A multimedia interactive display method and system for an exhibition hall

Abstract

The application relates to the technical field of exhibition hall display, and discloses a multimedia interactive display method and system for an exhibition hall. The system comprises an interactive data acquisition module, a display partition module, a resource mapping module, a content combination module, a dynamic updating module and a display management module. The interactive data acquisition module acquires behavior data and preference data of visitors in the exhibition hall; the display partition module partitions display content according to themes and types according to the interactive information; the resource mapping module establishes a corresponding relationship between the display partition and a multimedia resource library; the content combination module extracts key resources and calculates resource correlation degrees to generate display combinations; the dynamic updating module updates the above mapping relationship according to real-time interactive data of the visitors; and the display management module determines a display sequence and a target according to the updated mapping relationship and generates a display plan. The system can improve the adaptability of display content to visitor demands and is suitable for display scenes of various exhibition halls.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of exhibition hall display, in particular to a multimedia interactive display method and system for an exhibition hall. BACKGROUND

[0002] In the current operation of the exhibition hall, the display system as the core carrier of information transmission and tourist attraction, the adaptability between its operation mode and the demand of tourists gradually becomes the key factor affecting the experience effect of the exhibition hall. The traditional exhibition display system adopts a fixed content presentation mode, and the display partition is usually determined at the initial stage of the construction of the exhibition hall, and it is difficult to adjust according to the actual interactive situation of the tourists. In this mode, the corresponding relationship between the display content and the multimedia resources is relatively single, and a fixed resource combination is usually preset, which cannot be dynamically optimized according to the behavior characteristics and preference differences of different tourist groups.

[0003] From the perspective of tourist experience, the traditional system cannot accurately capture the interactive information of tourists in the exhibition hall, such as the behavior data of tourists, such as the length of stay of tourists on specific display content, operation frequency, and the preference data expressed by tourists through feedback devices. Due to the lack of effective use of these data, the display partition remains fixed, which may cause some tourists to have aesthetic fatigue on the repeated display content, while the content interested by other tourists cannot be fully presented.

[0004] In terms of resource utilization, the traditional system lacks a flexible mapping mechanism between the multimedia resource library and the display partition, and once the corresponding relationship between the resource and the partition is determined, it is difficult to modify. This makes it impossible to timely allocate a large number of high-quality multimedia resources according to the needs of tourists, resulting in resource waste and failing to fully play the display value of the resources. In addition, the traditional system does not have the ability to update the mapping relationship according to the real-time interactive data of tourists, and the display content is always in a static state, which cannot keep up with the dynamic changes of the needs of tourists, further reducing the attraction of the exhibition hall and the participation of tourists.

[0005] From the perspective of the operation efficiency of the exhibition hall, the traditional system needs to manually adjust the display partition and the resource combination by the staff, which not only consumes a lot of manpower and time cost, but also the adjustment effect often depends on the experience judgment of the staff, and cannot guarantee the accuracy and timeliness of the adjustment. With the continuous improvement of the requirements of tourists on the exhibition hall experience, the limitations of the traditional display system are increasingly prominent, and a new type of display system that can realize interactive data collection, dynamic partition adjustment and flexible resource mapping is needed to meet the needs of modern exhibition hall operation. SUMMARY

[0006] The purpose of the present application is to provide a multimedia interactive display method and system for an exhibition hall to solve the problems raised in the background.

[0007] To achieve the above object, the present application provides a multimedia interactive display system for an exhibition hall, which comprises:

[0008] An interactive data collection module for collecting interactive information of visitors in the exhibition hall, including behavior data and preference data;

[0009] A display partition module for partitioning display content according to themes and types based on the interactive information;

[0010] A resource mapping module for mapping the display partitions and a multimedia resource library to establish a corresponding relationship between the partitions and the resources;

[0011] A content combination module for extracting key resources from the mapping relationship and calculating the correlation between multiple resources to generate a display combination;

[0012] A dynamic updating module for updating the mapping relationship between the display partitions and the multimedia resource library based on real-time interactive data of visitors;

[0013] A display management module for determining a display order and a display target based on the updated mapping relationship to generate a display plan.

[0014] Preferably, the implementation of the interactive data collection module comprises:

[0015] For any interactive information of a visitor, a classification model corresponding to the interactive information is obtained;

[0016] The classification model is used to classify the interactive information to obtain at least one interactive category;

[0017] Key words and associated words in the interactive information corresponding to the interactive category are identified to form an interactive sample set;

[0018] The key words and associated words in the interactive sample set are analyzed to obtain a key word area and an associated word area as part of the interactive information.

[0019] Preferably, the implementation of obtaining the key word area and the associated word area of the interactive sample set further comprises:

[0020] The key words and associated words in the interactive sample set are merged according to the interactive category to obtain multiple merging results;

[0021] The key word pairs in the merging results are extracted and compared with a key word dictionary to obtain the key word area;

[0022] The association strength of the associated words in the merging results is extracted, and the merging results are divided according to the association strength to obtain the associated word area.

[0023] Preferably, the implementation of the display partition module further comprises:

[0024] The display partition is judged, the visiting frequency and the staying time of the visitors in the display partition are analyzed, the display partition is fitted according to the visiting frequency and the staying time, and the mapping relationship of the display partition to the visitor interest is constructed.

[0025] Preferably, the implementation of the resource mapping module comprises:

[0026] The multimedia resources corresponding to the display partition and the historical display records are called to generate a plurality of unlabeled resource identification results;

[0027] It is judged whether the plurality of unlabeled resource identification results is the target resource identification result, if yes, the target resource identification result is regarded as the multimedia resource library of the display partition.

[0028] Preferably, the implementation of establishing the mapping relationship between the display partition and the multimedia resource library comprises: the information of the keyword area and the associated word area existing in the display partition, the description information and the category of the multimedia resource library are used to represent, and the mapping relationship between the display partition and the multimedia resource library is constructed.

[0029] Preferably, the implementation of the content combination module comprises:

[0030] The display partition and the multimedia resource library are clustered and analyzed according to the content type, the resource format and the display function, and the maximum clustering center after the clustering analysis is set as the key resource;

[0031] The keywords of the key resource are extracted, the similarity between each keyword is calculated, and the public sequence related to the similarity between each keyword is set;

[0032] The resources existing in the public sequence are extracted by using the public sequence related to the similarity between each keyword, and the matching degree between each resource is set;

[0033] The matching degree between each resource is set according to the heat distribution of each resource.

[0034] Preferably, the implementation of the dynamic updating module comprises:

[0035] The heat distribution of each resource is extracted from the display combination, and the target path of the display combination is set according to the time period corresponding to the heat distribution of each resource;

[0036] The target path of each resource in the display combination is fitted to obtain the fitted target path, and the probability value of the fitted target path in each time period is set as the appearance probability of the display combination;

[0037] The occurrence probability of the display combination is compared with the reference in the multimedia resource library, a difference value is identified, and the resources in the multimedia resource library are classified according to the difference value, so that the updating of the multimedia resource library is completed.

[0038] Preferably, the display management module is implemented as:

[0039] According to the mapping relationship between the updated display partition and the multimedia resource library, the display target in the interactive information is extracted, the display target is sorted according to the occurrence probability of the display target, and the display order is obtained;

[0040] The display target and the display order are combined in a structured form to obtain a display plan.

[0041] Preferably, the present application also includes a multimedia interactive display method for an exhibition hall, which includes all the modules and method processes of the above-mentioned multimedia interactive display system for an exhibition hall.

[0042] Compared with the prior art, the present application has the following advantages:

[0043] The multimedia interactive display system for an exhibition hall can comprehensively obtain the interactive information of tourists in the exhibition hall, including behavior data and preference data, through the interactive data acquisition module, so that the system can accurately understand the needs and interest points of tourists, and avoid the problem that the display content is disconnected with the needs of tourists due to the lack of effective perception of the needs of tourists in the traditional system. Based on these collected interactive information, the display partition module can partition the display content according to the theme and type, so that the classification of the display content is more in line with the actual interests of tourists, and tourists can quickly find the display area of their interest, reduce the invalid browsing time in the exhibition hall, and improve the browsing efficiency and experience of tourists.

[0044] The resource mapping module maps the display partition and the multimedia resource library to establish the corresponding relationship between the partition and the resource, breaking the limitation that the corresponding relationship between the resource and the partition in the traditional system is fixed. The establishment of this mapping relationship enables the resources in the multimedia resource library to be reasonably distributed according to the needs of the display partition, avoids the waste of resources, and also enables each display partition to be equipped with high-quality resources that match it, enhancing the attractiveness of the display partition. The content combination module extracts key resources from the mapping relationship and calculates the correlation degree between multiple resources to generate a display combination, which can integrate resources with strong correlation for display, so that tourists can obtain more coherent and systematic information in the browsing process, deepen the understanding and memory of the display content, and instead of only seeing scattered and isolated resource display in the traditional system, improving the effectiveness of information transmission.

[0045] The dynamic updating module updates the mapping relationship between the display partitions and the multimedia resource library according to the real-time interaction data of the tourists, so that the system can respond to the changes in the tourist demand in a timely manner. When the interest points of the tourists change, the system can update the mapping relationship to allocate the corresponding multimedia resources to the corresponding display partitions, so as to ensure that the display content is always synchronized with the real-time demand of the tourists, and avoid the problem of aesthetic fatigue of the tourists caused by the static display content of the traditional system, and maintain the freshness and participation enthusiasm of the tourists for the display content of the exhibition hall.

[0046] The display management module determines the display order and the display target according to the updated mapping relationship, generates a display plan, and makes the entire display process more orderly and organized. The module can optimize the display order in combination with the real-time updated mapping relationship, ensure that the tourists can obtain information in a reasonable logical order during browsing, and clearly define the display target of each display partition, so that each display link can revolve around the target, and the pertinence and effectiveness of the display are improved. The operation of the entire system does not require a large amount of manual operation by the staff, reduces the investment in labor and time cost, improves the operation efficiency of the exhibition hall, enables the exhibition hall to more efficiently provide high-quality display services for the tourists, enhances the competitiveness of the exhibition hall in similar places, and better meets the demand of modern exhibition halls for personalized and intelligent display. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 a timing diagram of the multimedia interactive display system for an exhibition hall according to the present application;

[0048] Figure 2 a working principle diagram of an implementation mode of the interactive data acquisition module;

[0049] Figure 3 a working principle diagram of a supplementary implementation mode of the display partition module;

[0050] Figure 4 a working principle diagram of an implementation mode of the resource mapping module. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0052] Please refer to Figure 1 The present application provides a multimedia interactive display method and system for an exhibition hall, which comprises:

[0053] The interactive data collection module is responsible for obtaining the interactive information of the visitors in the exhibition hall, including behavior data and preference data. The display partition module partitions the display content according to themes and types according to the interactive information. The resource mapping module maps the display partition and the multimedia resource library to establish the corresponding relationship between the partition and the resource. The content combination module extracts key resources from the mapping relationship, calculates the correlation degree between multiple resources, and generates a display combination. The dynamic update module updates the mapping relationship between the display partition and the multimedia resource library according to the real-time interactive data of the visitors. The display management module determines the display order and the display target according to the updated mapping relationship, and generates a display plan. The entire system cooperates through modularization to realize dynamic adjustment and personalized display of the display content.

[0054] Embodiment 1: see Figure 2 When a visitor operates in front of an interactive touch screen displaying dinosaur fossils in the exhibition hall, his / her behavior data (such as clicking the "Tyrannosaurus" 3D model to rotate and view, and staying in front of the "Cretaceous ecosystem" video for a long time) and preference data (such as subsequently inputting "Jurassic large carnivorous dinosaur" in the search box) are recorded by the system to form an original, multi-modal interactive information record. For such an interactive information, the system will first obtain its corresponding classification model. These classification models are machine learning models pre-trained based on a large amount of exhibition theme data, which may include text classification models for identifying interest themes, behavior analysis models for judging behavior intentions, etc. The system will automatically select and call one or more corresponding models according to the data type of the interactive information (such as text input, touch behavior sequence, and stay time). In this example, the text classification model will process the search term "Jurassic large carnivorous dinosaur", and the behavior analysis model will analyze the click and stay behavior sequence. After these classification models process the current interactive information, one or more interactive category labels will be output. The text model may output theme categories such as "paleontology", "dinosaur", and "predator"; the behavior model may output behavior intention categories such as "deep exploration" and "content consumption". These categories provide a framework and direction for subsequent fine-grained analysis.

[0055] The system will perform further text and semantic analysis on the interaction information under each identified interaction category. For the text content, natural language processing techniques will be applied to identify the core keywords and associated words. For the behavior data, it will be converted into analyzable metadata, such as the set of associated tags for the clicked "Tyrannosaurus Rex" model object, or the list of title and keywords for the watched video. All these extracted words and tags from the raw information form an initial interaction sample set for the current interaction event. The sample set for a single interaction is sparse and noisy. Therefore, the implementation contains key merging and analysis steps. The system will merge the current interaction information generated sample set with the historical interaction sample sets from the same visitor or other visitors with similar category labels. This process is not a simple addition, but a merging and aggregation by interaction category. For example, the keywords and associated words from all interactions labeled with "Dinosaur" will be merged together. After merging, a larger vocabulary set is formed, which better reflects the group interest patterns.

[0056] The system will extract high-frequency and stable word pairs from the merged set and compare them with a pre-built keyword dictionary covering the knowledge domain of the exhibition. Those word pairs that highly match the dictionary entries are identified as "keyword zones" with clear directionality, representing the explicit and core concepts in visitor interest. The system will analyze the co-occurrence relationships and statistical correlation strength between words in the vocabulary set. For example, "Tyrannosaurus Rex" and "Rex" may always appear together with a very high correlation strength, and "Cretaceous" and "extinction event" may also show a strong statistical correlation. The system will divide these closely related word groups into different "associated word zones" based on a pre-set correlation strength threshold. Each associated word zone reveals some potential and associated theme clusters or concept networks in visitor interest.

[0057] The original, unstructured visitor interaction information is transformed into a structured data representation containing "keyword zones" and "associated word zones". These zone information, as part of the enriched interaction information, accurately depicts the current and potential interest focus and knowledge exploration path of the visitor. This structured data provides high-quality and computable input for the downstream display partitioning module and resource mapping module, enabling the system to truly understand visitor intent and lay a solid data foundation for subsequent content personalization and dynamic display. The entire process is automated and continuously running, ensuring real-time and accurate analysis and characterization of each visitor's interaction.

[0058] Embodiment 2: see Figure 3, which is responsible for partitioning the exhibit content based on the visitors' interaction information and further analyzing the visitors' behavior data in these partitions to construct the mapping relationship between the partitions and visitors' interests. The entire implementation process is based on continuous monitoring and analysis of the partition access frequency and dwell time, with the goal of dynamically adapting the setting of the exhibit partitions to the actual interest patterns of visitors. During system initialization, the physical or logical space of the exhibition hall has been divided into multiple exhibit partitions, each of which is organized around a core theme or content type. For example, in a large natural history museum, there may be "Paleontology Evolution Hall", "Modern Ecological Diversity Corridor", "Geology and Mineral Exhibition Area", and "Human and Environment Interaction Museum" as the main partitions. Each partition contains several exhibits, such as fossil specimens, ecological landscaping, interactive screens, and cultural relic displays. The division of these partitions is not only based on content themes, but also considers the rationality of space layout and the smoothness of visitor flow.

[0059] The interactive data collection module continuously collects the original behavior data of visitors in each partition, which is obtained through a variety of sensor networks deployed in the exhibition hall, including but not limited to: infrared or camera flow counters located at the entrances of partitions and in front of key exhibits, used to count the number of visitors entering the partition and staying in front of specific exhibits; touch sensors attached to interactive exhibits, recording the number of operations and duration; and positioning beacons worn on visitors or anonymous mobile trajectory data obtained through Wi-Fi probes, used to calculate the overall dwell time of visitors in the partition. All sensor data is indexed by timestamp and partition ID to form a complete visitor behavior log. These data streams are transmitted in real time to the data processing center, and after data cleaning and preprocessing, they are aggregated and stored according to the partition identifier.

[0060] For the calculation of "access frequency", the system uses a sliding time window mechanism to count the number of independent visitors accessing each partition in fixed time intervals (such as every hour, every day, or every week). The system uses a HyperLogLog-based cardinality estimation algorithm to accurately count the number of independent visitors while ensuring computational efficiency. For example, the system may find that "Paleontology Evolution Hall" has a total of 500 independent visitors entering during the afternoon on Saturday, while "Geology and Mineral Exhibition Area" has only 150 visitors accessing during the same period. The system not only records the total number, but also analyzes the pattern of its change over time, such as the low access volume in the afternoon on weekdays, but a significant increase in the influx of family visitors on weekends. The system will establish an access frequency time series model for each partition to predict future access trends.

[0061] For the calculation of "dwell time", the system employs an accurate analysis method based on trajectory data. By analyzing the positioning data sequence of visitors, the system uses a dwell point detection algorithm (such as a density-based clustering algorithm) to identify the actual dwell area and time of visitors within a subarea. When calculating the average duration from the first entry to the final exit, the system excludes abnormally brief stays (such as passing through). For example, analysis found that the average dwell time of visitors in the "Modern Ecological Diversity Corridor" was 25 minutes, while the average dwell time in the "Human-Environment Interaction Gallery" reached 40 minutes due to the presence of many interactive exhibits. The system also analyzes the distribution characteristics of dwell time and plots a dwell time distribution histogram to identify the main dwell time intervals and outliers.

[0062] The core operation of the subarea display module is the "fitting" process, which performs multidimensional comprehensive operations on the two types of numerical indicators mentioned above, access frequency and dwell time. The system uses a multi-index comprehensive evaluation method based on the entropy weight method to calculate a single "interest heat value" for each subarea. This method determines the weight of each index by calculating its entropy value, avoiding the bias of subjective weighting. In the calculation process, the system gives a higher weight to dwell time, as long-term stays often reflect deeper interest more than high-frequency brief visits. For example, the "Geological Mineral Exhibition Area" may not have a high access frequency, but a few mineral enthusiasts stay for a very long time, and its final calculated interest heat value may not be low. The system also considers the coefficient of variation of access frequency and dwell time to assess the stability and reliability of the data.

[0063] Based on the interest heat values calculated for all subareas, the system performs subarea-level judgments. It uses the K-means clustering algorithm to divide all subareas into different interest levels, such as "high-interest subareas", "medium-interest subareas", and "low-interest subareas". This process establishes a mapping relationship from physical subareas to abstract visitor interest. The system maintains a dynamically updated mapping table inside, using a Redis database to store each subarea ID and its current corresponding interest level label, and sets a reasonable expiration time to achieve automatic data updating. This mapping relationship is not static, and the system continuously monitors new interaction data streams, using a streaming computing framework (such as Apache Flink) to process sensor data in real time. The system recalculates the access frequency, dwell time, and interest heat value of all subareas regularly (e.g., every two hours). It uses the CUSUM (Cumulative Sum) control chart algorithm to detect changes in heat values, and once the heat value of a subarea changes beyond a preset threshold, the system automatically updates the interest level in the mapping table. For example, a temporary "Special Exhibition: Deep Sea Spectacle" may have high access frequency and dwell time at the beginning, and be mapped as a "high-interest subarea"; after a few weeks, as the novelty wears off, its data falls back, and the system will remap it as a "medium-interest subarea".

[0064] The module also supports more granular analysis, which employs spatial clustering algorithms (e.g. DBSCAN) to identify the heat difference of different exhibits within a zone. For example, in the "Hall of Ancient Life", the system might find through local sensors that the dinosaur skeleton exhibit area has a high crowd density and long dwell time, while the early trilobite fossil exhibit has a fast flow rate. These micro-level data within a zone are visualized through heat maps, providing data support for the internal layout optimization of the zone. The system also establishes association rules between exhibits, analyzing the transition probability of visitors between different exhibits, and optimizing the layout of exhibits and guide routes.

[0065] This dynamically constructed mapping relationship becomes the key input of the exhibition hall content management system, which enables the management system to identify the current most attractive content theme to visitors, thereby providing decision-making basis for resource allocation, guide suggestion, and subsequent content combination. The system generates zone heat reports based on the mapping relationship, including the real-time heat ranking of each zone, heat change trend, and predicted future heat trend. These reports are provided to other system modules through RESTful API. The entire implementation embodies a closed-loop process from data collection to semantic mapping, enabling static display zones to dynamically reflect and adapt to the changing interests of the flowing visitor groups.

[0066] Example 3: Refer to Figure 4 , a dynamic and semantic correspondence is established and maintained between the display zones and the multimedia resource library. Its implementation is a multi-step reasoning and matching process, starting with extensive calling and preliminary identification of zone-related resources. During system initialization, the resource mapping module receives the processing results from the display zone module, which is the semantic representation of each display zone, including the keyword area and the associated word area generated by the interactive data collection module. For example, for the "Hall of Ancient Life" zone, its keyword area may contain "dinosaur", "fossil", "extinction", and its associated word area may contain "excavation", "stratum", "climate change". These semantic information is stored in the form of vectorization, using pre-trained models such as Word2Vec or BERT to convert text into high-dimensional vector representation. At the same time, the system accesses a centralized multimedia resource library, which is built based on a distributed file system and stores all available display resources. Each resource is accompanied by rich metadata and records of its history of being displayed. All metadata are stored in Elasticsearch to support efficient full-text search and complex queries.

[0067] The first step of the module is to call multimedia resources related to the current target showcase partition and their history, which is based on preliminary semantic screening. The system uses the keyword area of the partition and resource metadata for rough matching based on vector similarity, calculates the similarity of the partition vector and the resource description vector using cosine similarity, and sets a lower threshold to generate an initial resource candidate set. Each resource in this set is considered as an "unlabeled resource identification result". The next step is a critical judgment step. The system needs to filter out the most relevant and high-quality resources from the batch of unlabeled resource identification results. This judgment process uses a multi-factor comprehensive evaluation model to build an evaluation function to calculate the comprehensive matching degree of each candidate resource and the target partition. The evaluation function considers three main dimensions, and its calculation formula is as follows:

[0068] ;

[0069] Wherein: represents the resource and the comprehensive matching degree value of the target partition. represents the text semantic similarity score, which is calculated by the fine-tuned BERT model to calculate the semantic correlation between the partition keywords and the resource description. represents the time freshness score, which is calculated based on the latest use time of the resource, using an exponential decay function , wherein represents the time difference between the current time and the latest use time, is the decay coefficient. represents the popularity score, which is calculated by weighting the min-max standardized historical use frequency, user stay time, interaction times and user rating, etc. , , is the weight coefficient, which satisfies , which can be dynamically adjusted through the management interface.

[0070] The system calculates the comprehensive matching degree value of all candidate resources, and uses an adaptive threshold setting method to dynamically set the threshold according to the overall quality distribution of the current resource library. If the matching degree value of any resource exceeds this threshold, it is determined as the target resource identification result. The system also considers the diversity of resources to avoid selecting too many similar resources. By calculating the similarity matrix between resources, it ensures that the final selected resource set is both relevant and diverse. The last step in establishing the mapping relationship is to persistently associate these target resources with the display partitions. The implementation is based on vector space nearest neighbor search, using the vectorized representation of the partition's keyword area and associated word area as the query vector, and the vectorized representation of the target resource's description information and category label as the database vector. Through the approximate nearest neighbor search library, the proximity of the two in the vector space is efficiently calculated. The system records the matching strength of each resource with the partition, which is the normalized result of the comprehensive matching degree value. This mapping relationship is usually stored in the form of a graph database, establishing a "partition-resource" relationship graph, where the edge weight represents the matching strength.

[0071] In Example 4, a display partition and its mapped multimedia resource library are described. Taking a display partition named "Ocean Ecological Mystery" as an example, its mapped resource library may include the following resources: a 4K high-definition documentary video about coral reefs (resource ID: V_001), an interactive marine food chain simulation software (resource ID: S_002), a set of high-definition photos of rare deep-sea organisms (resource ID: P_003), an expert interpretation article on ocean acidification (resource ID: T_004), and an immersive experience program that allows visitors to virtually drive a submarine to explore underwater mountain ranges (resource ID: I_005). The module first performs clustering analysis, which reads the metadata of all resources and automatically groups them based on three preset dimensions: content type (such as video, software, image, text, immersive experience), resource format (such as MP4, EXE, JPEG, PDF, VR), and display function (such as information transmission, interactive operation, visual appreciation, knowledge deepening, experience simulation). The analysis process uses unsupervised machine learning algorithms to calculate the distances of all resources in these dimension feature spaces, and groups resources with similar features into the same cluster. For example, the algorithm may group V_001 (video, MP4, information transmission) and P_003 (image set, JPEG, visual appreciation) into a category based on "visual media" features; at the same time, S_002 (software, EXE, interactive operation) and I_005 (program, VR, experience simulation) are grouped into another category based on "interactive experience" features; and T_004 (text, PDF, knowledge deepening) may temporarily form a separate group.

[0072] After clustering, the system calculates the center point of each cluster and sets the center point of the cluster with the most representative members or the most representative cluster as the key resource of the current analysis. Assuming that the "interactive experience" cluster contains S_002 and I_005, and the center point feature is closest to S_002, then the resource S_002 (ocean food chain simulation software) is established as a key resource of the current partition resource library. The module extracts the core keywords from the metadata of the key resource S_002, such as "food chain", "energy flow", "predator", and "plankton". The system also extracts keywords from all other resources, such as "coral reef", "symbiosis", and "biodiversity" from V_001, and "submarine", "exploration", and "terrain" from I_005. Then, the system calculates the semantic similarity between the keywords of the key resource and the keywords of each other resource. This calculation is based on a word vector model that evaluates the proximity of words in semantic space. Through analysis, the system may find that "food chain" has a high semantic association with "symbiosis" (from V_001) and some conceptual relevance with "exploration" (from I_005).

[0073] Based on these pairwise similarity relationships, the system aims to find a common sequence that maximally covers these highly associated keywords. This sequence is essentially a content theme or narrative thread, such as "ocean life interaction and energy exploration". This common sequence serves as a filter through which the system extracts resources from the resource library whose keywords have the highest matching degree with this sequence. Then, the system calculates the matching degree between these extracted resources to evaluate their logical connection and thematic consistency in content with each other. For example, the matching degree between V_001 (coral reef ecosystem) and S_002 (food chain simulation) in content will be high, while their matching degree with I_005 (submarine exploration) may be moderate. The module needs to sort and combine these resources into an attractive display scheme. The system introduces historical heat distribution data of each resource, such as the average user engagement, dwell time, or interaction frequency in the past week. The display combination settings will preferentially arrange resources with high matching degree and high heat next to each other to form a smooth experience flow; at the same time, it will also consider strategically placing resources with high heat but moderate matching degree in the sequence.

[0074] Table 1: Content combination analysis process.

[0075]

[0076] The operation of the dynamic updating module begins with a deep analysis of the current generated presentation combinations, from which the heat distribution data embedded in each resource is extracted from the output of the content combination module. Heat distribution is a multi-dimensional measure that not only contains the historical access frequency of a resource, but also incorporates patterns in the time dimension, such as the popularity intensity of a resource at a specific time of day (e.g. afternoon), a specific day of the week (e.g. weekend), or a specific seasonal exhibition period. The system uses time series analysis methods to decompose the heat data into trend components, seasonal components, and residual components, thereby more accurately capturing the regularity of heat changes. These heat distribution data are associated with precise time periods, outlining an expected "target path" for each presentation combination. This path depicts the expected trajectory of the combination in the future time periods under ideal conditions, in which it is called and presented by the system.

[0077] The next step is to integrate and correct these expected paths. The system does not consider the target path of a single presentation combination in isolation, but rather synthesizes the target paths of all currently active presentation combinations. The fitting process uses a Kalman filter-based data fusion algorithm that can effectively handle noise and uncertainty between multiple prediction paths, extracting the most likely future presentation trend of the system as a whole from a large number of individual predictions. Through fitting, a comprehensive target path is obtained that represents the overall intention of the system. For this fitted path, the system uses Monte Carlo simulation methods to assign a probability value to each future time period (such as the next one-hour window), which is called the "appearance probability of the presentation combination". This probability value quantifies the likelihood of a specific presentation combination being actually triggered in that time period.

[0078] The dynamic updating module initiates a comparison and verification process, which compares the calculated appearance probabilities of each presentation combination with the actual reference situation recorded in the multimedia resource library. The actual reference situation includes objective indicators such as the recent actual number of calls, the actual position in the presentation sequence, and the completion rate of being completely browsed by tourists. The comparison process uses statistical hypothesis testing methods to calculate the standardized residual between the observed value and the expected value, producing a quantitative "difference value" that reveals the degree of deviation and its statistical significance between the system's prediction and the actual situation. For example, a presentation combination with the theme of deep sea exploration, which the system predicts to have a high appearance probability in the evening period, but the actual reference data shows that it is rarely triggered in that period, resulting in a positive difference value. Conversely, a combination that is not sufficiently valued by the system may be active in reality, resulting in a negative difference value.

[0079] These difference values become the key signals to drive the update of the multimedia resource library. The system reclassifies and identifies the resources in the library according to the size and sign of the difference values. The system establishes a three-layer classification system based on machine learning: the first layer divides the resources into "overestimated resources" and "underestimated resources" according to the sign of the difference values; the second layer performs fine-grained classification according to the magnitude of the difference values; and the third layer dynamically adjusts in combination with the duration and trend of the difference values. For resources that continuously have large positive difference values, it means that the system overestimates their attractiveness, and their priority may be lowered or they may be given a lower weight in subsequent mapping. For resources that have large negative difference values, it means that their actual popularity exceeds the system's expectations, and these resources will be marked as "emerging hotspots", their priority will be raised, and they may be more frequently associated with related partitions in subsequent mapping relationships. In addition, the system also establishes a resource life cycle management mechanism, which triggers the introduction of new resources or the archiving of old resources according to the pattern of the difference values, thereby updating the multimedia resource library and the evaluation system.

[0080] The display management module starts running after the dynamic update module completes its work, and it receives the updated mapping relationship between the display partitions and the multimedia resource library, which is a dynamic view reflecting the latest tourist interests and resource attractiveness. The module first extracts the core "display targets" that need to be responded to from the real-time incoming interactive information stream using natural language processing techniques and behavior pattern recognition algorithms. These targets may come directly from the active queries of tourists, or may be derived from potential interest points inferred by behavior data analysis, or even include display requirements predicted based on spatio-temporal context.

[0081] The module sorts all identified display targets according to the appearance probability data provided by the dynamic update module. The sorting algorithm not only considers the level of appearance probability, but also takes into account the diversity, novelty and educational value of the display targets as multiple optimization objectives. Display targets with high appearance probability are considered to be more consistent with the overall tourist interest trends, and therefore are given higher priority and thus get a higher position in the display order. At the same time, the system also uses an exploration-exploitation strategy to reserve a certain proportion of opportunities for display targets with low appearance probability but potential value or novelty, in order to maintain the diversity and exploratory nature of the display content.

[0082] The module encapsulates the sorted sequence of display targets with their display order in a structured data format, generating a clear "display plan" that can be directly parsed and executed by the display terminal device. The plan is encoded in JSON-LD format, which not only specifies in detail which content resources should be presented in sequence at a specific time, facing a specific audience or situation, but also includes transition effect suggestions between resources, display duration guidance, and alternative information. After the plan is generated, it is distributed to each display terminal through the message queue, and the terminal device makes appropriate adaptive adjustments according to the local context (such as the current audience size, device status, etc.), realizing fine guidance of the visitor experience.

[0083] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0084] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, alternatives, and variations can be made in the embodiments without departing from the spirit and scope of the present application as defined by the appended claims and their equivalents.

Claims

1. A multimedia interactive display system for exhibition halls, characterized in that, include: The interactive data collection module is used to acquire interactive information of visitors within the exhibition hall, including behavioral data and preference data; The display partitioning module is used to partition the displayed content according to themes and types based on the interactive information. The resource mapping module is used to map the display partitions to the multimedia resource library, establishing a correspondence between the partitions and the resources; The content combination module is used to extract key resources from the mapping relationship, calculate the correlation between multiple resources, and generate a display combination; The dynamic update module is used to update the mapping relationship between the display area and the multimedia resource library based on the real-time interaction data of tourists; The display management module is used to determine the display order and display objectives based on the updated mapping relationship, and to generate a display plan; The resource mapping module is implemented in the following ways: The system retrieves the multimedia resources and historical display records corresponding to the display partition, generating multiple unlabeled resource identification results. Determine whether multiple unlabeled resource identification results are the target resource identification results. If so, treat the target resource identification results as the multimedia resource library of the display partition. The content combination module is implemented in the following ways: The display partitions and multimedia resource library are clustered according to content type, resource format, and display function. The largest cluster center after the clustering analysis is set as the key resource. Extract keywords from key resources, calculate the similarity between keywords, and set common sequences related to the similarity between keywords; By utilizing the common sequences related to the similarity between keywords, resources existing in the common sequences are extracted, and the matching degree between each resource is set; The matching degree between various resources is set according to the popularity distribution of each resource, and the display combination is set accordingly; The implementation methods of the dynamic update module include: Extract the popularity distribution of each resource from the display combination; set the target path of the display combination according to the time period corresponding to the popularity distribution of each resource; The target paths of each resource in the display combination are fitted to obtain the fitted target paths. The probability value of the fitted target paths in each time period is set as the probability of the display combination appearing. The probability of the displayed combination is compared with the references in the multimedia resource library, the difference value is identified, and the resources in the multimedia resource library are classified according to the difference value to complete the update of the multimedia resource library.

2. The multimedia interactive display system for exhibition halls according to claim 1, characterized in that, The interactive data acquisition module is implemented in the following ways: For any interactive information from a tourist, obtain the corresponding classification model for that interactive information; Use a classification model to classify interactive information to obtain at least one interactive category; Identify keywords and related terms in the interactive information under the corresponding interactive category to form an interactive sample set; By analyzing the keywords and related words in the interactive sample set, keyword areas and related word areas are obtained, which are used as part of the interactive information.

3. The multimedia interactive display system for exhibition halls according to claim 2, characterized in that, Other methods for obtaining the keyword and related word regions of the interactive sample set include: The keywords and related terms in the interaction sample set are merged according to the interaction category to obtain multiple merged results; Extract keyword pairs from the merged results and compare them with the keyword dictionary to obtain the keyword region; Extract the association strength of related words in the merged results, and divide the merged results according to the association strength to obtain the related word region.

4. The multimedia interactive display system for exhibition halls according to claim 1, characterized in that, The implementation of the display partition module also includes: The display zones are evaluated, and the frequency of visit and dwell time of visitors in the display zones are analyzed. The display zones are then fitted according to the frequency of visit and dwell time to construct a mapping relationship between the display zones and visitors' interests.

5. The multimedia interactive display system for exhibition halls according to claim 3, characterized in that, The methods for establishing a mapping relationship between display zones and multimedia resource libraries include: using the information representation of keyword areas and related word areas in the display zones, and the descriptive information and categories of the multimedia resource libraries, to construct a mapping relationship between the display zones and the multimedia resource libraries.

6. The multimedia interactive display system for exhibition halls according to claim 1, characterized in that, The display management module is implemented as follows: Based on the updated mapping relationship between the display partitions and the multimedia resource library, display targets are extracted from the interactive information, and the display targets are sorted according to their probability of occurrence to obtain the display order; The presentation plan is obtained by combining the presentation objectives and presentation order in a structured manner.

7. A multimedia interactive display method for exhibition halls, characterized in that, It includes all modules and method flows of the multimedia interactive display system for exhibition halls as described in any one of claims 1 to 6.

Images