge background and immediate emotional needs, thereby weakening the depth of cultural resonance and the strength of emotional connection. Second, the display logic is strictly confined to a closed, predefined framework. The relationship between cultural elements and products is statically solidified, making it impossible to dynamically reconstruct the topology and optimize the narrative path based on users' real-time interactive behavior. This makes the exploration process lack flexibility and inspiration, and users find it difficult to obtain a rich and layered cultural cognitive experience.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide a smart display method and system for cultural and creative products, aiming to improve user participation and the depth of cultural dissemination.

[0005] To achieve the above objectives, this application proposes a method for intelligent display of cultural and creative products, the method comprising: Acquire the user's explicit interest data and implicit interest data, and perform weighted fusion processing on the explicit interest data and implicit interest data to generate an initial user cognitive vector; Based on the target cultural and creative product, cultural nodes and relationships associated with the target cultural and creative product are retrieved from a pre-trained multi-dimensional cultural knowledge graph database to obtain initial association network data. The initial user cognition vector is matched with each cultural node in the initial association network data to calculate vector similarity. Based on the matching results, each cultural node in the initial association network data is assigned an initial association weight. The cultural nodes are then sorted and filtered based on the initial association weight to generate initial narrative path data. A digital model of the target cultural and creative product is loaded into the display interface, and based on the initial narrative path data, visual narrative elements distributed around the digital model are dynamically rendered and generated in the display interface; the spatial layout and visual salience of the visual narrative elements are determined by the initial association weight of their corresponding cultural nodes. The system captures user interaction data in the display interface in real time, and dynamically adjusts the initial narrative path data based on the interaction data and the current narrative state to generate updated narrative path data. Based on this, the system updates the visual narrative elements and their content in the display interface in real time.

[0006] The steps of acquiring users' explicit and implicit interest data, and performing weighted fusion processing on the explicit and implicit interest data to generate an initial user cognitive vector include: The system presents users with preset cultural level options for selection, records the user's selected options, and parses the option information into explicit interest vector data. When a user views a preview image of the target cultural and creative product for the first time, facial image sequence data of the user during this period is collected by an image acquisition device. The facial image sequence data is then used for fixed-point region identification and change analysis to obtain the user's visual attention distribution data and emotional characteristic data, which are then used as latent interest data. The explicit interest data and the implicit interest data are weighted and fused to generate the initial user cognition vector.

[0007] The facial image sequence data is subjected to pinpoint region identification and change analysis to obtain the user's visual attention distribution data and emotional feature data, including: The user's eye region image is identified frame by frame from the facial image sequence data; Based on a pre-trained image analysis model, the eye region image is analyzed to obtain pupil size change sequence data. Based on the correspondence between the pupil size change sequence data and different regions in the preview image, the relative attention intensity of the user to different regions in the preview image is determined to obtain the visual attention distribution data. The system identifies key facial feature points of the user from the facial image sequence data, matches the displacement changes of the key facial feature points with a pre-trained micro-expression feature library to determine whether the user exhibits a preset type of micro-expression, and generates the emotion feature data based on the determination result.

[0008] The process of sorting and filtering cultural nodes based on the initial association weights to generate initial narrative path data includes: Cultural nodes in the initial association network data whose initial association weight is greater than a first preset threshold are marked as core narrative nodes; Cultural nodes in the initial association network data whose initial association weight is less than the first preset threshold but greater than the second preset threshold are marked as associated narrative nodes; Based on the connection relationship between the core narrative node and the associated narrative node in the pre-trained multi-dimensional cultural knowledge graph database, a tree-like or graph-like structure data is constructed with the core narrative node as the main line and the associated narrative nodes as branches, as the initial narrative path data.

[0009] Based on the initial narrative path data, visual narrative elements distributed around the digital model are dynamically rendered and generated in the display interface, including: Based on the weights and types of cultural nodes in the initial narrative path data, they are mapped to a preset visual template library to determine the corresponding visual element type, basic style, and dynamic behavior parameters for each cultural node. Based on the relationships between cultural nodes in the initial narrative path data, the relative positional relationships of each visual element in the three-dimensional or two-dimensional display space are determined; Based on the visual element type, basic style, dynamic behavior parameters, and relative positional relationships of each visual element determined for each cultural node, the cultural node is rendered as a visual narrative element with corresponding style, position, and dynamic behavior in the display interface.

[0010] Based on the interactive operation data and the currently displayed narrative state, the initial narrative path data is dynamically adjusted to generate updated narrative path data, including: When the interactive operation data is a user's trigger operation on a specific visual narrative element in the display interface, the cultural node corresponding to that specific visual narrative element is identified as the currently active node. From the pre-trained multi-dimensional cultural knowledge graph database, retrieve the neighboring cultural nodes that are directly connected to the currently activated node; Calculate the real-time correlation between the adjacent cultural nodes and the initial user cognition vector, and calculate the content difference between the adjacent cultural nodes and the currently displayed narrative node set; the currently displayed narrative node set is defined by the initial narrative path data and dynamically updated according to historical adjustments; Based on the weighted sum of the real-time relevance and the content difference, a dynamic recommendation weight is calculated for each adjacent cultural node; Based on the dynamic recommendation weights, the corresponding adjacent cultural nodes are inserted or promoted in the position of the initial narrative path data to generate the updated narrative path data.

[0011] Calculate the content difference between the adjacent cultural nodes and the currently displayed set of narrative nodes, including: Using a pre-trained text feature extraction model, extract the text feature vectors of the adjacent cultural nodes and each node in the currently displayed set of narrative nodes; Calculate the average cosine distance between the text feature vector of the adjacent cultural node and the text feature vectors of all nodes in the currently displayed narrative node set, and use the average cosine distance as a measure of the content difference.

[0012] The method further includes: During the demonstration, the narrative field complexity index is continuously calculated; the calculation of the narrative field complexity index is based on the number of currently active nodes in the updated narrative path data, the number of relationships between the active nodes, and the average similarity between the active nodes and the initial user cognition vector; When the narrative field complexity index exceeds a preset upper limit threshold, a path convergence operation is triggered; the path convergence operation includes: Increase the weight of nodes directly associated with the currently active node in the updated narrative path data; Reduce the weight of nodes that are indirectly associated with the currently active node and whose dynamic recommendation weight is lower than a third preset threshold, and regenerate the updated narrative path data accordingly. When the narrative field complexity index is lower than a preset lower threshold, a path expansion operation is triggered; the path expansion operation includes: From the pre-trained multi-dimensional cultural knowledge graph database, retrieve nodes that have multi-hop associations with the currently activated node and whose similarity to the initial user cognition vector is greater than a fourth preset threshold, and add them as potential expansion nodes to subsequent calculations; For each potential expansion node, calculate its real-time correlation with the initial user cognition vector and its content difference with the currently displayed set of narrative nodes; The dynamic recommendation weight is calculated based on the weighted sum of the real-time relevance and the content difference, and the narrative path data is updated based on the dynamic recommendation weight.

[0013] The method further includes: The memory data recorded during the interaction process is based on a preset instantaneous narrative memory; the memory data includes the sequence of cultural nodes identified as the currently active node, the timestamp of each node being activated, and the type of interaction operation that triggered the activation of the node; Based on the memory data recorded in the instantaneous narrative memory, the user's interaction patterns are analyzed; Based on the analysis results of the interaction mode, a pre-trained sequence prediction model is used to predict a set of potential cultural nodes from the pre-trained multi-dimensional cultural knowledge graph database, according to the currently activated node and the sequence of cultural nodes recorded in the instantaneous narrative memory. Before the user performs the next interaction, some or all of the associated data of the potential cultural node set is loaded into the cache from the pre-trained multi-dimensional cultural knowledge graph database.

[0014] A smart display system for cultural and creative products, comprising: a memory, a processor, and a smart display program for cultural and creative products stored in the memory and executable on the processor, wherein the smart display program for cultural and creative products is configured to implement the steps of the smart display method for cultural and creative products.

[0015] The intelligent display method and system for cultural and creative products proposed in this application can achieve personalized display by acquiring user interest data, generating cognitive vectors, retrieving cultural nodes, matching similarities, generating narrative paths, dynamically rendering elements, and adjusting display content in real time. It can dynamically adjust the display content according to user interests, thereby improving user participation and the depth of cultural dissemination. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating an embodiment of the intelligent display method for cultural and creative products according to this application; Figure 2 This is a structural diagram of an embodiment of the intelligent display system for cultural and creative products provided in this application.

[0019] Explanation of icon numbers: 10. Memory; 20. Processor.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of this application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be understood that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] In existing technologies, online displays of cultural and creative products generally treat the display as a one-way output process of the product's inherent information. They fail to perceive and integrate users' existing cultural background, knowledge reserves, and immediate emotional feedback, resulting in a one-size-fits-all display that fails to evoke deep cultural resonance and emotional connection. Secondly, their display logic is static and closed, with the connection between cultural elements and products pre-fixed. This prevents dynamic expansion and narrative reorganization based on users' real-time interactive behavior, limiting the depth and inspiration of exploration.

[0024] Based on this, the embodiments of this application provide a method for intelligent display of cultural and creative products, referring to... Figure 1 The intelligent display method for cultural and creative products includes steps S100 to S500, wherein: Step S100: Obtain the user's explicit interest data and implicit interest data, and perform weighted fusion processing on the explicit interest data and the implicit interest data to generate an initial user cognitive vector. Step S200: Based on the target cultural and creative product, retrieve cultural nodes and relationships associated with the target cultural and creative product from a pre-trained multi-dimensional cultural knowledge graph database to obtain initial association network data; Step S300: Perform vector similarity matching calculation between the initial user cognition vector and each cultural node in the initial association network data; assign initial association weights to each cultural node in the initial association network data according to the matching results; and sort and filter the cultural nodes based on the initial association weights to generate initial narrative path data. Step S400: Load the digital model of the target cultural and creative product into the display interface, and dynamically render and generate visual narrative elements distributed around the digital model in the display interface based on the initial narrative path data; the spatial layout and visual salience of the visual narrative elements are determined by the initial association weight of their corresponding cultural nodes. Step S500: Capture user interaction data in the display interface in real time, and dynamically adjust the initial narrative path data based on the interaction data and the current narrative state to generate updated narrative path data, and update the visual narrative elements and their presentation content in the display interface in real time accordingly.

[0025] In this embodiment, explicit interest data refers to interest information acquired by users through active selection, input, or explicit expression, such as user choices in cultural level options and collection preferences. Implicit interest data refers to interest information displayed by users unconsciously or involuntarily, such as data passively captured through facial expressions, eye movements, and interactive behaviors. The initial user cognitive vector is a vector representation formed by weighted fusion of explicit and implicit interest data, used to quantify the user's comprehensive cognitive state and interest preferences for specific cultural content. The multi-dimensional cultural knowledge graph database refers to a pre-constructed structured database containing rich cultural knowledge, where cultural knowledge is stored in the form of nodes and relationships. Nodes represent cultural entities, and relationships represent the connections between these entities. This graph supports multi-dimensional queries and reasoning. A cultural node refers to a basic unit representing a specific cultural entity or concept in the multi-dimensional cultural knowledge graph database.

[0026] In this embodiment, the initial association network data refers to the network structure data composed of cultural nodes and their interrelationships that are directly or indirectly associated with the target cultural and creative product, retrieved from a multi-dimensional cultural knowledge graph database based on the target cultural and creative product. The initial narrative path data refers to the cultural narrative sequence with logical order and strong association, formed after sorting and filtering the cultural nodes in the initial association network data according to the matching results of user cognitive vectors and cultural nodes. This data defines the initial narrative clues for the display of the cultural and creative product. Visual narrative elements refer to the visual representations of specific cultural nodes dynamically generated and distributed around the digital model of the cultural and creative product on the display interface, such as icons, text bubbles, dynamic effects, and small 3D models. Interactive operation data refers to the data generated when users interact with visual narrative elements or digital models on the display interface, such as clicking, swiping, zooming, and rotating. Narrative state refers to the presentation status of visual narrative elements on the current display interface, the narrative paths explored by the user, and the system's real-time assessment of user interest.

[0027] The intelligent display method for cultural and creative products first acquires users' explicit and implicit interest data, and then performs a weighted fusion process on this explicit and implicit interest data to generate an initial user cognitive vector. Specifically, explicit interest data can be obtained through users actively selecting cultural themes, filling out interest questionnaires, or recording users' browsing history and collection preferences. For example, the system can present users with a series of cultural category options, which users can select to express their interests. Implicit interest data can be obtained by analyzing users' behavioral patterns during browsing, such as recording users' mouse movement paths, click hotspots, and page dwell time. This data can be converted into numerical or vector forms. Subsequently, the explicit and implicit interest data are weighted and fused, for example, by combining them through simple arithmetic averaging or adding them with preset weights. The fused data is then further processed, for example, through dimensionality reduction algorithms or encoders, to convert it into a fixed-dimensional vector. This vector serves as the initial user cognitive vector, used to quantify the user's overall cognitive state and interest preferences regarding specific cultural content.

[0028] Furthermore, based on the target cultural and creative product, the system retrieves cultural nodes and their relationships associated with the target product from a pre-trained multi-dimensional cultural knowledge graph database to obtain initial association network data. Specifically, the system receives a cultural and creative product to be displayed as input, such as through its product identifier or detailed description. Subsequently, the system searches for cultural nodes directly or indirectly related to the product in the multi-dimensional cultural knowledge graph database through methods such as keyword matching, semantic similarity calculation, or graph traversal. For example, when the target cultural and creative product is "Silk Road murals," the system can retrieve cultural nodes such as "Dunhuang," "Mogao Grottoes," "Buddhist art," and "Zhang Qian." The retrieved cultural nodes and their connections in the knowledge graph are extracted to form a temporary network structure, which serves as the initial association network data.

[0029] Based on this, vector similarity matching is performed between the initial user cognition vector and each cultural node in the initial association network data. Initial association weights are assigned to each cultural node in the initial association network data according to the matching results. These initial association weights are then used to sort and filter the cultural nodes to generate initial narrative path data. Specifically, cosine similarity, the inverse of Euclidean distance, or other distance metrics can be used to calculate the similarity between the initial user cognition vector and the vector corresponding to each cultural node in the initial association network data. Based on the similarity matching results, each cultural node is assigned a numerical value reflecting the strength of the association between the cultural node and the user cognition vector, serving as the initial association weight. Subsequently, the cultural nodes can be sorted in descending order according to the magnitude of the initial association weights, and nodes with weights higher than a preset threshold, or the top N nodes with the highest weights, can be selected for filtering. The filtered cultural nodes can be constructed into a preliminary narrative sequence or graph structure according to their connection relationships or weight order in the knowledge graph; this structure serves as the initial narrative path data.

[0030] The system loads a digital model of the target cultural and creative product onto the display interface. Based on the initial narrative path data, it dynamically renders and generates visual narrative elements distributed around the digital model. The spatial layout and visual salience of these visual narrative elements are determined by the initial association weights of their corresponding cultural nodes. Specifically, a 3D model or a high-precision 2D image of the target cultural and creative product can be loaded into the display interface using a 3D rendering engine. Based on the cultural node information contained in the initial narrative path data, the system can select corresponding icons, text boxes, lighting effects, etc., from a preset universal visual element library and dynamically present them around the digital model. For example, a simple circular icon can be assigned to each cultural node, displaying its name. The initial association weights of the cultural nodes can affect the position, size, color, transparency, or animation effects of the visual narrative elements in the display space. For example, elements corresponding to nodes with higher weights can be placed near the digital model and displayed in a larger size or brighter color to improve their visual salience.

[0031] Finally, the system captures user interaction data on the display interface in real time. Based on this data and the current narrative state, it dynamically adjusts the initial narrative path data to generate updated data, and accordingly updates the visual narrative elements and their content on the interface. Specifically, the system listens for all user input events on the interface, such as mouse clicks, keyboard input, and touch gestures, and records them as interaction data. The current narrative state can include activated cultural nodes, displayed narrative path segments, and the user's current focus. Based on the captured interaction data and the current narrative state, the system reassesses the relevance of cultural nodes and modifies the initial narrative path data. For example, if a user clicks on a visual narrative element, the system can mark its corresponding cultural node as high priority and add its associated nodes to the narrative path. The adjusted narrative path data forms a new narrative sequence or graph structure, serving as the updated narrative path data. Based on the updated data, the visual narrative elements on the interface are re-rendered, including their quantity, position, style, and content, to achieve real-time updates of the narrative content.

[0032] This intelligent display method for cultural and creative products integrates users' explicit and implicit interest data to construct a personalized user cognitive model. Combined with a multi-dimensional cultural knowledge graph, it dynamically generates and adjusts the narrative path surrounding the cultural and creative products. As a result, the displayed content deeply resonates with users' personalized cultural backgrounds and immediate emotional responses, effectively avoiding the one-size-fits-all approach of traditional displays. This enhances users' cultural resonance and depth of exploration of the cultural and creative products, achieving a dynamic and open display logic.

[0033] In one feasible implementation, the steps of acquiring a user's explicit interest data and implicit interest data, and performing weighted fusion processing on the explicit interest data and implicit interest data to generate an initial user cognitive vector include: presenting the user with preset cultural level options for selection, recording the user's selected option information, and parsing the option information into explicit interest vector data; when the user first views the preview image of the target cultural and creative product, acquiring facial image sequence data of the user during this period through an image acquisition device, performing fixed-point region identification and change analysis on the facial image sequence data to obtain the user's visual attention distribution data and emotional feature data, and using these as implicit interest data; and performing weighted fusion processing on the explicit interest data and implicit interest data to generate the initial user cognitive vector.

[0034] In this embodiment, regarding the acquisition of explicit interest data, the system presents users with preset cultural level options for selection. These options can cover multiple dimensions such as the cultural background, historical period, artistic style, and thematic elements involved in cultural and creative products. For example, they can be displayed in the form of multi-level menus, tag clouds, or questionnaires, showing specific categories such as "Tang Dynasty Culture," "Song Dynasty Aesthetics," "Bronze Age Art," and "Silk Road." Users express their explicit interest preferences by clicking, checking boxes, or inputting information, and the system records the option information of these selections. Subsequently, this option information is parsed into explicit interest vector data. For example, through techniques such as one-hot encoding, word embedding, or TF-IDF, the user's selection is quantified into a multi-dimensional vector, where each dimension of the vector corresponds to a cultural level option, and its value reflects the user's interest intensity in that option.

[0035] In acquiring latent interest data, when a user first views a preview image of the target cultural and creative product, the system collects a sequence of facial images of the user during this period using an image acquisition device. The preview image can be a static picture, short video, or thumbnail of a 3D model of the cultural and creative product, designed to create a strong first impression. The image acquisition device is typically a built-in camera on the user's device (such as a smartphone, tablet, or computer) or an external high-definition camera. The system captures the user's facial images in consecutive frames within a specific time window of the preview image viewing, forming a time-series data stream for subsequent analysis of the dynamic changes in the user's facial expressions and eye movements.

[0036] Fixed-point region identification and change analysis are performed on the facial image sequence data to obtain user visual attention distribution data and emotional feature data, which are then used as latent interest data. Fixed-point region identification and change analysis refers to using computer vision and image processing technology to identify key areas of the user's face from the facial image sequence data, such as the eyes, eyebrows, and mouth, and analyzing the dynamic changes of these areas over time. For example, eye-tracking technology can be used to analyze the time the user's gaze lingers, the scanning path, and the fixation point in different areas of the preview image, thereby generating visual attention distribution data and quantifying the user's level of attention to different components of the cultural and creative products. Simultaneously, by analyzing subtle facial muscle movements, such as slight eyebrow raising and minor changes in the corners of the mouth, combined with a pre-trained facial expression recognition model, the user's emotional state, such as pleasure, surprise, and confusion, can be identified, generating emotional feature data. This visual attention distribution data and emotional feature data represent the physiological and behavioral responses generated by the user in an unconscious state, and can objectively and truthfully reflect their latent interest in cultural and creative products.

[0037] The explicit interest data and the implicit interest data are weighted and fused to generate the initial user cognition vector. The weighted fusion process integrates the acquired explicit interest vector data and implicit interest data (including visual attention distribution data and emotional feature data). Fusion methods may include, but are not limited to: concatenating the explicit and implicit interest vectors and then performing feature extraction and dimensionality reduction through a fully connected neural network; or assigning different weights to the explicit and implicit interest data and then performing linear weighted summation or nonlinear fusion. The weights can be set based on empirical values, machine learning model training results, or dynamically adjusted according to the characteristics of the user interaction scenario. For example, when a user first encounters a product, the weight of implicit interest data may be higher to capture their initial reaction; while when a user has a clear selection history, the weight of explicit interest data may be higher. The final generated initial user cognition vector is a high-dimensional vector that can comprehensively and quantitatively represent the user's overall cognition and interest preferences for cultural and creative products.

[0038] Through the aforementioned technical solution, this application overcomes the limitations of traditional methods that rely solely on user input or behavioral logs when acquiring user interest data, making it difficult to capture deep, unconscious user interests. By introducing the selection of preset cultural level options, the system can directly acquire explicit user interest data, ensuring the reflection of the user's subjective will. Simultaneously, when the user first views the preview image of the cultural and creative product, the system uses an image acquisition device to capture facial image sequence data in real time, and performs fixed-point region identification and change analysis. This allows for the objective and non-invasive acquisition of the user's visual attention distribution data and emotional characteristic data, thereby accurately quantifying the user's implicit interests. This combination of explicit and implicit interest data, through weighted fusion processing, ensures that the generated initial user cognitive vector not only includes the user's explicit preferences but also incorporates their subconscious, genuine physiological and emotional responses. This significantly improves the comprehensiveness, accuracy, and real-time nature of the user cognitive vector, providing a more solid and personalized foundation for subsequent narrative path generation based on cultural knowledge graphs. This enables the intelligent display of cultural and creative products to more accurately match the user's deep interests, providing a more attractive and immersive personalized narrative experience.

[0039] In one feasible implementation, the facial image sequence data is subjected to fixed-point region identification and change analysis to obtain the user's visual attention distribution data and emotional feature data, including: identifying the user's eye region image frame by frame from the facial image sequence data; analyzing the eye region image based on a pre-trained image analysis model to obtain pupil size change sequence data, and determining the relative attention intensity of the user to different regions in the preview image based on the correspondence between the pupil size change sequence data and different regions in the preview image, so as to obtain the visual attention distribution data; identifying the user's facial key feature points from the facial image sequence data, matching the displacement changes of the facial key feature points with a pre-trained micro-expression feature library based on the displacement changes of the facial key feature points, determining whether the user exhibits a preset type of micro-expression, and generating the emotional feature data based on the determination result.

[0040] In this embodiment, the eye region is accurately located and extracted from continuous user facial images, providing foundational data for subsequent visual attention analysis. Specifically, various computer vision techniques can be employed. For example, a deep learning-based face detection model (such as MTCNN, RetinaFace, etc.) can be used to first detect the face, and then a facial landmark detection model (such as Dlib's 68-point or more refined 106-point model) can be used to identify the key points of the eye contour, thereby cropping out the image region containing the eyes. Alternatively, specialized eye detection algorithms, such as those based on Haar feature classifiers or cascaded classifiers, can be used to directly identify the eye region in the facial image. Frame-by-frame recognition ensures real-time or near-real-time tracking of the user's eye state, laying the foundation for dynamic analysis of pupil changes.

[0041] The pre-trained image analysis model can be a deep learning model specifically designed for pupil detection and size measurement (e.g., semantic segmentation based on U-Net or Mask R-CNN architectures for accurate pupil region identification), or a visual analysis module combining traditional image processing algorithms (such as Hough transform, edge detection, and circle fitting). This model processes frame-by-frame identified eye region images to continuously acquire dimensional data such as pupil diameter or area, forming a sequence of pupil size changes. Subsequently, by analyzing this sequence of pupil size changes and combining it with the content of the preview image currently viewed by the user—for example, if the user's pupil size changes significantly when viewing a specific area of ​​the preview image (e.g., magnification)—it indicates a high cognitive load or interest in that area. By correlating pupil size changes with different content areas in the preview image (e.g., different details of cultural and creative products, background elements, etc.), the relative intensity of the user's attention to these areas can be quantified, thus constructing data on the user's visual attention distribution.

[0042] Facial landmark detection techniques (e.g., using convolutional neural network-based feature point detectors such as FAN or OpenPose's facial module) are used to accurately locate and track a series of key facial feature points, such as the contour points of eyebrows, eyes, nose, and mouth, in a sequence of facial images. The displacement changes of these key feature points—that is, their relative position or shape changes between consecutive frames—are important indicators of micro-expressions. A pre-trained micro-expression feature library is a database containing typical facial landmark displacement patterns for various known micro-expressions (such as surprise, disgust, happiness, sadness, anger, fear, contempt, etc.). This feature library can be trained using a large number of labeled micro-expression datasets to build classification models (such as Support Vector Machines (SVM), Random Forests, or deep learning classifiers) capable of recognizing different micro-expressions. By matching real-time captured facial landmark displacement changes with this feature library, the system can determine whether the user exhibits a preset type of micro-expression. Based on the matching results, the user's emotional feature data can be generated, for example, represented as discrete emotion category labels, emotion intensity values, or emotion probability distributions.

[0043] Through the aforementioned technical solution, this application overcomes the challenge of accurately capturing and quantifying users' visual attention and emotional state when acquiring latent interest data. By identifying eye regions frame by frame and analyzing pupil size changes, the system can meticulously perceive the relative intensity of user attention to different areas of the preview image of cultural and creative products, thereby obtaining more refined and accurate visual attention distribution data. Simultaneously, by identifying displacement changes of key facial feature points and matching them with a pre-trained micro-expression feature library, this application can effectively capture users' unconscious and genuine micro-expressions, thereby generating reliable emotional feature data. These precise visual attention distribution data and emotional feature data, as high-quality latent interest data, can improve the accuracy and richness of the initial user cognitive vector. Based on this, the weighted fusion-processed user cognitive vector will more realistically reflect the user's deep interests and preferences, thus ensuring that the subsequently generated initial narrative path data highly matches the user's potential needs and provides a more solid foundation for dynamically adjusting the narrative path, ultimately greatly enhancing the personalized, immersive experience and user satisfaction of the intelligent display of cultural and creative products.

[0044] In one feasible implementation, the step of sorting and filtering cultural nodes based on the initial association weights to generate initial narrative path data includes: marking cultural nodes in the initial association network data whose initial association weights are greater than a first preset threshold as core narrative nodes; marking cultural nodes in the initial association network data whose initial association weights are less than the first preset threshold but greater than a second preset threshold as associated narrative nodes; and constructing a tree-like or graph-like structure data with the core narrative nodes as the main line and the associated narrative nodes as branches, based on the connection relationship between the core narrative nodes and the associated narrative nodes in the pre-trained multi-dimensional cultural knowledge graph database, as the initial narrative path data.

[0045] In this embodiment, the first preset threshold is a numerical boundary used to distinguish the importance of cultural nodes. Its setting can be based on various factors, such as expert experience or automatic learning through analysis of large amounts of user behavior data using machine learning models (such as clustering or classification algorithms). This threshold aims to filter out cultural nodes that highly align with user interests and have high salience in the cultural knowledge graph, ensuring that these nodes can serve as the backbone of the narrative. Core narrative nodes refer to cultural elements that the system determines have the highest relevance to user interests and a core position in the cultural knowledge graph during the intelligent display of cultural and creative products. These nodes typically represent the most important and attractive cultural connotations or storylines of the cultural and creative products. They are the skeleton of the narrative path, carrying key information conveying the core value of the product to the user.

[0046] The second preset threshold is also a numerical limit, its value being less than the first preset threshold. The method for setting this threshold is similar to that of the first preset threshold, and can be based on experience or data-driven learning. Its purpose is to further filter out cultural nodes that, while not as important as the core narrative nodes, still have a certain relevance to user interests and can supplement and expand the core narrative. Relevant narrative nodes refer to cultural elements that, during the intelligent display of cultural and creative products, are determined by the system to have a moderate relevance to user interests and can enrich the core narrative content. These nodes typically serve as supplements or extensions to the core narrative nodes, providing background knowledge, related allusions, derivative stories, etc., making the narrative content more complete and three-dimensional. They are branches of the narrative path, used to expand the depth and breadth of users' understanding of the cultural connotations of cultural and creative products.

[0047] The initial narrative path data is constructed using tree-like or graph-like structures, with core narrative nodes as the main thread and related narrative nodes as branches. This is based on the connection relationships in a pre-trained multi-dimensional cultural knowledge graph database. The pre-trained multi-dimensional cultural knowledge graph database is a structured knowledge base containing a large number of cultural entities (such as historical figures, events, art styles, geographical locations, myths, and legends) and the complex relationships between them. This knowledge graph, after pre-training and construction, can represent the hierarchical structure and semantic connections of cultural knowledge. Connection relationships refer to the semantic associations between different cultural nodes in the cultural knowledge graph, such as "belongs to," "contains," "influences," "originates from," and "related to." These relationships define the inherent logical structure of cultural knowledge. When constructing tree-like or graph-like structures, tree-like structures typically represent a hierarchical relationship, with core narrative nodes as the trunk and related narrative nodes as branches, forming a clear narrative hierarchy. Graph-like structures are more flexible, capable of representing more complex non-linear relationships, allowing narrative paths to jump and intersect between different nodes. The construction process may include: starting from the core narrative node, traversing the related narrative nodes in the knowledge graph that are directly or indirectly connected to the core narrative node, and determining their positions and connection methods in the tree or graph structure based on their connection relationships (such as parent-child relationships, causal relationships, chronological order, etc.). For example, breadth-first search (BFS) or depth-first search (DFS) algorithms can be used to start from the core node, extend along the edges in the knowledge graph to related nodes, and record the path.

[0048] Through the aforementioned technical solution, this application effectively distinguishes cultural nodes with varying degrees of relevance to user interests, clearly dividing them into core narrative nodes and related narrative nodes. This hierarchical approach avoids the problem of presenting all relevant cultural nodes in a flat, one-dimensional manner, enabling the system to identify the core cultural connotations of cultural and creative products and the key points of user interest. Based on this, by utilizing the rich connections in a pre-trained multi-dimensional cultural knowledge graph database, a tree-like or graph-like data structure with clear logical hierarchy is constructed as the initial narrative path, using core narrative nodes as the main thread and related narrative nodes as branches. This structured narrative path, compared to simple weighted sorting, provides a more coherent and easily understood cultural storyline. It not only highlights the core value of cultural and creative products but also enriches the narrative details through branch content, allowing users to gradually gain a deeper understanding of the cultural background and story of cultural and creative products, thereby improving the efficiency of users' understanding of the cultural connotations of cultural and creative products and enhancing their immersive experience.

[0049] In one feasible implementation, based on the initial narrative path data, visual narrative elements distributed around the digital model are dynamically rendered in the display interface, including: mapping the weights and node types of cultural nodes in the initial narrative path data to a preset visual template library, and determining the corresponding visual element type, basic style, and dynamic behavior parameters for each cultural node; determining the relative positional relationship of each visual element in a three-dimensional or two-dimensional display space based on the correlation between cultural nodes in the initial narrative path data; and rendering the cultural node as the visual narrative element with corresponding style, position, and dynamic behavior in the display interface based on the visual element type, basic style, dynamic behavior parameters, and relative positional relationship of each visual element determined for each cultural node.

[0050] In this embodiment, based on the weights and types of cultural nodes in the initial narrative path data, a preset visual template library is mapped to determine the corresponding visual element type, basic style, and dynamic behavior parameters for each cultural node. For each cultural node in the initial narrative path data, the system selects or generates a corresponding visualization from the preset visual template library based on its calculated initial association weight and predefined node type (e.g., historical events, figures, geographical locations, artifacts, art styles, etc.). This visual template library can contain various visual element types, such as text labels, images, video clips, 3D models, particle effects, or abstract graphics. Simultaneously, the basic styles of these visual elements are determined, such as color, size, transparency, font, and texture, as well as their dynamic behavior parameters, such as appearance animations, disappearance animations, looping animations, interactive response animations (e.g., click to zoom, hover highlight), or specific motion trajectories. This mapping mechanism ensures that cultural nodes of different importance or types can be presented with differentiated visual effects, thereby intuitively conveying their information.

[0051] Based on the relationships between cultural nodes in the initial narrative path data, the system determines the relative positions of each visual element in the three-dimensional or two-dimensional display space. After determining the visual representation of each cultural node, the system further analyzes the relationships between cultural nodes in the initial narrative path data. These relationships may include temporal sequence, spatial proximity, logical causality, and conceptual inclusion. Based on these relationships, the system calculates and determines the relative positions of each visual narrative element in the three-dimensional or two-dimensional display space. For example, closely related nodes may be placed close to each other, forming visual clusters; nodes with hierarchical relationships may be arranged in a parent-child structure; and nodes with a time sequence may be arranged linearly along a timeline. This spatial layout not only enhances visual organization but also intuitively reveals the intrinsic connections between cultural nodes, helping users understand the narrative structure.

[0052] Based on the visual element type, basic style, dynamic behavior parameters, and relative positional relationships of each visual element determined for each cultural node, the cultural node is rendered as a visual narrative element with corresponding style, position, and dynamic behavior in the display interface. Finally, the system comprehensively utilizes the visual element type, basic style, dynamic behavior parameters determined for each cultural node, as well as the calculated relative positional relationships of each visual element in the display space, to render the data in the display interface. The rendering process transforms abstract cultural node data into visual narrative elements visible to the user, possessing specific visual styles, precise spatial positions, and preset dynamic behaviors. These elements are distributed around the digital model of the target cultural and creative product, collectively constructing a layered and interactive narrative scene, enabling users to intuitively perceive the cultural background and story of the cultural and creative product.

[0053] Through the aforementioned technical solution, this application can systematically transform abstract initial narrative path data into intuitive and expressive visual narrative elements. By mapping cultural nodes to a pre-defined visual template library based on their weight and type, it ensures that cultural nodes of different importance or attributes are presented with differentiated visual styles and dynamic behaviors, thereby highlighting core information and guiding users to focus on key points. Simultaneously, by determining the relative positions of each visual element based on the relationships between cultural nodes, the narrative elements form a meaningful spatial layout, intuitively revealing the intrinsic connections and narrative logic between cultural nodes. This refined rendering mechanism not only enhances the expressiveness and information delivery efficiency of the visual narrative elements but also greatly strengthens users' understanding of the cultural connotations of cultural and creative products and enhances their immersive experience, effectively solving the technical problem of how to transform abstract narrative data into intuitive and visual presentations.

[0054] In one feasible implementation, the initial narrative path data is dynamically adjusted based on the interactive operation data and the currently displayed narrative state to generate updated narrative path data. This includes: when the interactive operation data is a user's trigger operation on a specific visual narrative element in the display interface, the cultural node corresponding to that specific visual narrative element is identified as the currently active node; from the pre-trained multi-dimensional cultural knowledge graph database, adjacent cultural nodes directly connected to the currently active node are retrieved; the real-time correlation between the adjacent cultural node and the initial user cognitive vector is calculated, and the content difference between the adjacent cultural node and the currently displayed narrative node set is calculated; the currently displayed narrative node set is defined by the initial narrative path data and dynamically updated according to historical adjustments; a dynamic recommendation weight is calculated for each adjacent cultural node based on the weighted sum of the real-time correlation and the content difference; and the corresponding adjacent cultural node is inserted or moved up in the position of the initial narrative path data according to the dynamic recommendation weight to generate the updated narrative path data.

[0055] In this embodiment, when the interactive operation data is a user's trigger operation on a specific visual narrative element in the display interface, the system identifies the cultural node corresponding to that specific visual narrative element as the currently active node. This step aims to accurately capture the user's current focus of interest. The user's trigger operation can take various forms, including but not limited to clicking, hovering, voice commands, and gesture recognition. By listening to these interactive events and reversing the analysis based on the visual narrative element associated with the event, the system deciphers the corresponding cultural node in the multi-dimensional cultural knowledge graph database, thereby determining the core narrative that the user is currently actively exploring or focusing on.

[0056] The system retrieves adjacent cultural nodes directly connected to the currently active node from a pre-trained multi-dimensional cultural knowledge graph database. This step aims to expand outwards from the user's current point of interest to explore related cultural information. The multi-dimensional cultural knowledge graph database stores a wealth of cultural nodes and their relationships, which can be chronological, spatially proximate, involve interactions between characters, cause and effect between events, or thematic relevance. Through the graph database's query mechanism, the system can efficiently identify all cultural nodes directly connected to the currently active node via an edge, providing a candidate set for subsequent narrative path adjustments.

[0057] The system calculates the real-time relevance between adjacent cultural nodes and the initial user cognitive vector. This calculation aims to assess the degree of alignment between these candidate adjacent cultural nodes and the user's overall, relatively stable interests and preferences. Real-time relevance can be obtained by calculating the similarity between the semantic representation of the adjacent cultural node (e.g., a vector obtained through word embedding or knowledge graph embedding) and the initial user cognitive vector. Commonly used similarity metrics include cosine similarity and the reciprocal of Euclidean distance. A higher relevance indicates that the adjacent node aligns more closely with the user's deeper interests. Simultaneously, the system also calculates the content difference between adjacent cultural nodes and the currently displayed set of narrative nodes. The currently displayed set of narrative nodes is defined by the initial narrative path data and dynamically updated based on historical adjustments. The calculation of content difference aims to measure the novelty or complementarity of adjacent nodes relative to the narrative content the user has already encountered, thereby avoiding repetition and redundancy in narrative content. For example, a pre-trained text feature extraction model can be used to extract the text feature vectors of adjacent cultural nodes and each node in the currently displayed set of narrative nodes. Then, the mean cosine distance between the text feature vectors of the adjacent cultural nodes and the text feature vectors of all nodes in the currently displayed set of narrative nodes can be calculated, and this mean cosine distance can be used as a measure of content dissimilarity. The greater the dissimilarity, the more new information or perspectives the adjacent node can bring to the user.

[0058] Based on this, the system calculates a dynamic recommendation weight for each adjacent cultural node according to a weighted sum of real-time relevance and content difference. The dynamic recommendation weight is a quantitative indicator that comprehensively considers user interest matching and content novelty. By assigning different weight coefficients to real-time relevance and content difference and then weighting and summing them, the system can flexibly balance the deepening (closer to user interests) and expansion (introducing novel content) of the narrative path. For example, when users tend to explore a particular topic in depth, the weight of real-time relevance can be appropriately increased; when users show a tendency to explore new areas, the weight of content difference can be appropriately increased.

[0059] Based on dynamic recommendation weights, the system inserts or elevates the positions of corresponding adjacent cultural nodes in the initial narrative path data to generate updated narrative path data. The system sorts adjacent cultural nodes according to their dynamic recommendation weights. Nodes with higher weights are given priority and can be inserted after the currently active node to form new narrative branches, or their priority in the original narrative path is increased, giving them higher visual prominence or a higher display order in the display interface. In this way, the narrative path can be adjusted in real time to better adapt to changes in the user's current interaction behavior and interests.

[0060] Through the aforementioned technical solution, this application effectively addresses the potential lack of adaptability of initial narrative path data during actual user interaction. By capturing user interactions in real time and converting them into activations of specific cultural nodes, the system can accurately perceive the user's current focus of interest. Based on this, and combining the real-time correlation between adjacent cultural nodes and the user's initial cognitive vector, as well as the difference from already displayed content, the system can intelligently recommend cultural nodes that align with the user's deep interests and possess novelty. This dynamic recommendation mechanism, through weighted calculation of dynamic recommendation weights and adjustments to the narrative path accordingly, ensures that the narrative content of cultural and creative product displays consistently maintains a high degree of alignment with the user's real-time interests, avoiding rigidity and repetition in the narrative content. This not only enhances the user's immersive and personalized experience when exploring the cultural connotations of cultural and creative products but also makes the transmission of cultural knowledge more efficient and engaging, transforming static cultural displays into a dynamic, adaptive, and personalized exploration journey.

[0061] In one feasible implementation, calculating the content difference between the adjacent cultural node and the currently displayed set of narrative nodes includes: using a pre-trained text feature extraction model to extract the text feature vectors of the adjacent cultural node and each node in the currently displayed set of narrative nodes; calculating the average cosine distance between the text feature vector of the adjacent cultural node and the text feature vectors of all nodes in the currently displayed set of narrative nodes; and using the average cosine distance as a measure of the content difference.

[0062] In this embodiment, the pre-trained text feature extraction model is a model trained on a large amount of text data, capable of converting text content into fixed-dimensional numerical vectors. These models learn the distribution patterns of words or sentences in context, mapping discrete text information to a continuous vector space, making semantically similar texts closer together in the vector space. This model can be a deep learning-based word embedding model (such as Word2Vec, GloVe, FastText), or a more complex sentence or paragraph embedding model (such as BERT, RoBERTa, etc.). Its role is to convert the textual description of cultural nodes (e.g., their name, introduction, related stories, historical background, etc.) into a computable numerical form, laying the foundation for subsequent similarity calculations. The text feature vector is a numerical sequence representing the semantic information of the text, output by the pre-trained text feature extraction model. For a cultural node, its text feature vector can be a single vector obtained by encoding all its related text information, or an aggregation of multiple text fragment vectors, serving as a numerical representation of the cultural node's content.

[0063] In this embodiment, after obtaining the text feature vectors of adjacent cultural nodes and each node in the currently displayed narrative node set, the content difference is quantified by calculating the average cosine distance between the text feature vectors of the adjacent cultural nodes and the text feature vectors of all nodes in the currently displayed narrative node set. Cosine distance is a measure of the cosine of the angle between two non-zero vectors, used to assess their directional similarity; the closer the cosine value is to 1, the more consistent the directions, and the higher the similarity. Here, the average cosine distance refers to the average of the cosine distances between the text feature vectors of adjacent cultural nodes and the text feature vectors of each node in the currently displayed narrative node set. Using this average cosine distance as a measure of content difference means that the lower the average cosine similarity, the higher the content difference, thus effectively identifying nodes with high novelty to the current narrative content.

[0064] This application utilizes a pre-trained text feature extraction model to transform cultural node content into semantically rich text feature vectors. Furthermore, by calculating the mean cosine distance between the text feature vectors of adjacent cultural nodes and the text feature vectors of all nodes in the currently displayed narrative node set, this application can accurately quantify the semantic differences between adjacent cultural nodes and the current narrative content. This semantic vector-based difference calculation method avoids errors that may arise from keyword matching alone, ensuring the accuracy of content difference. Therefore, when dynamically adjusting the narrative path, the system can effectively identify and recommend cultural nodes with appropriate novelty to the user's currently known content, avoiding repetition and redundancy in narrative content. This enhances the coherence, richness, and exploratory nature of the user experience, making the intelligent display of cultural and creative products more attractive.

[0065] In one feasible implementation, the method further includes: continuously calculating a narrative field complexity index during the display process; the narrative field complexity index is calculated based on the number of currently active nodes in the updated narrative path data, the number of associations between the active nodes, and the average similarity between the active nodes and the initial user cognitive vector; when the narrative field complexity index exceeds a preset upper limit threshold, a path convergence operation is triggered; the path convergence operation includes: increasing the weight of nodes directly associated with the currently active node in the updated narrative path data; decreasing the weight of nodes indirectly associated with the currently active node and whose dynamic recommendation weight is lower than a third preset threshold, and regenerating the... The updated narrative path data; when the narrative field complexity index is lower than a preset lower threshold, a path expansion operation is triggered; the path expansion operation includes: retrieving from the pre-trained multi-dimensional cultural knowledge graph database nodes that have multi-hop associations with the currently active node and whose similarity to the initial user cognition vector is greater than a fourth preset threshold, and adding them as potential expansion nodes to subsequent calculations; calculating the real-time correlation degree between each potential expansion node and the initial user cognition vector, and its content difference degree with the currently displayed set of narrative nodes; calculating its dynamic recommendation weight based on the weighted sum of the real-time correlation degree and the content difference degree, and updating the narrative path data according to the dynamic recommendation weight.

[0066] In this embodiment, the narrative field complexity index is calculated based on the number of currently active nodes in the updated narrative path data, the number of relationships between active nodes, and the average similarity between the active nodes and the initial user cognitive vector. The narrative field complexity index aims to quantify the richness of the narrative content in the current display interface and the difficulty of user comprehension. For example, it can be calculated using the following formula: "Complexity Index = w1 × (Number of Active Nodes) + w2 × (Number of Relationships Between Active Nodes) - w3 × (Average Similarity Between Active Nodes and the Initial User Cognitive Vector)", where w1, w2, and w3 are adjustable weight coefficients. The number of active nodes can be directly counted by determining the number of cultural nodes corresponding to the visualized narrative elements in the current display interface. The number of relationships between active nodes can be obtained by analyzing the number of connection edges between these active nodes in a multi-dimensional cultural knowledge graph database. The average similarity between active nodes and the initial user cognitive vector can be obtained by calculating the similarity (e.g., cosine similarity) between each active node and the initial user cognitive vector and taking their average.

[0067] When the narrative complexity index exceeds a preset upper threshold, the system will trigger a path convergence operation. The path convergence operation includes: increasing the weight of nodes directly associated with the currently active node in the updated narrative path data; decreasing the weight of nodes indirectly associated with the currently active node whose dynamic recommendation weight is below a third preset threshold; and regenerating the updated narrative path data accordingly. For example, the weight of directly associated nodes can be increased by a fixed value or a proportional factor, giving them higher priority in the narrative path, thus achieving higher visual salience or a higher priority display order in the display interface. Nodes indirectly associated with a dynamic recommendation weight below the third preset threshold can have their weight reduced, or even be temporarily removed from the current narrative path, to reduce unnecessary interference information. The third preset threshold can be set according to the actual application scenario and user experience requirements; for example, it can be set to 0.2 or 0.3. Nodes below this value are considered to have a lower relevance to user interests.

[0068] When the narrative complexity index falls below a preset lower threshold, the system triggers a path expansion operation. The path expansion operation includes: retrieving nodes from the pre-trained multi-dimensional cultural knowledge graph database that have multi-hop associations with the currently active node and whose similarity to the initial user cognition vector is greater than a fourth preset threshold, and adding them as potential expansion nodes for subsequent calculations; calculating the real-time association degree between each potential expansion node and the initial user cognition vector, and its content difference degree with the currently displayed set of narrative nodes; calculating its dynamic recommendation weight based on the weighted sum of the real-time association degree and the content difference degree, and updating the narrative path data according to the dynamic recommendation weight. Multi-hop association refers to an association relationship connected through one or more intermediate nodes; for example, a two-hop association indicates a connection through one intermediate node. The fourth preset threshold is used to filter out potential expansion nodes that are somewhat related to user interests, avoiding the introduction of completely irrelevant cultural content. The calculation method for the real-time association degree and content difference degree of potential expansion nodes is similar to the calculation method for adjacent cultural nodes, ensuring that the expanded content is both in line with user interests and novel.

[0069] Through the aforementioned technical solution, this application can effectively manage the complexity of narrative content during the intelligent display of cultural and creative products. By continuously calculating the narrative field complexity index, the system can perceive the richness of the current narrative state and the potential burden on user comprehension in real time. When the narrative content is too complex, path convergence can focus on the core narrative clues, reduce redundant information, help users better understand the cultural connotations of the current cultural and creative products, and avoid confusion caused by information overload. Conversely, when the narrative content is too simple or monotonous, path expansion can proactively introduce novel cultural nodes related to user interests, enrich the narrative content, stimulate users' desire to explore, and thus enhance users' immersion and satisfaction during the intelligent display of cultural and creative products.

[0070] In one feasible implementation, the method further includes: recording memory data during the interaction process based on a preset instantaneous narrative memory; the memory data includes a sequence of cultural nodes identified as the currently active node, a timestamp of each node's activation, and the type of interaction operation that triggered the node's activation; analyzing the user's interaction pattern based on the memory data recorded in the instantaneous narrative memory; based on the analysis results of the interaction pattern, using a pre-trained sequence prediction model, predicting a set of potential cultural nodes from the pre-trained multi-dimensional cultural knowledge graph database according to the currently active node and the sequence of cultural nodes recorded in the instantaneous narrative memory; and loading some or all of the associated data of the potential cultural node set into a cache from the pre-trained multi-dimensional cultural knowledge graph database before the user performs the next interaction operation.

[0071] In this embodiment, the instantaneous narrative memory is a temporary storage structure whose main function is to record a series of user interactions and system state changes within the display interface. By capturing the user's trajectory and preferences during the narrative exploration process, it provides a data foundation for subsequent prediction and optimization. This memory can be implemented as a circular buffer, queue, or stack structure to store a fixed number of recent interaction events. Each memory data entry typically contains a unique identifier for the cultural node identified as the currently active node, a timestamp recording the specific time the node was activated, and the type of interaction operation indicating how the user interacted with the node, such as click, hover, drag, or voice command. The size and storage granularity of the memory can be flexibly configured according to the actual application scenario and performance requirements. For example, it can be chosen to record only the activation events of core nodes or to record all types of interaction events.

[0072] Based on the recorded memory data, the system analyzes the user's interaction patterns using the memory data recorded in the instantaneous narrative memory. This analysis aims to reveal the user's behavioral habits, interest shift paths, and underlying intentions when exploring the cultural connotations of cultural and creative products. Specific analysis methods may include sequence pattern mining to identify frequently occurring cultural node sequences or interaction operation sequences; time interval analysis to determine the user's interest depth between different nodes; interaction type preference statistics to understand which interaction methods the user prefers; and state transition analysis, such as constructing a Markov chain model to analyze the probability of a user moving from one cultural node to another.

[0073] Furthermore, based on the analysis results of the interaction pattern, the system utilizes a pre-trained sequence prediction model to predict a set of potential cultural nodes from the pre-trained multi-dimensional cultural knowledge graph database, according to the currently activated node and the cultural node sequence recorded in the instantaneous narrative memory. The core of this step lies in predicting the cultural nodes that the user might be interested in or want to explore next, thereby achieving pre-loading and pre-recommendation of narrative content. The sequence prediction model used can be a recurrent neural network (RNN), a long short-term memory network (LSTM), or a Transformer, etc. These models are pre-trained on a large amount of user interaction history data to learn the transfer patterns between cultural nodes and user behavior patterns. The model's input typically includes the currently activated node and the cultural node sequence in the instantaneous narrative memory as historical context, and the output is a set containing multiple cultural nodes, each node accompanied by a predicted probability or score, representing the likelihood of the user visiting that node next. The model's pre-training process is completed offline, using large-scale user interaction log data (e.g., historical user browsing paths and click sequences in similar cultural and creative product displays) for training to learn the semantic relationships between cultural nodes and the temporal dependencies of user behavior.

[0074] Before the user performs the next interaction, the system loads some or all of the associated data of the potential cultural nodes from the pre-trained multi-dimensional cultural knowledge graph database into the cache. This aims to pre-load the predicted potential cultural nodes and their related data into the system cache before the user actually performs the next interaction, thereby reducing user waiting time and improving response speed and narrative fluency. The loaded associated data may include detailed text descriptions of cultural nodes, related multimedia resources such as images, audio, and video, connections with other cultural nodes, and style and layout information used to render visual narrative elements. Caching strategies can employ LRU (Least Recently Used) or LFU (Least Frequently Used) algorithms. Depending on the probability of the predicted nodes, system resources (memory, bandwidth), and user experience requirements, the system can choose to load data for all predicted nodes or only load data for the N nodes with the highest probabilities. Preloading typically occurs after the user completes the current interaction but before initiating the next interaction, utilizing idle time.

[0075] By introducing a transient narrative memory to record user interaction trajectories and analyzing these patterns, the system can proactively predict potential cultural nodes of interest to users before their next action using a pre-trained sequence prediction model. Subsequently, the associated data of these predicted nodes is pre-loaded into a cache. This proactive data preparation mechanism effectively solves the latency problems caused by real-time computation and data loading in traditional dynamic adjustment methods. When a user actually triggers an interaction, the required data is already ready in the cache, significantly shortening response time and making the switching and presentation of narrative content smoother and seamless. This not only enhances the coherence and immersion of the user experience but also makes the entire intelligent display system for cultural and creative products more intelligent and predictive, better guiding users to explore the cultural connotations of these products.

[0076] In the embodiments of this application, the intelligent display method for cultural and creative products can achieve personalized display by acquiring user interest data, generating cognitive vectors, retrieving cultural nodes, matching similarity, generating narrative paths, dynamically rendering elements, and adjusting display content in real time. It can dynamically adjust the display content according to user interests, thereby improving user participation and the depth of cultural dissemination.

[0077] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the intelligent display method of cultural and creative products in this application. Any simple transformations based on this technical concept are within the protection scope of this application.

[0078] This application also provides an intelligent display system for cultural and creative products, for reference. Figure 2The intelligent display system for cultural and creative products includes: a memory 10, a processor 20, and an intelligent display program for cultural and creative products stored on the memory 10 and capable of running on the processor 20. The intelligent display program for cultural and creative products is configured to implement the steps of the intelligent display method for cultural and creative products.

[0079] The intelligent display system for cultural and creative products provided in this application, employing the intelligent display method for cultural and creative products in the above embodiments, can improve user participation and the depth of cultural dissemination. Compared with the prior art, the beneficial effects of the intelligent display system for cultural and creative products provided in this application are the same as those of the intelligent display method for cultural and creative products provided in the above embodiments, and other technical features of the intelligent display system for cultural and creative products are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0080] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0081] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. All equivalent structural transformations made under the technical concept of this application using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included within the scope of patent protection of this application.

Claims

1. A method for intelligent display of cultural and creative products, characterized in that, The method includes: Acquire the user's explicit interest data and implicit interest data, and perform weighted fusion processing on the explicit interest data and implicit interest data to generate an initial user cognitive vector; Based on the target cultural and creative product, cultural nodes and relationships associated with the target cultural and creative product are retrieved from a pre-trained multi-dimensional cultural knowledge graph database to obtain initial association network data. The initial user cognition vector is matched with each cultural node in the initial association network data to calculate vector similarity. Based on the matching results, each cultural node in the initial association network data is assigned an initial association weight. The cultural nodes are then sorted and filtered based on the initial association weight to generate initial narrative path data. A digital model of the target cultural and creative product is loaded into the display interface, and based on the initial narrative path data, visual narrative elements distributed around the digital model are dynamically rendered and generated in the display interface; the spatial layout and visual salience of the visual narrative elements are determined by the initial association weight of their corresponding cultural nodes. The system captures user interaction data in the display interface in real time, and dynamically adjusts the initial narrative path data based on the interaction data and the current narrative state to generate updated narrative path data. Based on this, the system updates the visual narrative elements and their content in the display interface in real time.

2. The intelligent display method for cultural and creative products as described in claim 1, characterized in that, The steps of acquiring users' explicit and implicit interest data, and performing weighted fusion processing on the explicit and implicit interest data to generate an initial user cognitive vector include: The system presents users with preset cultural level options for selection, records the user's selected options, and parses the option information into explicit interest vector data. When a user views a preview image of the target cultural and creative product for the first time, facial image sequence data of the user during this period is collected by an image acquisition device. The facial image sequence data is then used for fixed-point region identification and change analysis to obtain the user's visual attention distribution data and emotional characteristic data, which are then used as latent interest data. The explicit interest data and the implicit interest data are weighted and fused to generate the initial user cognition vector.

3. The intelligent display method for cultural and creative products as described in claim 2, characterized in that, The facial image sequence data is subjected to pinpoint region identification and change analysis to obtain the user's visual attention distribution data and emotional feature data, including: The user's eye region image is identified frame by frame from the facial image sequence data; Based on a pre-trained image analysis model, the eye region image is analyzed to obtain pupil size change sequence data. Based on the correspondence between the pupil size change sequence data and different regions in the preview image, the relative attention intensity of the user to different regions in the preview image is determined to obtain the visual attention distribution data. The system identifies key facial feature points of the user from the facial image sequence data, matches the displacement changes of the key facial feature points with a pre-trained micro-expression feature library to determine whether the user exhibits a preset type of micro-expression, and generates the emotion feature data based on the determination result.

4. The intelligent display method for cultural and creative products as described in claim 1, characterized in that, The process of sorting and filtering cultural nodes based on the initial association weights to generate initial narrative path data includes: Cultural nodes in the initial association network data whose initial association weight is greater than a first preset threshold are marked as core narrative nodes; Cultural nodes in the initial association network data whose initial association weight is less than the first preset threshold but greater than the second preset threshold are marked as associated narrative nodes; Based on the connection relationship between the core narrative node and the associated narrative node in the pre-trained multi-dimensional cultural knowledge graph database, a tree-like or graph-like structure data is constructed with the core narrative node as the main line and the associated narrative nodes as branches, as the initial narrative path data.

5. The intelligent display method for cultural and creative products as described in claim 1, characterized in that, Based on the initial narrative path data, visual narrative elements distributed around the digital model are dynamically rendered and generated in the display interface, including: Based on the weights and types of cultural nodes in the initial narrative path data, they are mapped to a preset visual template library to determine the corresponding visual element type, basic style, and dynamic behavior parameters for each cultural node. Based on the relationships between cultural nodes in the initial narrative path data, the relative positional relationships of each visual element in the three-dimensional or two-dimensional display space are determined; Based on the visual element type, basic style, dynamic behavior parameters, and relative positional relationships of each visual element determined for each cultural node, the cultural node is rendered as a visual narrative element with corresponding style, position, and dynamic behavior in the display interface.

6. The intelligent display method for cultural and creative products as described in claim 1, characterized in that, Based on the interactive operation data and the currently displayed narrative state, the initial narrative path data is dynamically adjusted to generate updated narrative path data, including: When the interactive operation data is a user's trigger operation on a specific visual narrative element in the display interface, the cultural node corresponding to that specific visual narrative element is identified as the currently active node. From the pre-trained multi-dimensional cultural knowledge graph database, retrieve the neighboring cultural nodes that are directly connected to the currently activated node; Calculate the real-time correlation between the adjacent cultural nodes and the initial user cognition vector, and calculate the content difference between the adjacent cultural nodes and the currently displayed narrative node set; the currently displayed narrative node set is defined by the initial narrative path data and dynamically updated according to historical adjustments; Based on the weighted sum of the real-time relevance and the content difference, a dynamic recommendation weight is calculated for each adjacent cultural node; Based on the dynamic recommendation weights, the corresponding adjacent cultural nodes are inserted or promoted in the position of the initial narrative path data to generate the updated narrative path data.

7. The intelligent display method for cultural and creative products as described in claim 6, characterized in that, Calculate the content difference between the adjacent cultural nodes and the currently displayed set of narrative nodes, including: Using a pre-trained text feature extraction model, extract the text feature vectors of the adjacent cultural nodes and each node in the currently displayed set of narrative nodes; Calculate the average cosine distance between the text feature vector of the adjacent cultural node and the text feature vectors of all nodes in the currently displayed narrative node set, and use the average cosine distance as a measure of the content difference.

8. The intelligent display method for cultural and creative products as described in claim 6, characterized in that, The method further includes: During the demonstration, the narrative field complexity index is continuously calculated; the calculation of the narrative field complexity index is based on the number of currently active nodes in the updated narrative path data, the number of relationships between the active nodes, and the average similarity between the active nodes and the initial user cognition vector; When the narrative field complexity index exceeds a preset upper limit threshold, a path convergence operation is triggered; the path convergence operation includes: Increase the weight of nodes directly associated with the currently active node in the updated narrative path data; Reduce the weight of nodes that are indirectly associated with the currently active node and whose dynamic recommendation weight is lower than a third preset threshold, and regenerate the updated narrative path data accordingly. When the narrative field complexity index is lower than a preset lower threshold, a path expansion operation is triggered; the path expansion operation includes: From the pre-trained multi-dimensional cultural knowledge graph database, retrieve nodes that have multi-hop associations with the currently activated node and whose similarity to the initial user cognition vector is greater than a fourth preset threshold, and add them as potential expansion nodes to subsequent calculations; For each potential expansion node, calculate its real-time correlation with the initial user cognition vector and its content difference with the currently displayed set of narrative nodes; The dynamic recommendation weight is calculated based on the weighted sum of the real-time relevance and the content difference, and the narrative path data is updated based on the dynamic recommendation weight.

9. The intelligent display method for cultural and creative products as described in claim 6, characterized in that, The method further includes: The memory data recorded during the interaction process is based on a preset instantaneous narrative memory; the memory data includes the sequence of cultural nodes identified as the currently active node, the timestamp of each node being activated, and the type of interaction operation that triggered the activation of the node; Based on the memory data recorded in the instantaneous narrative memory, the user's interaction patterns are analyzed; Based on the analysis results of the interaction mode, a pre-trained sequence prediction model is used to predict a set of potential cultural nodes from the pre-trained multi-dimensional cultural knowledge graph database, according to the currently activated node and the sequence of cultural nodes recorded in the instantaneous narrative memory. Before the user performs the next interaction, some or all of the associated data of the potential cultural node set is loaded into the cache from the pre-trained multi-dimensional cultural knowledge graph database.

10. A smart display system for cultural and creative products, characterized in that, The intelligent display system for cultural and creative products includes: a memory, a processor, and an intelligent display program for cultural and creative products stored in the memory and capable of running on the processor. The intelligent display program for cultural and creative products is configured to implement the steps of the intelligent display method for cultural and creative products as described in any one of claims 1 to 9.

Images