Non-residual element feature extraction method based on knowledge graph

Abstract

The invention relates to the technical field of data processing, and provides a non-residual element feature extraction method based on a knowledge graph. The method comprises the following steps: defining interface arrangement parameters; verifying the interface arrangement parameters, and developing an interface arrangement management system based on the interface arrangement parameters after verification is passed; determining a to-be-accessed system, and acquiring a multi-source non-residual element set through interface arrangement management system integration; initializing a non-residual element knowledge graph, identifying a multi-source non-residual element set, and obtaining key non-residual elements; carrying out multi-scale feature extraction on the key non-residual elements, and outputting multi-scale fusion non-residual elements; and fusing the multi-scale fused non-residual elements, and outputting a non-residual element extraction result, so as to solve the technical problems of low non-residual element feature recognition accuracy and poor feature fusion effect caused by high interface arrangement complexity and difficulty in coping with the complexity and isomerism of multi-source data in the existing non-residual element feature extraction process.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method for extracting features of intangible cultural heritage elements based on a knowledge graph. Background Art

[0002] Intangible cultural heritage (referred to as ICH) refers to various traditional cultural expressions that have been passed down from generation to generation by people of all ethnic groups and are closely related to their lives, including oral legends, performing arts, festivals, traditional handicrafts, etc. These cultural expressions are an important part of national culture and have important historical, cultural and social values. However, with the acceleration of globalization and the advancement of modernization, many ICH projects are facing difficulties in inheritance and the crisis of extinction, and they are in urgent need of protection and inheritance through modern scientific and technological means. Existing ICH protection methods mainly rely on manual collection and sorting, and lack systematic and intelligent means to comprehensively and accurately identify and extract ICH elements. This method is not only inefficient, but also prone to omissions and errors, and cannot meet the needs of large-scale ICH protection and utilization. In addition, when processing multi-source heterogeneous data, existing technologies often face problems such as difficulty in data integration, inaccurate element identification, and poor feature extraction, resulting in unsatisfactory inheritance and promotion of ICH elements in different application scenarios. Summary of the invention

[0003] The present invention provides a method for extracting features of intangible cultural heritage elements based on a knowledge graph, aiming to solve the technical problems in the existing process of extracting features of intangible cultural heritage elements, such as high interface arrangement complexity, difficulty in coping with the complexity and heterogeneity of multi-source data, resulting in low accuracy of intangible cultural heritage element feature recognition and poor feature fusion effect.

[0004] In view of the above problems, the present invention provides a method for extracting features of intangible cultural heritage elements based on knowledge graph.

[0005] The present invention provides a method for extracting features of intangible cultural heritage elements based on a knowledge graph, the method comprising: defining interface orchestration parameters, the interface orchestration parameters comprising the number of interfaces, interface classification, and transmission mode; performing verification through the defined interface orchestration parameters, and if the verification passes, developing an interface orchestration management system based on the defined interface orchestration parameters; determining a plurality of systems to be connected, integrating the plurality of systems to be connected through the interface orchestration management system, and obtaining a multi-source intangible cultural heritage element set; initializing an intangible cultural heritage element knowledge graph, identifying the multi-source intangible cultural heritage element set according to the intangible cultural heritage element knowledge graph, and obtaining a plurality of key intangible cultural heritage elements; performing multi-scale element feature extraction on the plurality of key intangible cultural heritage elements, and outputting multi-scale fused intangible cultural heritage elements; fusing the multi-scale fused intangible cultural heritage elements, and outputting intangible cultural heritage element extraction results.

[0006] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0007] The above-mentioned method for extracting features of intangible cultural heritage elements based on knowledge graphs first defines parameters related to interfaces, including quantity, classification, and transmission methods, etc., to guide how to arrange these interfaces. Subsequently, verification is performed based on these defined parameters to ensure that they are correct, and then an interface arrangement management system is constructed based on these parameters. Then, through the constructed interface arrangement management system, multiple systems that need to be connected are integrated, and diversified intangible cultural heritage element information, i.e., a multi-source intangible cultural heritage element set, is obtained from these systems. After that, an intangible cultural heritage element knowledge graph is established to help identify and organize these intangible cultural heritage elements. Using this intangible cultural heritage element knowledge graph, multiple key intangible cultural heritage elements are extracted from the obtained multi-source intangible cultural heritage elements. For these key intangible cultural heritage elements, feature extraction is performed using multiple different scales to more comprehensively understand the characteristics of each element. Then, the features at these different scales are fused to obtain a more comprehensive and rich representation of intangible cultural heritage elements, and these fused intangible cultural heritage elements are integrated to form an intangible cultural heritage element extraction result. This process helps to improve the accuracy and efficiency of intangible cultural heritage element feature extraction, so as to better protect, inherit and display intangible cultural heritage.

[0008] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of a flow chart of a method for extracting features of intangible cultural heritage elements based on a knowledge graph in one embodiment;

[0011] Figure 2 A flowchart of constructing a similarity matrix for an intangible cultural heritage element feature extraction method based on a knowledge graph in one embodiment. DETAILED DESCRIPTION

[0012] The embodiments of the present invention provide a method for extracting features of intangible cultural heritage elements based on a knowledge graph, thereby solving the technical problems in the existing process of extracting features of intangible cultural heritage elements, such as high complexity of interface arrangement, difficulty in coping with the complexity and heterogeneity of multi-source data, resulting in low accuracy of intangible cultural heritage element feature recognition and poor feature fusion effect.

[0013] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0014] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0015] Examples, such as Figure 1 As shown, the present invention provides a method for extracting features of intangible cultural heritage elements based on a knowledge graph, the method comprising:

[0016] Define interface arrangement parameters, which include the number of interfaces, interface classification, and transmission mode.

[0017] In an embodiment of the present invention, the requirements and functions of the entire system are first analyzed, and the number of systems that need to be integrated and the interaction requirements between the systems are preliminarily determined, thereby determining the total number of interfaces. The determination of the number of interfaces needs to take into account the scalability of the system and future integration requirements, so that the interface arrangement can support the seamless integration of multiple systems. Subsequently, the interfaces are classified according to the functions, data types, communication protocols and other characteristics of each system. Interface classification includes data interface, service interface, management interface and the like. Each type of interface may involve different communication methods, data formats, and security requirements, so classification can help clarify the functional positioning and technical implementation path of the interface. After that, a transmission method suitable for each type of interface is selected. The transmission method can be synchronous communication, asynchronous communication, message queue, RESTfulAPI, SOAP and the like. When selecting a transmission method, it is necessary to comprehensively consider the real-time, reliability, security and system architecture requirements of data transmission. For example, for a real-time data interface that requires a fast response, a WebSocket-based transmission method can be selected; for an interface with a large amount of data and the need to ensure transmission reliability, a message queue method can be selected. After the interface classification, transmission mode, etc. are determined, multiple interfaces are generated, and a set of interface orchestration parameters suitable for the current system requirements are defined based on these interfaces, providing clear guidance for subsequent interface development and system integration.

[0018] Verify through the defined interface orchestration parameters. If the verification passes, develop an interface orchestration management system based on the defined interface orchestration parameters.

[0019] In one embodiment, after obtaining the interface orchestration parameters, a verification model is established for the defined interface orchestration parameters, and this model includes a consistency verification channel and a security verification channel. By inputting the interface orchestration parameters into the pre-built verification model, consistency and security verification are performed respectively through the two channels inside the verification model. When the verification results of the two channels are greater than or equal to the preset thresholds, it means that the interface orchestration parameters are set reasonably and meet the system requirements, and an instruction is issued that the verification is passed and the next development stage is carried out. After receiving the verification pass instruction, the interface orchestration management system is developed. This interface orchestration management system is used to integrate regional e-commerce, smart tourism platforms, regional intangible cultural heritage archive management and other systems. Specifically, the interface parameters are defined as Interface(I1, I2, …I n; Type), where I represents a single interface, n represents the number of interfaces, and Type represents the interface classification. For each interface I, its specific functions and behaviors can be abstractly defined as I(A(c), A(p), Desr, Tans, Tag), where A(c) represents the abstract list of front-end request services, A(p) represents the abstract list of back-end response services, Desr represents different interface descriptions, Tans represents the transmission connection method, and Tag represents the abstract interface identifier. These definitions provide the basis for interface orchestration and ensure the coordination between different interfaces. The interface orchestration method is defined as Method(Sim(I i ,I j ),Mat(Max(Sim),I),Arr(Uni,Sec)), where Sim(I i ,I j ) calculates the interface similarity, Mat(Max(Sim),I) is used to obtain the interface with the greatest similarity and match it, and Arr(Uni,Sec) is used to detect the consistency and security of interface arrangement. i ,I j ) calculates the similarity between different interfaces, determines the interface pairs that can be matched, and then uses Mat(Max(Sim),I) to match the interfaces to ensure smooth data transmission. Subsequently, consistency and security detection is performed through Arr(Uni,Sec) to ensure the overall stability and security of the orchestration process. When interface I passes the orchestration operation of Method, that is, after similarity calculation, interface matching, and consistency and security verification, the interface orchestration logic is completed. Afterwards, according to the defined interface and orchestration method, the system architecture is designed, including the interface management module, the data processing module, etc., to construct an interface orchestration management system to achieve efficient identification and processing of multi-source intangible cultural heritage elements, ensuring that the system has a high degree of stability and security in complex application scenarios.

[0020] Further, if Figure 2 As shown, the present invention provides a method for verifying the defined interface arrangement parameters, including:

[0021] Perform pairwise similarity calculation on all interfaces to obtain a similarity set, wherein the similarity calculation includes interface functions, input and output parameters, and data types; generate a similarity matrix based on the similarity set, wherein each element in the matrix represents a similarity value between two interfaces; based on the similarity matrix, obtain a first interface pair with the greatest similarity, and perform matching based on the first interface pair.

[0022] Preferably, in order to more accurately verify the interface arrangement parameters, the system terminal first determines the standard for calculating the interface similarity, including interface functions, input and output parameters and data types. Among them, by comparing whether the two interface functions are similar, it can be determined whether they perform similar operations or achieve similar goals. By comparing the input and output parameters of the two interfaces, it can be checked whether their types, quantities and structures are consistent. By comparing the data types transmitted by the two interfaces, it can be determined whether their data formats are compatible. Subsequently, all interfaces in the system are combined in pairs, and the similarity of each pair of interfaces is calculated using the corresponding similarity calculation method according to the above standards. For the interface function, first define a set of functional features for each interface, such as the main operations of the interface (such as reading, writing, deleting, etc.), the frequency and timeliness of the operation, etc., and vectorize the functional features to form feature vectors of the two interfaces, for example, using binary values ​​to indicate whether a certain function exists (1 means existence, 0 means non-existence). Subsequently, the Jaccard similarity coefficient or cosine similarity is used to calculate the functional similarity of the two interfaces to obtain the interface function similarity. For the input and output parameters, the input parameters and output parameters are first listed. Common parameters include integers, strings, floating point numbers, JSON objects, etc. Then compare the input parameters and output parameters of the two interfaces respectively to check whether their types are consistent, and calculate the similarity of the parameter names through Levenshtein distance. Subsequently, the similarity of the input parameters and output parameters is processed by the mean to obtain the similarity of the input and output parameters. For the data type, first list the specific data types transmitted by each interface, such as JSON, XML, CSV, text, etc., and then use Levenshtein distance to calculate the similarity of the data types of the two interfaces to obtain the data type similarity. After obtaining the interface function similarity, input and output parameter similarity and data type similarity, the three similarities are processed by the mean to obtain the similarity of the two interfaces, which is added to the similarity set, and the above process is repeated until all interface pairs are calculated. After that, a similarity matrix is ​​constructed based on the obtained similarity set. This matrix is ​​a symmetric matrix, and the rows and columns represent the interfaces in the system. Each element in the matrix represents the similarity value between the corresponding row and column interfaces. Then, in the similarity matrix, find the non-diagonal element with the largest similarity value. This value represents the pair of interfaces with the highest similarity among all interface combinations. And according to the maximum similarity value found, the corresponding interface pair I is determined. i and I j, this pair of interfaces is most similar in terms of functions, input and output parameters, and data types. Finally, based on the interface pairs with the greatest similarity, they are preliminarily matched together. This means that these interfaces can work together or replace each other in the calling and processing of the system. Through the above process, the system can efficiently identify and match the most similar interface pairs, laying the foundation for the subsequent development of the interface orchestration management system. This method not only improves the compatibility of the interface, but also optimizes the overall performance and data processing capabilities of the system.

[0023] Furthermore, the present invention provides verification through defined interface arrangement parameters, including:

[0024] Obtain an orchestration verification model, the orchestration verification model comprising a consistency verification channel and a security verification channel, wherein the consistency verification channel is used to perform data transmission verification on the defined interface orchestration parameters and output a consistency verification index, and the security verification channel is used to perform data security verification on the defined interface orchestration parameters and output a security verification index; if both the consistency verification index and the security verification index are greater than or equal to a preset threshold, obtain a verification pass result; if either the consistency verification index or the security verification index is less than the preset threshold, obtain an orchestration update instruction, and redefine the interface orchestration parameters according to the orchestration update instruction until a verification pass result is obtained.

[0025] Preferably, after the interface orchestration parameters pass the similarity calculation, the pre-built orchestration verification model is obtained and activated, and the orchestration verification model includes a consistency verification channel and a security verification channel. Among them, the consistency verification channel is used to verify whether the data transmission between the interface orchestration parameters is consistent and reliable. This channel mainly verifies the data transmission integrity and transmission efficiency of the interface. The security verification channel is used to evaluate the security of the interface orchestration parameters to ensure that the encryption level, user authentication complexity and other aspects of the data during the transmission process meet the requirements. When the orchestration verification model is activated, the interface orchestration parameters are input into the orchestration verification model for verification. In the consistency verification channel, by weighting the data transmission integrity, transmission efficiency and other aspects, a consistency verification index is obtained and transmitted to the input layer. In the security verification channel, by weighting the encryption level, user authentication complexity and other aspects, a security verification index is obtained and transmitted to the input layer, and output together with the consistency verification index. After the orchestration verification model outputs the consistency verification index and the security verification index, the size relationship between the consistency verification index and the security verification index and the preset threshold is compared. If both indicators are greater than or equal to the preset threshold, it means that the verification is passed, and the result of the verification passing is generated. On the contrary, if any of the consistency check indicators or the security check indicators is lower than the preset threshold, an orchestration update instruction will be generated, prompting the need to redefine and adjust the interface orchestration parameters. According to this orchestration update instruction, adjust the number, classification, transmission method or other parameters of the interface, re-orchestrate, and verify again through the orchestration verification model until both verification indicators meet the preset threshold and obtain the verification result. Through this process, the reliability and security of the interface orchestration management system when integrating and calling multiple interfaces can be ensured, providing a solid foundation for the stable operation of the system.

[0026] Furthermore, the present invention provides a consistency verification channel and a security verification channel, including:

[0027] The consistency verification channel is used to perform data transmission verification on the defined interface orchestration parameters, obtain data transmission integrity, data transmission efficiency and transmission mode matching, calculate the data transmission integrity, data transmission efficiency and transmission mode matching through the first weighted network layer, and output a consistency verification index; the security verification channel is used to perform data security verification on the defined interface orchestration parameters, obtain data encryption level, user authentication complexity and authority control protection, calculate the data encryption level, user authentication complexity and authority control protection through the second weighted network layer, and output a security verification index.

[0028] Optionally, the consistency verification channel is used to verify various indicators of the interface orchestration parameters in data transmission, such as data transmission integrity, data transmission efficiency, transmission mode matching, etc., to ensure that the data transmission in the system is reliable and efficient. In order to build a consistency verification channel, a large amount of historical interface orchestration case data is first obtained, including successful and failed cases. Each case contains the actual interface orchestration parameters and their corresponding transmission integrity, transmission efficiency, transmission mode matching and other results. Then, the transmission integrity, transmission efficiency and transmission mode matching are annotated for each case for supervised learning. Subsequently, based on the multi-layer perceptron, the consistency verification channel structure is set in combination with business requirements and data characteristics, including input layer, hidden layer and output layer. Then, the historical interface orchestration case data is divided to generate training set, validation set and test set, and the weight and bias of the consistency verification channel are initialized. Common initialization methods include random initialization, Gaussian distribution initialization and Xavier initialization. After that, the training set is input into the consistency verification channel, passed through the input layer, hidden layer, and finally passed to the output layer to generate a prediction result. Then, the mean square error (MSE) is used to calculate the error between the predicted value and the true value. After calculating the error, the back-propagation algorithm is used to calculate the gradient of the mean square error loss function with respect to the consistency verification channel parameters. These gradients indicate the direction and magnitude of the consistency verification channel parameters that need to be adjusted. Based on the calculated gradient, the selected optimizer (such as SGD or Adam) is used to update the weights and biases of the model so that the loss function value gradually decreases. This training process is performed in multiple batches. After each batch, the consistency verification channel is evaluated using the validation set to calculate the validation loss. This helps monitor the performance of the consistency verification channel on unseen data and avoid overfitting or underfitting. If the loss of the validation set does not decrease significantly in several consecutive batches, stop training early and use the test set to perform a final evaluation of the consistency verification channel to calculate the loss on the test set and various performance indicators (such as R 2, MAE, etc.), to ensure that the model can still maintain good performance on unseen data. When the verification is passed, the current consistency verification channel is saved. For the first weighted network layer, it is constructed by setting the weights of each verification parameter according to the actual needs of the system. For example, data transmission integrity may be more important than transmission efficiency, so it is given a higher weight. After constructing the consistency verification channel and the first weighted network layer, the output layer of the consistency verification channel is connected to the first weighted network layer, and the data transmission integrity, data transmission efficiency and transmission mode matching generated by the consistency verification channel are weighted and summed through the weights set inside the first weighted network layer to generate consistency verification indicators. Similarly, using the same construction method as above, based on the historical interface orchestration case data and its corresponding data encryption level, user authentication complexity and permission control protection, a security verification channel is constructed and connected to the second weighted network layer to evaluate the protection capabilities of the interface orchestration parameters in data security, including data encryption level, user authentication complexity and permission control protection, and generate security verification indicators to ensure the confidentiality, integrity and availability of data during transmission. By integrating the consistency verification channel and the security verification channel, an orchestration verification model is constructed. This model evaluates the overall effectiveness of interface orchestration parameters through the consistency verification indicators and security verification indicators output by two verification channels, ensuring that the system has high reliability and data security in actual operation.

[0029] Determine a plurality of systems to be connected, integrate the plurality of systems to be connected through the interface orchestration management system, and obtain a multi-source intangible cultural heritage element set.

[0030] In one embodiment, various data sources related to intangible cultural heritage elements are determined according to business needs. These data sources may include regional e-commerce systems, smart tourism platforms, intangible cultural heritage archive management systems, etc. Then clarify the role and function of each system in the extraction and processing of intangible cultural heritage elements. For example, the regional e-commerce system may provide sales data of intangible cultural heritage products, the smart tourism platform provides tourist behavior data, and the intangible cultural heritage archive management system provides specific information on intangible cultural heritage. Subsequently, the interface of each system to be connected is called through the interface orchestration management system to obtain intangible cultural heritage element data from these systems. Ensure that the data can be transmitted according to the predetermined interface orchestration parameters. After that, the data obtained from each system is synchronized and integrated. In this process, duplicates and abnormal data will be removed, and data from different systems will be converted into the same data format. Then, according to the timestamp, the data from different systems will be aligned to ensure the consistency and accuracy of the data. Then, the cleaned and aligned data will be aggregated into a multi-source intangible cultural heritage element set. This set will contain intangible cultural heritage element information from all systems to be connected, providing a reliable data basis for subsequent data analysis and application.

[0031] Furthermore, the present invention provides a plurality of systems to be accessed, including:

[0032] The multiple systems to be connected include at least the regional e-commerce, regional smart tourism platform and regional integrated media platform, regional intangible cultural heritage archive management system, regional intangible cultural heritage ontology and regional knowledge graph system, regional intangible cultural heritage and cultural tourism resource distribution GIS system and regional intangible cultural heritage virtual simulation interactive system in the target area.

[0033] Preferably, the multiple systems to be connected include regional e-commerce, regional smart tourism platform and regional integrated media platform, regional intangible cultural heritage archive management system, regional intangible cultural heritage ontology and regional knowledge graph system, regional intangible cultural heritage and cultural tourism resource distribution GIS system and regional intangible cultural heritage virtual simulation interactive system in the target area. Among them, regional e-commerce, regional smart tourism platform and regional integrated media platform are existing systems, and regional intangible cultural heritage archive management system, regional intangible cultural heritage ontology and regional knowledge graph system, regional intangible cultural heritage and cultural tourism resource distribution GIS system and regional intangible cultural heritage virtual simulation interactive system are newly developed systems. The regional intangible cultural heritage archive management system is developed based on MongoDB, and data is stored and managed according to the intangible cultural heritage classification. The regional intangible cultural heritage ontology and regional knowledge graph system are also developed based on MongoDB to ensure that the system can provide accurate intangible cultural heritage element identification and knowledge association. The regional intangible cultural heritage and cultural tourism resource distribution GIS system is developed using GIS technology and MongoDB to realize the geographic information management and display of resources. The intangible cultural heritage virtual simulation interactive system uses 3D and VR technology for intangible cultural heritage virtual simulation and interactive design to enhance user experience.

[0034] Initialize the intangible cultural heritage element knowledge graph, identify the multi-source intangible cultural heritage element set according to the intangible cultural heritage element knowledge graph, and obtain multiple key intangible cultural heritage elements.

[0035] In one embodiment, the nodes and relationships in the intangible cultural heritage element structure sub-graph are first determined, the nodes represent different intangible cultural heritage elements, and the relationships represent the associations between these elements. Then, according to the data type of the interface orchestration management system, multiple intangible cultural heritage element structure sub-graphs are constructed. Subsequently, according to the similarity of the node attributes between different intangible cultural heritage element structure sub-graphs, the node information is merged to generate an intangible cultural heritage element knowledge graph. Afterwards, in the intangible cultural heritage element knowledge graph, a multi-source intangible cultural heritage element set is matched, and according to the matching results, a screening mark is performed in combination with a preset matching degree to generate multiple key intangible cultural heritage elements for subsequent feature extraction, analysis, and application.

[0036] Furthermore, the present invention provides initialization of the intangible cultural heritage element knowledge graph, including:

[0037] Define the node attribute information and node dependency of each intangible cultural heritage element structural sub-graph; integrate the data types of the interface orchestration management system, and obtain multiple intangible cultural heritage element structural sub-graphs according to the node attribute information and the node dependency, wherein each intangible cultural heritage element structural sub-graph corresponds to a data type, and the nodes of each intangible cultural heritage element structural sub-graph correspond to an intangible cultural heritage element; calculate the node attribute similarity between different intangible cultural heritage element structural sub-graphs, merge the information of nodes with a similarity greater than a preset similarity, fuse the multiple intangible cultural heritage element structural sub-graphs, and generate an intangible cultural heritage element knowledge graph.

[0038] Preferably, attribute information is defined for the nodes in each intangible cultural heritage element structure sub-graph according to different data types, such as video, image, text, audio, etc. For example, video node attributes may include the name, length, content description, related characters or events of the video. Image node attributes may include the shape, color, cultural significance, historical background, etc. of the image. Text node attributes may include the content, language, source, events or stories described by the text, etc. Audio node attributes may include the title, length, sound quality, content description, sound source (such as the call of a dragon), etc. of the audio. Dependencies are then defined for the nodes in each sub-graph. For example, the relationship between an image and a text or a historical event, such as a dragon-shaped image and a dragon story. The association between text and audio, such as a story describing a dragon and an audio recording of a dragon's call. Subsequently, the various types of data obtained by the interface orchestration management system from different data sources are integrated. Ensure the consistency of data types and map them to corresponding node attributes and dependencies, dog name and multiple intangible cultural heritage element structure sub-graphs, each intangible cultural heritage element structure sub-graph corresponds to a data type, and each node of the intangible cultural heritage element structure sub-graph corresponds to an intangible cultural heritage element. Taking video and text as an example, a sub-graph describing a specific skill or historical event is generated based on video data. Each node represents an element in the video, such as a skill action or an event scene. Based on text data, a sub-graph describing a story or historical event is generated, and each node represents text content, such as the story of a dragon. Afterwards, similarity calculation standards are defined for nodes in different sub-graphs, such as based on attributes such as node name, description, historical background, and cultural significance, and similarity is calculated using similar similarity calculation methods described above. For example, node names are extracted from dragon-shaped images and texts describing dragon stories, name similarity is performed using Levenshtein distance, feature vectors of dragon-shaped images and texts describing dragon stories are obtained, and cosine similarity is used to calculate similarity to obtain cultural similarity, and then the two similarities are averaged to obtain node attribute similarity. After calculating the node attribute similarity of each two nodes, these node attribute similarities are compared with the preset similarity, and the information of nodes whose node attribute similarity is greater than the preset similarity is merged. For example, a dragon image, a dragon story text, and a dragon's call audio are merged into one node to form a comprehensive node with multiple meanings. After merging similar nodes in multiple intangible cultural heritage element structural subgraphs, a complete intangible cultural heritage element knowledge graph is generated. Each node in the graph may contain information from multiple data sources such as video, image, text, audio, etc., forming a node with multiple meanings, thereby achieving the effect of processing multi-source data at the same time.

[0039] Furthermore, the present invention provides a method for identifying the multi-source intangible cultural heritage element set according to the intangible cultural heritage element knowledge graph to obtain multiple key intangible cultural heritage elements, including:

[0040] Based on the intangible cultural heritage element knowledge graph, node matching is performed on the multi-source intangible cultural heritage element set to obtain the data feature vector matching degree of each node; according to the size of the data feature vector matching degree, k nodes with a vector matching degree greater than a preset degree are screened, the k nodes are marked on the intangible cultural heritage element knowledge graph, the intangible cultural heritage elements corresponding to the k nodes are extracted, and multiple key intangible cultural heritage elements are obtained.

[0041] Preferably, for each element in the multi-source intangible cultural heritage element set, the corresponding data feature vector is extracted. These feature vectors are obtained from data forms such as image features (extracted by existing convolutional neural networks), text features (extracted by existing embedding models), and audio features. Each element in the multi-source intangible cultural heritage element set is then matched with a node in the intangible cultural heritage element knowledge graph, and the cosine similarity is used to calculate the matching degree between the feature vector of each element in the multi-source intangible cultural heritage element set and the feature vector of each node in the graph, and the data feature vector matching degree of each node is obtained. Subsequently, according to business needs, a preset vector matching degree is set, and the data feature vector matching degree of each node is screened using this preset vector matching degree, and k nodes whose data feature vector matching degree is greater than the preset vector matching degree are obtained. Afterwards, the k nodes screened out are marked on the intangible cultural heritage element knowledge graph. The marking method includes changing the node color, thickening the node border, adding annotations, etc., so that it is highlighted in the graph for easy identification. Then, the intangible cultural heritage elements corresponding to the K nodes are input into the pre-built user test evaluation model, and the intangible cultural heritage elements are evaluated in terms of usability, satisfaction, accuracy, etc. through the user test evaluation model, and then the evaluation results are weighted and summed to obtain k evaluation indicators. Finally, based on the preset evaluation indicators, the k evaluation indicators are screened to obtain multiple key intangible cultural heritage elements. These key intangible cultural heritage elements will serve as the core data for subsequent analysis, display or application, representing the most representative cultural elements in the multi-source intangible cultural heritage element collection, and providing a solid data foundation for cultural protection, inheritance and promotion.

[0042] Furthermore, the present invention provides a method for extracting the intangible cultural heritage elements corresponding to the k nodes to obtain a plurality of key intangible cultural heritage elements, and further includes:

[0043] Establish a user testing and evaluation model; input the intangible cultural heritage elements corresponding to the k nodes into the user testing and evaluation model, including usability evaluation, satisfaction evaluation, accuracy evaluation, communication evaluation, and aesthetics evaluation, and obtain k evaluation indicators corresponding to the k intangible cultural heritage elements; output the intangible cultural heritage elements that are greater than the preset evaluation indicators as the multiple key intangible cultural heritage elements.

[0044] Optionally, first collect a large number of intangible cultural heritage element data sets with known evaluation results. Each intangible cultural heritage element has a corresponding score in the evaluation dimensions of usability, satisfaction, accuracy, dissemination, and aesthetics. Among them, usability is to evaluate the ease of use of intangible cultural heritage elements, such as ease of operation and user experience. Satisfaction is to evaluate the overall satisfaction of users with intangible cultural heritage elements. Accuracy is to evaluate the degree of match between intangible cultural heritage elements and their actual cultural background or history. Dissemination is to evaluate the dissemination effect of intangible cultural heritage elements in social media or educational scenarios. Aesthetics is to evaluate the visual or auditory beauty of intangible cultural heritage elements. Subsequently, the same method as the aforementioned construction of the consistency verification channel is used to train the user test evaluation model based on the multilayer perceptron, the third weight network layer, and the collected intangible cultural heritage element data set. After the training is completed, the intangible cultural heritage elements corresponding to the selected k nodes are input into the trained user test evaluation model. The user test evaluation model outputs the scores of each intangible cultural heritage element in the five dimensions of usability, satisfaction, accuracy, dissemination, and aesthetics according to the learned mapping relationship, and weighted sums are performed through the third weight network layer to obtain k evaluation indicators of k intangible cultural heritage elements. Afterwards, by comparing the k evaluation indicators with the preset evaluation indicators respectively, the intangible cultural heritage elements whose evaluation indicators are greater than the preset evaluation indicators are screened out, and the screened out multiple intangible cultural heritage elements are used as multiple key intangible cultural heritage elements.

[0045] Perform multi-scale element feature extraction on the multiple key intangible cultural heritage elements, and output multi-scale fused intangible cultural heritage elements; fuse the multi-scale fused intangible cultural heritage elements, and output intangible cultural heritage element extraction results.

[0046] In one embodiment, after obtaining multiple key intangible cultural heritage elements, these key intangible cultural heritage elements are subjected to multi-scale element features, including local scale features and global scale features of each key intangible cultural heritage element. Local scale features can extract local detail features of intangible cultural heritage elements, such as texture, detail structure, features of specific parts, etc. Global scale features can extract the overall features of intangible cultural heritage elements, focusing on macro information, such as shape, contour, overall structure, etc. Subsequently, the local features and global features of each key intangible cultural heritage element extracted are integrated to generate local feature vectors and global feature vectors of each key intangible cultural heritage element, and thereby obtain multi-scale fused intangible cultural heritage elements. Afterwards, the local features and global features of each key intangible cultural heritage element in the multi-scale fused intangible cultural heritage element are fused and spliced ​​to generate the final intangible cultural heritage element extraction result. This intangible cultural heritage element extraction result contains the key information of multiple intangible cultural heritage elements at different scales, and has a strong expressive ability.

[0047] Furthermore, the present invention provides a method for extracting multi-scale element features from the multiple key intangible cultural heritage elements, further comprising:

[0048] Perform local feature extraction on the multiple key intangible cultural heritage elements to obtain a local feature vector corresponding to each key intangible cultural heritage element; perform global feature extraction on the multiple key intangible cultural heritage elements to obtain a global feature vector corresponding to each key intangible cultural heritage element; splice the local feature vector and the global feature vector corresponding to each key intangible cultural heritage element to output a comprehensive feature vector; output the intangible cultural heritage element extraction result based on the comprehensive feature vector.

[0049] Optionally, for data in multiple key intangible cultural heritage elements, local feature extraction is performed by taking image data, text data, and audio data as examples. For image data, key points in the image are located by feature point detection algorithms such as SIFT, SURF, and ORB. These key points represent important local areas in the image, such as edges, corners, etc. The detected feature points are then described to generate image local feature vectors. Descriptors such as SIFT descriptors or SURF descriptors can capture texture and shape information of local areas of the image. For text data, natural language processing technology is used to extract key phrases or phrases in the text, which represent the core content of the text. These phrases or phrases are then input into existing text embedding models, such as Word2Vec and BERT, to generate corresponding text local feature vectors. For audio data, the audio signal is divided into short time segments, and local spectral features are extracted using short-time Fourier transform or wavelet transform. The spectral features extracted from each time segment are then encoded to generate audio local feature vectors, which reflect the change information of the audio in a specific time period. Subsequently, the image local feature vector, text local feature vector and audio local feature vector are concatenated to form a complete local feature vector, which contains the information of the intangible cultural heritage elements at the local scale. For each key intangible cultural heritage element, its corresponding local feature vector is output for subsequent feature concatenation. Similarly, global feature extraction is performed on the image data, text data and audio data in multiple key intangible cultural heritage elements. For image data, the existing convolutional neural network, such as ResNet and VGG, is used to extract the global features of the image. The image global feature vector is then output through the fully connected layer or the last convolutional layer of the network. For text data, a text embedding model is also used to generate a text global feature vector representing the overall semantics of the text. For audio data, Mel frequency cepstral coefficient (MFCC) is used to extract the global features of the audio and generate an audio global feature vector. MFCC captures the overall frequency distribution and pitch characteristics of the audio. Afterwards, the extracted image global feature vector, text global feature vector and audio global feature vector are concatenated to obtain a complete global feature vector, which reflects the information of the intangible cultural heritage elements at the overall level. For each key intangible cultural heritage element, output its corresponding global feature vector as the basis for comprehensive feature splicing. Then, splice the local feature vector and the global feature vector of each key intangible cultural heritage element. The splicing method is direct series connection, that is, connecting the local feature vector and the global feature vector into a longer comprehensive feature vector. Then, for each key intangible cultural heritage element, output its corresponding comprehensive feature vector. This vector combines local and global feature information and is the most comprehensive description of the intangible cultural heritage element. Finally, all comprehensive feature vectors are summarized to generate the final intangible cultural heritage element extraction results. These results combine the local and global characteristics of the intangible cultural heritage elements and can fully reflect the cultural connotation and expression of the intangible cultural heritage elements.

[0050] In summary, the embodiments of the present invention have at least the following technical effects:

[0051] The embodiment of the present invention first defines the interface arrangement parameters, and develops the interface arrangement management system after passing the verification. Through the system, multiple systems to be connected are integrated, a multi-source intangible cultural heritage element set is obtained, and the multi-source intangible cultural heritage elements are identified based on the initialized intangible cultural heritage element knowledge graph, and the key intangible cultural heritage elements are extracted. Multi-scale element feature extraction is further performed on these key intangible cultural heritage elements, and the intangible cultural heritage element extraction results are output after fusion. In addition, the method also includes calculating the similarity of the interface, using the arrangement verification model to verify the consistency and security of the interface parameters, and redefining the interface arrangement parameters as needed. The initialization of the intangible cultural heritage element knowledge graph is performed by integrating data types and calculating the similarity of node attributes. Finally, the key intangible cultural heritage elements are evaluated by the user test evaluation model, and the comprehensive feature vector is output according to the splicing of local and global features to obtain the extraction result of the intangible cultural heritage elements. These technical effects jointly solve the technical problems in the existing intangible cultural heritage element feature extraction process, such as high interface orchestration complexity, difficulty in coping with the complexity and heterogeneity of multi-source data, resulting in low accuracy in intangible cultural heritage element feature recognition and poor feature fusion effect. By defining interface orchestration parameters and optimizing the interface orchestration management system, the feature fusion process is optimized, and the accuracy and efficiency of intangible cultural heritage element feature extraction are improved.

[0052] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0053] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0054] This specification and the accompanying drawings are merely exemplary illustrations of the present invention and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the scope of the present invention. Thus, the present invention is intended to include such changes and modifications if they fall within the scope of the present invention and its equivalents.

Claims

1. The feature extraction method of intangible cultural heritage elements based on knowledge graph is characterized by: The method comprises: Defining interface arrangement parameters, wherein the interface arrangement parameters include the number of interfaces, interface classification, and transmission mode; Verify through defined interface orchestration parameters. If the verification passes, develop an interface orchestration management system based on the defined interface orchestration parameters. Determine a plurality of systems to be connected, integrate the plurality of systems to be connected through the interface orchestration management system, and obtain a multi-source intangible cultural heritage element set; Initialize the intangible cultural heritage element knowledge graph, identify the multi-source intangible cultural heritage element set according to the intangible cultural heritage element knowledge graph, and obtain multiple key intangible cultural heritage elements; Performing multi-scale element feature extraction on the multiple key intangible cultural heritage elements, and outputting multi-scale fused intangible cultural heritage elements; The multi-scale fused intangible cultural heritage elements are fused and the intangible cultural heritage element extraction result is output.

2. The method according to claim 1, characterized in that Before verifying the parameters through the defined interface orchestration, the method includes: Perform pairwise similarity calculation on all interfaces to obtain a similarity set, wherein the similarity calculation includes interface functions, input and output parameters, and data types; Generate a similarity matrix according to the similarity set, wherein each element in the matrix represents a similarity value between two interfaces; Based on the similarity matrix, a first interface pair with the greatest similarity is obtained, and matching is performed according to the first interface pair.

3. The method according to claim 1, characterized in that Verify through defined interface orchestration parameters, including: Acquire an orchestration verification model, wherein the orchestration verification model includes a consistency verification channel and a security verification channel, wherein the consistency verification channel is used to perform data transmission verification on the defined interface orchestration parameters and output a consistency verification index, and the security verification channel is used to perform data security verification on the defined interface orchestration parameters and output a security verification index; If the consistency check index and the security check index are both greater than or equal to the preset threshold, a check pass result is obtained; If any one of the consistency check index and the security check index is less than the preset threshold, an orchestration update instruction is obtained, and the interface orchestration parameters are redefined according to the orchestration update instruction until a verification result is obtained.

4. The method according to claim 3, characterized in that The consistency check channel is used to perform data transmission verification on the defined interface orchestration parameters, obtain data transmission integrity, data transmission efficiency and transmission mode matching, calculate the data transmission integrity, data transmission efficiency and transmission mode matching through the first weighted network layer, and output a consistency check indicator; The security verification channel is used to perform data security verification on the defined interface orchestration parameters, obtain data encryption level, user authentication complexity and authority control protection, calculate the data encryption level, user authentication complexity and authority control protection through the second weighted network layer, and output security verification indicators.

5. The method according to claim 1, characterized in that Initialize the knowledge graph of intangible cultural heritage elements. The methods include: Define the node attribute information and node dependency of each intangible cultural heritage element structure subgraph; Integrate the data types of the interface orchestration management system, and obtain multiple intangible cultural heritage element structure sub-graphs according to the node attribute information and the node dependency, wherein each intangible cultural heritage element structure sub-graph corresponds to a data type, and a node of each intangible cultural heritage element structure sub-graph corresponds to an intangible cultural heritage element; The node attribute similarity between different intangible cultural heritage element structural sub-graphs is calculated, the information of nodes with similarity greater than a preset similarity is merged, and the multiple intangible cultural heritage element structural sub-graphs are fused to generate an intangible cultural heritage element knowledge graph.

6. The method according to claim 1, characterized in that The multi-source intangible cultural heritage element set is identified according to the intangible cultural heritage element knowledge graph to obtain a plurality of key intangible cultural heritage elements, the method comprising: Perform node matching on the multi-source intangible cultural heritage element set based on the intangible cultural heritage element knowledge graph to obtain the matching degree of the data feature vector of each node; According to the size of the data feature vector matching degree, k nodes with a matching degree greater than a preset vector are screened, the k nodes are marked on the intangible cultural heritage element knowledge graph, the intangible cultural heritage elements corresponding to the k nodes are extracted, and multiple key intangible cultural heritage elements are obtained.

7. The method according to claim 6, characterized in that Extracting the intangible cultural heritage elements corresponding to the k nodes to obtain a plurality of key intangible cultural heritage elements, the method further comprises: Establish user testing and evaluation model; Input the intangible cultural heritage elements corresponding to the k nodes into the user test evaluation model, including usability evaluation, satisfaction evaluation, accuracy evaluation, communication evaluation, and aesthetic evaluation, to obtain k evaluation indicators corresponding to the k intangible cultural heritage elements; The intangible cultural heritage elements that are greater than the preset evaluation index are regarded as the multiple key intangible cultural heritage elements.

8. The method according to claim 7, characterized in that Performing multi-scale element feature extraction on the multiple key intangible cultural heritage elements, the method further comprises: Performing local feature extraction on the multiple key intangible cultural heritage elements to obtain a local feature vector corresponding to each key intangible cultural heritage element; Performing global feature extraction on the multiple key intangible cultural heritage elements to obtain a global feature vector corresponding to each key intangible cultural heritage element; According to the local feature vector and global feature vector corresponding to each key intangible cultural heritage element, the comprehensive feature vector is output; Based on the comprehensive feature vector, the intangible cultural heritage element extraction result is output.

9. The method according to claim 1, characterized in that The multiple systems to be connected include at least the regional e-commerce, regional smart tourism platform and regional integrated media platform, regional intangible cultural heritage archive management system, regional intangible cultural heritage ontology and regional knowledge graph system, regional intangible cultural heritage and cultural tourism resource distribution GIS system and regional intangible cultural heritage virtual simulation interactive system in the target area.

Images