Intelligent management and application method of intangible cultural heritage digital resources

Abstract

The invention provides an intelligent management and application method for digital resources of intangible cultural heritage, and belongs to the technical field of crossing of data management and perpetual protection. Comprising the steps of multi-modal non-perpetual data collection and standardization, multi-modal data dynamic weighted fusion, block chain enhanced distributed storage, semantic association enhanced intelligent retrieval, resource dynamic management updating, dynamic value evaluation, intelligent pushing and the like. According to the method, accurate fusion, efficient retrieval and personalized pushing of multi-source data are realized through a self-created multi-modal feature fusion formula, a semantic association degree calculation function and a value evaluation formula; the fusion weight is dynamically adjusted in combination with an analytic hierarchy process, and different types of non-abandoned resource characteristics are adapted; and data security and sharing are guaranteed by adopting a licensed chain and open chain layered architecture. The defects of single-modal acquisition, fixed weight fusion, inaccurate retrieval and no dynamic evaluation in the prior art are overcome, and the method can be widely applied to digital protection and intelligent application scenes of various non-abandoned resources.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data management and intangible cultural heritage protection, and more specifically relates to an intelligent management and application method of intangible cultural heritage digital resources. BACKGROUND

[0002] Intangible cultural heritage (hereinafter referred to as "non-heritage") is an important carrier of national culture, and its digital protection and management has become the core demand of current cultural heritage. With the development of big data and artificial intelligence technology, non-heritage digital management has gradually upgraded from traditional single storage to intelligent application, but the existing technology still has many defects to be solved, which are specifically manifested as follows:

[0003] 1. Single data collection dimension, incomplete cultural connotation preservation: The existing method mainly takes single modal data (such as only collecting images or text) as the core, and cannot consider the multi-dimensional attributes of non-heritage resources. For example, only collecting work images for traditional skill type non-heritage, without synchronously recording the operation audio and skill process video of the inheritor, which leads to the inability to completely restore the skill process and cultural context, and loses the core value of non-heritage inheritance.

[0004] 2. Rigid multi-modal data fusion mode, low feature extraction precision: A few management methods using multi-modal data use fixed weight for data fusion, without considering the feature importance difference of different types of non-heritage resources. For example, the same weight is used for fusion of traditional skill type (emphasis on image and text) and folk performance type (emphasis on audio and video) non-heritage, which leads to the feature vector after fusion cannot accurately reflect the core attributes of non-heritage, affecting the accuracy of subsequent retrieval and evaluation.

[0005] 3. Retrieval mechanism relies on literal matching, lacks semantic association: The existing retrieval method is mainly based on keyword literal matching, and lacks semantic association analysis of non-heritage knowledge system. When the user searches for "random needle embroidery materials", it cannot be associated with "silk thread" and other semantically related resources, resulting in high missed detection rate and low accuracy of the search, which is difficult to meet the user's demand for accurate resource acquisition.

[0006] 4. Lack of dynamic value evaluation mechanism, insufficient precision of push: The existing technology does not establish a dynamic value quantification system of non-heritage digital resources, which cannot adjust the push strategy according to the inheritance characteristics, application demand and dissemination value of the resources. The same resources are pushed to designers, researchers and the public, resulting in low matching degree of push content and user demand, and the application value of non-heritage digital resources cannot be fully played.

[0007] 5. Insufficient balance between data security and sharing: Some methods focus only on data storage without considering the security protection of sensitive intangible cultural heritage resources (such as exclusive techniques and recipes, and unpublished transmission information); other methods excessively restrict data sharing, resulting in limited dissemination and application of digital resources of intangible cultural heritage, which violates the original intention of digital protection.

[0008] To address the shortcomings of existing technologies, there is an urgent need for an intelligent management method for digital resources of intangible cultural heritage that can achieve comprehensive multimodal data collection, dynamic and accurate fusion, efficient semantic retrieval, secure and controllable sharing, and personalized intelligent push notifications, so as to improve the quality and application effectiveness of digital protection of intangible cultural heritage. Summary of the Invention

[0009] To address the aforementioned technical problems, this invention provides an intelligent management and application method for digital resources of intangible cultural heritage. This method solves the technical problems of existing digital management methods for intangible cultural heritage, such as single data collection dimensions, rigid multimodal fusion methods, low accuracy of retrieval relying on literal matching, insufficient push accuracy due to lack of dynamic value assessment, inadequate balance between data security and sharing, and lack of timeliness and completeness of resources.

[0010] A method for intelligent management and application of digital resources of intangible cultural heritage includes the following steps:

[0011] S1. Multimodal Intangible Cultural Heritage Data Acquisition and Standardization: Intangible cultural heritage resource data is acquired collaboratively using 4K ultra-high-definition images, 3D laser scanning, 192kHz high-fidelity audio, and structured text. Based on the DC metadata framework, intangible cultural heritage-specific fields are expanded to generate standardized data units.

[0012] S2. Multimodal Data Weighted Fusion Processing: A comprehensive feature vector is calculated using a self-developed multimodal feature fusion formula, as follows:

[0013] F = α·F t +β·Fᵢ+γ·F a +δ·Fᵥ;

[0014] Where F is the comprehensive feature vector, F t F i F a Fᵥ and Fᵥ are the feature vectors of text, image, audio and video, respectively, and α, β, γ and δ are the fusion weights, satisfying α+β+γ+δ=1, and are dynamically adjusted according to the type of intangible cultural heritage resources (such as traditional crafts / folk performances) through the analytic hierarchy process.

[0015] S3. Blockchain-enhanced distributed storage: Generate a unique digital identifier for each standardized data unit, and construct a blockchain storage node for intangible cultural heritage resources by combining the comprehensive feature vector from step S2. Achieve data ownership confirmation and cross-domain sharing through a layered architecture of permissioned blockchain and open blockchain.

[0016] S4. Intelligent Retrieval Enhanced with Semantic Relevance: A semantic index is constructed based on the intangible cultural heritage knowledge graph, and a self-developed semantic relevance calculation function is used to optimize the retrieval ranking. The function is as follows:

[0017] R(x,y)=ω·Sim(x,y)+ω·Rel(x,y)·Depth(x,y)

[0018] Where R(x,y) is the relevance between the search keyword x and the resource y, Sim(x,y) is the literal similarity, Rel(x,y) is the node association strength in the knowledge graph, Depth(x,y) is the association path depth, and ω and ω are weight coefficients.

[0019] S5. Resource Management and Dynamic Updates: Based on the retrieval feedback data from step S4, dynamically optimize the index weight of data units, and regularly supplement and update resource metadata through interview data with inheritors to form a closed-loop management system.

[0020] Preferably, the intangible cultural heritage-specific fields mentioned in step S1 include technique process nodes, lineage of inheritors, material formula parameters, and regional cultural association attributes, and meet the 27 mandatory field requirements of the "Professional Standard for Digital Protection of Chinese Intangible Cultural Heritage".

[0021] Preferably, in step S2, the image feature vector F is extracted by an improved gray-level co-occurrence matrix, and after the target areas such as intangible cultural heritage patterns and craft tools are segmented, Gaussian smoothing filters are used for weighted noise reduction, and then texture feature parameters are calculated.

[0022] Preferably, in step S3, the blockchain evidence storage node includes a data collection timestamp, an inheritor's authorization signature, and a data modification log, and achieves heterogeneous data synchronization between different intangible cultural heritage databases through a cross-chain interoperability protocol.

[0023] Preferably, in step S4, the intangible cultural heritage knowledge graph extracts intangible cultural heritage entity labels through a BERT pre-trained model and constructs a semantic association network containing more than 32,000 nodes. The node types include intangible cultural heritage projects, inheritors, techniques and processes, and cultural symbols.

[0024] Preferably, the management method according to any one of claims 1-5 further includes the following steps:

[0025] S6. Dynamic Assessment of Resource Value: The resource value coefficient is calculated using a self-developed formula for assessing the value of intangible cultural heritage digital resources.

[0026] V = V·(1-λ·T) + μ·U + ν·I

[0027] Where V is the dynamic value coefficient, V is the basic value, λ is the inheritance attenuation coefficient, T is the digital storage duration, μ is the application frequency weight, U is the application scenario adaptability, and ν is the cultural dissemination gain coefficient.

[0028] S7. Intelligent Application Push: Based on user profile data and the value coefficient of step S6, customized intangible cultural heritage resource data interfaces are pushed to different users (researchers / designers / the general public) through collaborative filtering algorithms.

[0029] Preferably, in step S6, the basic value V is determined by combining expert scoring with big data statistics. The scoring dimensions include historical and cultural value, uniqueness of skills, and urgency of inheritance.

[0030] Preferably, the user profile data in step S7 includes browsing history, search keywords, and application scenario tags, and LBS technology is used to achieve accurate delivery of regional intangible cultural heritage resources.

[0031] Preferably, during the dynamic update process in step S5, an autoregressive integral moving average model is used to supplement the missing skill process data to ensure the integrity of the intangible cultural heritage skill lineage.

[0032] Preferably, it also includes data security safeguards: adopting a consent-based access control mechanism for digital intangible cultural heritage resources, generating access consent tokens, and allowing access to sensitive data (such as exclusive technique recipes) only when the token is verified to be valid.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] This invention employs 4K ultra-high-definition imaging, 3D laser scanning, 192kHz high-fidelity audio, and structured text collaborative acquisition. Combined with extended metadata fields specific to intangible cultural heritage (ICH), it achieves standardized processing, comprehensively covering the multi-dimensional attributes of ICH resources, including text, images, audio, and video. For example, the acquisition of Suzhou embroidery resources includes technique videos, pattern images, audio explanations by inheritors, and textual descriptions of the processes. After standardization, the data integrity is improved, fully preserving the cultural connotations of ICH.

[0035] This invention dynamically adjusts multimodal fusion weights based on the analytic hierarchy process (AHP) and achieves accurate fusion by combining a self-developed multimodal feature fusion formula. For traditional crafts (Suzhou embroidery), it emphasizes image and text weights; for folk performances (Peking Opera), it emphasizes audio and video weights. Compared to fixed-weight fusion methods, the retrieval accuracy is improved by 14-15 percentage points, solving the problems of rigidity and poor adaptability of existing fusion methods.

[0036] This invention constructs a semantic index based on an intangible cultural heritage knowledge graph, and achieves collaborative retrieval of literal matching and semantic association through a self-developed semantic relevance calculation function. Compared with traditional literal retrieval methods, the retrieval accuracy is improved, the false negative rate is effectively reduced, and users can quickly obtain semantically relevant intangible cultural heritage resources, thereby improving resource utilization efficiency.

[0037] This invention quantifies the dynamic value of intangible cultural heritage resources through a self-developed value assessment formula and combines it with user profiles to achieve precise targeting. It pushes resources such as Suzhou embroidery patterns and stitching techniques to designers, and resources such as lineage records and techniques to researchers, achieving an accuracy rate of ≥88%. This is an improvement of 36-38 percentage points compared to targeting methods without dynamic assessment, fully leveraging the application value of intangible cultural heritage resources.

[0038] This invention employs a permissioned blockchain to store sensitive data and an open blockchain to store public data, combined with a cross-chain interoperability protocol to achieve data synchronization. The rate of sensitive data leakage is 100%, and the synchronization latency is ≤500ms. This solves the problem of sensitive data leakage in existing technologies while enabling cross-domain sharing of intangible cultural heritage resources, balancing security and dissemination needs.

[0039] This invention optimizes index weights based on retrieval feedback, uses the ARIMA model to complete missing data, and updates the data after review and confirmation by inheritors, ensuring the timeliness of intangible cultural heritage resources and the integrity of the technical lineage. For example, after completing the missing data on the "random stitch embroidery" process of Suzhou embroidery, the resource integrity increased from 85% to 98%, providing complete and accurate data support for the inheritance of intangible cultural heritage. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the overall process of the present invention;

[0041] Figure 2 This is a schematic diagram of the multimodal intangible cultural heritage data collection and standardization process in this invention;

[0042] Figure 3 This is a schematic diagram of the multimodal data weighted fusion processing flow in this invention;

[0043] Figure 4 This is a schematic diagram of the blockchain-enhanced distributed storage process in this invention;

[0044] Figure 5 This is a schematic diagram of the intelligent retrieval process with enhanced semantic association in this invention;

[0045] Figure 6 This is a schematic diagram of the resource management and dynamic update process in this invention. Detailed Implementation

[0046] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0047] Please see Figures 1-6 This invention provides an intelligent management and application method for digital resources of intangible cultural heritage. This embodiment takes two typical types of intangible cultural heritage as research objects:

[0048] Example 1 is a traditional craft (Suzhou embroidery, a national intangible cultural heritage project, project number: VII-18), and Example 2 is a folk performance (Peking Opera, a national intangible cultural heritage project, project number: IV-2).

[0049] Two sets of comparative examples were set up (Comparative Example 1: Single-modal data management method; Comparative Example 2: Fixed-weight multimodal fusion management method) to verify the technical effect of the present invention through comparison.

[0050] Core technical parameters:

[0051] Data acquisition parameters: 4K ultra-high-definition video (resolution 3840×2160, frame rate 60fps), 3D laser scanning (point cloud density 100 points / mm², accuracy ±0.01mm), 192kHz high-fidelity audio (sampling bit depth 24bit, signal-to-noise ratio ≥110dB).

[0052] Analytic Hierarchy Process (AHP) Matrix Scaling: 1 - Equally important, 3 - Slightly important, 5 - Significantly important, 7 - Strongly important, 9 - Extremely important, 2 / 4 / 6 / 8 are intermediate values;

[0053] Semantic retrieval evaluation metrics: precision (P), recall (R), F1 score (F1 = 2PR / (P+R));

[0054] Value assessment parameters: the inheritance attenuation coefficient λ ranges from 0.005 to 0.02 (set according to the urgency of intangible cultural heritage inheritance), the application frequency weight μ ranges from 0.3 to 0.5, and the cultural transmission gain coefficient ν ranges from 0.2 to 0.4.

[0055] Example 1: Intelligent Management and Application of Suzhou Embroidery Intangible Cultural Heritage Resources

[0056] Step S1: Multimodal Suzhou Embroidery Data Acquisition and Standardization

[0057] Data collection subjects: representative Suzhou embroidery techniques (random stitch embroidery, flat stitch embroidery), inheritors (Zhang Moumou, national-level inheritor), and representative works ("Peony Painting").

[0058] Multimodal data acquisition:

[0059] Image data: 300 detailed images of Suzhou embroidery works (peony patterns, needlework textures) were taken using a 4K camera. The actual works and embroidery needles and thread tools were scanned using 3D laser scanning to generate point cloud data.

[0060] Audio data: Three audio clips, each 15 minutes long, were recorded using a 192kHz high-fidelity recording device, in which the inheritor explained the needlework techniques and material selection.

[0061] Text data: Structured text including the lineage of Suzhou embroidery, the process of random stitch embroidery (starting stitch - laying thread - pressing thread - finishing stitch), and parameters of embroidery thread materials (mulberry silk thread, thread diameter 0.12mm), totaling 5000 words;

[0062] Standardization processing: Extending Suzhou embroidery-specific fields based on the DC metadata framework, specifically including:

[0063] Technique type: Random stitch embroidery / Straight stitch embroidery;

[0064] Inheritor Level: National Level;

[0065] Material parameters: Mulberry silk thread, diameter 0.12mm;

[0066] Work steps: starting stitch / laying thread / pressing thread / ending stitch;

[0067] Geographical connection: Suzhou, Jiangsu Province;

[0068] 120 standardized data units are generated, each containing a unique number (e.g., SX-2024-001 to SX-2024-120).

[0069] Step S2: Weighted fusion processing of multimodal data

[0070] Single-modal feature vector extraction:

[0071] Text feature vector F t The TF-IDF algorithm is used to extract text keywords (such as "random stitch embroidery" and "silk thread"), generating a vector with a dimension of 512. Example: F t =[0.82,0.15,...,0.03];

[0072] Image feature vector Fᵢ: An improved gray-level co-occurrence matrix extraction method is used to divide the target area of ​​Suzhou embroidery pattern into blocks (8×8 pixel blocks). After denoising with a Gaussian smoothing filter (σ=1.2), four texture feature parameters such as contrast and correlation are calculated to generate a vector with a dimension of 256. Example: Fᵢ=[0.75,0.21,...,0.04];

[0073] Audio feature vector F aMel-frequency cepstral coefficients (MFCCs) are used for extraction, generating a 13-dimensional MFCC feature vector. Example: F a =[0.68,0.19,...,0.02];

[0074] Video feature vector Fᵥ: Extracts keyframes from the video (1 frame is extracted every 5 frames). Keyframe features are extracted using the same method as image features, generating a vector with a dimension of 256. Example: Fᵥ=[0.72,0.23,...,0.05];

[0075] Determining the fusion weights (based on the analytic hierarchy process):

[0076] Constructing a judgment matrix: As a traditional craft, Suzhou embroidery prioritizes both images (needlework, patterns) and text (processes, materials). The judgment matrix is ​​as follows:

[0077] Evaluation index F t (text) Fᵢ (image) F a (audio) Fᵥ (video) F t (text) 1 1 / 3 3 1 / 2 Fᵢ (image) 3 1 5 2 F a (audio) 1 / 3 1 / 5 1 1 / 4 Fᵥ (video) 2 1 / 2 4 1

[0078] Consistency test: The maximum eigenvalue λmax = 4.08, the consistency index CI = (λmax - n) / (n - 1) = 0.027, the random consistency index RI = 0.90, and the consistency ratio CR = CI / RI = 0.03 < 0.1, so the test is passed;

[0079] Weight calculation: The final fusion weights are α=0.2 (text), β=0.4 (image), γ=0.1 (audio), and δ=0.3 (video), which satisfy α+β+γ+δ=1;

[0080] Comprehensive eigenvector calculation (using the self-created formula F=α·F) t +β·Fᵢ+γ·F a +δ·Fᵥ):

[0081] Normalize each feature vector and then input it into the formula;

[0082] Example calculation:

[0083] F=0.2X[0.82,0.15,...0.03]+0.4X[0.75,0.21,...0.04]+0.1X[0.68,0.19,...0.02]+0.3X[0.72,0.23,...0.05];

[0084] The calculated value is F = [0.75, 0.20, ..., 0.04] (dimension 512), which comprehensively reflects the core features of Suzhou embroidery resources in terms of text, images, audio, and video.

[0085] Step S3: Blockchain-enhanced distributed storage

[0086] Generate numeric identifiers:

[0087] Generate a unique DOI identifier for each standardized data unit (e.g., 10.5281 / zenodo.SX-2024-001).

[0088] Constructing blockchain evidence storage nodes: Each node contains "data unit number + comprehensive feature vector F + collection timestamp (2024-05-2014:30:00) + inheritor's authorization signature (Zhang Moumou's electronic signature) + modification log (initial state: no modification)";

[0089] Tiered storage architecture: Sensitive data (such as exclusive color matching formulas for Suzhou embroidery) is stored on the Hyperledger Fabric permissioned blockchain, while public data (such as images of artworks) is stored on the Ethereum open blockchain. Data synchronization between the two chains is achieved through the cross-chain interoperability protocol (Polkadot), with a synchronization latency of ≤500ms.

[0090] Step S4: Intelligent retrieval with enhanced semantic association

[0091] Constructing a Suzhou embroidery knowledge graph: BERT pre-trained model (base version, 10,000 training steps) was used to extract entity labels and construct a semantic network containing 4,200 nodes. Node types include "Suzhou embroidery techniques", "inheritors", and "embroidery thread materials". Example association: Random stitch embroidery → Zhang Moumou (inheritor) → Mulberry silk thread (material).

[0092] Semantic relevance calculation (using the self-created formula R(x,y)=ω1·Sim(x,y)+ω2·Rel(x,y)·Depth(x,y)):

[0093] Parameter settings: ω1=0.3 (literal similarity weight), ω2=0.7 (semantic association weight), satisfying ω1+ω2=1;

[0094] Example calculation:

[0095] The search keyword is x = "random stitch embroidery material", the resource is y = "instructions for use of silk thread", Sim(x,y) = 0.6 (literal matching degree), Rel(x,y) = 0.9 (association strength between "random stitch embroidery" and "silk thread" in the knowledge graph), and Depth(x,y) = 2 (association path: random stitch embroidery → embroidery thread → silk thread). Therefore, R(x,y) = 0.3 × 0.6 + 0.7 × 0.9 × (1 / 2) = 0.18 + 0.315 = 0.495.

[0096] Search ranking: Search results are sorted in descending order by R(x,y) value, and the accuracy of the first 10 results is ≥92%.

[0097] Step S5: Resource Management and Dynamic Updates

[0098] Index weight optimization: Based on search feedback data (such as the frequency of users clicking on "random stitch embroidery process" being 60%), the index weight of the corresponding data unit will be increased by 20%;

[0099] Data completion and update: The missing data of the "random needle embroidery laying" process was completed using the ARIMA model (p=2, d=1, q=2). After the completed data was reviewed and confirmed by the inheritor Zhang, it was updated to the resource library, and the data integrity improved from 85% to 98%.

[0100] Steps S6-S7: Dynamic Value Assessment and Intelligent Push

[0101] Value assessment (using the self-created formula V=V0·(1-λ·T)+μ·U+ν·I):

[0102] Parameter settings: V0=0.9 (basic value, expert score 0.95 + big data statistics 0.85, weighted average), λ=0.01 (moderate urgency of Suzhou embroidery inheritance), T=1 year (digital storage duration), μ=0.4 (application frequency weight), U=0.8 (adapt to design scenarios), ν=0.3 (cultural dissemination gain).

[0103] Calculate: V = 0.9 × (1 - 0.01 × 1) + 0.4 × 0.8 + 0.3 × 0.9 = 0.891 + 0.32 + 0.27 = 1.481;

[0104] Intelligent push: Pushes high-definition images of Suzhou embroidery patterns, needlework videos and other resource interfaces to designer users, with a push accuracy rate of ≥88%.

[0105] Example 2: Intelligent Management and Application of Peking Opera Intangible Cultural Heritage Resources

[0106] The difference from Example 1 is as follows:

[0107] Data collection: The focus is on collecting audio of Peking Opera singing (192kHz), video of body movements (4K), and 3D scan data of facial makeup;

[0108] Weighting adjustment: Based on the analytic hierarchy process, as a folk performance-based intangible cultural heritage, audio (singing) and video (body movements) are more important for Peking Opera. The weights are determined as α=0.1 (text), β=0.2 (image), γ=0.4 (audio), and δ=0.3 (video).

[0109] Value assessment parameters: λ=0.008 (good foundation for Peking Opera inheritance), μ=0.35 (application frequency weight), ν=0.35 (cultural dissemination gain, wider audience);

[0110] Implementation results: Semantic retrieval accuracy ≥93%, dynamic push accuracy ≥90%, data integrity ≥97%.

[0111] Comparative Example 1: Single-Modal Data Management Method

[0112] Only Suzhou embroidery image data was collected, and traditional image retrieval methods were used without performing multimodal fusion and semantic association analysis.

[0113] Comparative Example 2: Fixed-weight multimodal fusion management method

[0114] Multimodal fusion was performed using fixed weights α=0.25, β=0.25, γ=0.25, and δ=0.25, with the remaining steps being the same as in Example 1.

[0115] Comparison of results:

[0116] Evaluation index Example 1 (embroidery) Example 2 (Peking opera) Comparative Example 1 Comparative Example 2 Search accuracy rate (%) 92 93 65 78 Data integrity (%) 98 97 72 85 Push accuracy rate (%) 88 90 52 70 Data storage security (no leakage rate %) 100 100 85 90

[0117] The formula in this invention has the following advantages:

[0118] Multimodal feature fusion formula F=α·F t +β·Fᵢ+γ·F a +δ·Fᵥ;

[0119] To address the issue of incomplete feature information in a single modality, multi-source data is fused using dynamic weighting to improve feature extraction accuracy;

[0120] F is the comprehensive feature vector, which is the core data foundation for subsequent retrieval, storage, and evaluation; F t / Fᵢ / F a / Fᵥ correspond to single-modal features of text, image, audio, and video, respectively, covering different dimensions of intangible cultural heritage resources; α / β / γ / δ are dynamic weights that adapt to the differences in feature importance among different types of intangible cultural heritage resources;

[0121] Compared to fixed weights (Comparative Example 2), dynamic weights improve search accuracy by 14-15 percentage points.

[0122] The semantic relevance calculation function is R(x,y)=ω1·Sim(x,y)+ω2·Rel(x,y)·Depth(x,y);

[0123] To address the "missed detection" problem of traditional literal retrieval, semantic association analysis based on knowledge graphs is used to improve the comprehensiveness and accuracy of retrieval.

[0124] Sim(x,y) ensures the accuracy of basic literal matching, while Rel(x,y) and Depth(x,y) together reflect the strength of semantic association (the shorter the path, the stronger the association, and the higher the weight); ω1 / ω2 balances the contributions of literal matching and semantic association.

[0125] Compared to single-word literal search (Comparative Example 1), the accuracy rate is improved by 27-28 percentage points.

[0126] The value assessment formula is V = V0·(1-λ·T) + μ·U + ν·I;

[0127] To achieve dynamic quantification of the value of intangible cultural heritage resources, providing a basis for precise targeting;

[0128] V0 safeguards the basic cultural value, (1-λ·T) reflects the characteristics of inheritance attenuation, and μ·U and ν·I respectively consider application needs and dissemination value;

[0129] Compared to push methods without dynamic evaluation (Comparative Example 1), the push accuracy rate is improved by 36-38 percentage points.

[0130] For sensitive data such as the exclusive color matching formula of Suzhou embroidery and the unpublished aria scores of Peking Opera, an access consent token based on SM2 encryption is generated. The token contains the visitor's identity identifier, access permission level, and validity period (24 hours). The validity of the token is verified by blockchain during access. Access is only allowed when the identity matches and the permissions are met. The test results show that the sensitive data leakage rate is 0 and the access response time is ≤300ms.

[0131] This invention solves the problems of incomplete data, inaccurate retrieval, imprecise push notifications, and lack of security in the existing digital management of intangible cultural heritage by integrating multimodal dynamic fusion, blockchain evidence storage, semantic retrieval, and dynamic evaluation. Verification by examples shows that the retrieval accuracy, data integrity, and push notification accuracy are all superior to existing technologies, and it can be widely applied to various digital protection and intelligent application scenarios of intangible cultural heritage resources.

[0132] This invention, through empirical research on two typical intangible cultural heritage projects—Suzhou embroidery (a traditional craft) and Peking Opera (a folk performance)—comprehensively verifies the scientific nature, adaptability, and efficiency of the intelligent management and application method for digital resources of intangible cultural heritage. In terms of implementation results, this invention significantly outperforms existing single-modal management methods and fixed-weight multimodal fusion methods in core technical indicators: the retrieval accuracy rates for Suzhou embroidery and Peking Opera reach 92% and 93% respectively, an improvement of 27-28 percentage points compared to single-modal methods; data integrity is improved to over 97%, addressing the pain point of missing intangible cultural heritage skill information; dynamic push accuracy is ≥88%, achieving precise matching of resource supply and user needs; and the sensitive data leakage rate remains 100%, balancing the core requirements of data security and cross-domain sharing, fully demonstrating the practical value of the technical solution of this invention.

[0133] The core innovation of this invention lies in constructing a fully intelligent closed-loop system encompassing data acquisition, fusion, storage, retrieval, management, evaluation, push notifications, and security. In the data acquisition stage, multimodal technologies such as 4K ultra-high-definition imaging and 3D laser scanning are used for collaborative acquisition. Combined with the DC metadata framework to expand intangible cultural heritage (ICH)-specific fields, this overcomes the limitations of traditional single-modal acquisition that leads to a lack of cultural connotation, achieving complete preservation of the multidimensional attributes of ICH resources. In the multimodal data fusion stage, the innovative introduction of the analytic hierarchy process (AHP) to dynamically adjust weights, along with a self-developed feature fusion formula, enables the accurate extraction of core features from different types of ICH resources, solving the industry pain point of poor adaptability in fixed-weight fusion. The blockchain-enhanced distributed storage, through a layered architecture of permissioned and open chains, not only achieves the confirmation and traceability of ICH data but also achieves efficient synchronization of heterogeneous data through cross-chain interoperability protocols, balancing the protection of sensitive resources with public... The system addresses the contradictions in resource sharing; its semantically enhanced intelligent retrieval, built upon a BERT pre-trained model to construct an intangible cultural heritage knowledge graph and combined with a self-developed semantic relevance calculation function, breaks through the limitations of traditional literal retrieval, significantly reducing the false negative rate and allowing users to quickly access semantically relevant resources; its dynamic resource management and update mechanism, through a closed-loop system of retrieval feedback to optimize index weights, ARIMA model to complete missing data, and inheritor verification and confirmation, ensures the timeliness and integrity of intangible cultural heritage resources and their technical lineage; and its dynamic value assessment and intelligent push, by quantifying the characteristics of resource inheritance decay, application needs, and dissemination value, combined with user profiling and LBS technology, achieves personalized services tailored to each individual, fully releasing the application potential of digital intangible cultural heritage resources.

[0134] The technological advantages of this invention are not only reflected in the innovation of its core algorithm and architecture design, but also in its strong practicality and scalability. Regarding the adaptation to different types of intangible cultural heritage, by dynamically adjusting the multimodal fusion weights, it can flexibly adapt to the characteristics of different types of intangible cultural heritage resources such as traditional crafts, folk performances, and oral literature, solving the problem of the one-size-fits-all adaptation defects of existing technologies. In terms of meeting user needs, it can accurately match the resource demands of different groups such as researchers, designers, and the general public, providing complete inheritance genealogy and craft process data for academic research, providing visual resources such as patterns and needlework for creative design, and providing easily understandable cultural symbols and performance excerpts for public dissemination. Regarding technical compatibility, the algorithms used, such as TF-IDF, MFCC, and BERT, all have good technical compatibility with blockchain cross-chain protocols, allowing seamless integration with existing intangible cultural heritage digitization platforms and reducing the cost of technology promotion.

[0135] From an industry value perspective, this invention not only provides a standardized and intelligent technical paradigm for the digital protection of intangible cultural heritage (ICH), but also promotes the transformation and upgrading of ICH protection from "passive storage" to "active application." By using digital means to achieve precise replication and dynamic transmission of ICH techniques, it provides a sustainable protection solution for endangered ICH projects. Through intelligent push notifications and quantitative assessment of cultural dissemination gain coefficients, it helps ICH resources integrate into modern life scenarios, promoting the deep integration of traditional culture with creative industries and academic research. The data security guarantee mechanism provides full life-cycle security protection for sensitive ICH resources (such as exclusive technique recipes and unpublished transmission materials), dispelling the concerns of inheritors about data sharing, and aligning with the core principle of digital protection: "transmission as the primary focus, dissemination as a secondary focus."

[0136] In the future, this invention can be further expanded to include more diverse applications, adapting to the digital management needs of various intangible cultural heritage projects (such as traditional medicine and folk literature). Furthermore, it can be combined with new technologies such as large-scale artificial intelligence models and digital twins to enhance the intelligent interaction and immersive experience of intangible cultural heritage resources. The widespread application of this invention will provide strong technical support for the digital protection and living transmission of my country's intangible cultural heritage, helping these national cultural treasures achieve sustainable development in the digital age and injecting new momentum into the building of cultural confidence and the innovative dissemination of traditional culture.

[0137] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A method for intelligent management and application of digital resources of intangible cultural heritage, characterized in that, Includes the following steps: S1. Multimodal Intangible Cultural Heritage Data Acquisition and Standardization: Intangible cultural heritage resource data is acquired collaboratively using 4K ultra-high-definition images, 3D laser scanning, 192kHz high-fidelity audio, and structured text. Based on the DC metadata framework, intangible cultural heritage-specific fields are expanded to generate standardized data units. S2. Multimodal Data Weighted Fusion Processing: A comprehensive feature vector is calculated using a self-developed multimodal feature fusion formula, as follows: F=α·F t +β·Fᵢ+γ·F a +δ·Fᵥ; Where F is the comprehensive feature vector, F t F i F a Fᵥ and Fᵥ are the feature vectors of text, image, audio and video, respectively, and α, β, γ and δ are the fusion weights, which satisfy α+β+γ+δ=1 and are dynamically adjusted according to the type of intangible cultural heritage resources through the analytic hierarchy process. S3. Blockchain-enhanced distributed storage: Generate a unique digital identifier for each standardized data unit, and construct a blockchain storage node for intangible cultural heritage resources by combining the comprehensive feature vector from step S2. Achieve data ownership confirmation and cross-domain sharing through a layered architecture of permissioned blockchain and open blockchain. S4. Intelligent Retrieval Enhanced with Semantic Relevance: A semantic index is constructed based on the intangible cultural heritage knowledge graph, and a self-developed semantic relevance calculation function is used to optimize the retrieval ranking. The function is as follows: R(x,y)=ω·Sim(x,y)+ω·Rel(x,y)·Depth(x,y) Where R(x,y) is the relevance between the search keyword x and the resource y, Sim(x,y) is the literal similarity, Rel(x,y) is the node association strength in the knowledge graph, Depth(x,y) is the association path depth, and ω and ω are weight coefficients. S5. Resource Management and Dynamic Updates: Based on the retrieval feedback data from step S4, dynamically optimize the index weight of data units, and regularly supplement and update resource metadata through interview data with successors to form a closed-loop management system.

2. The method according to claim 1, characterized in that, The intangible cultural heritage-specific fields mentioned in step S1 include technique process nodes, lineage of inheritors, material formula parameters, and regional cultural association attributes.

3. The method according to claim 1, characterized in that, In step S2, the image feature vector F is extracted through an improved gray-level co-occurrence matrix. After segmenting the target areas such as intangible cultural heritage patterns and craft tools, Gaussian smoothing filters are used for weighted noise reduction, and then texture feature parameters are calculated.

4. The method according to claim 1, characterized in that, In step S3, the blockchain evidence storage node includes data collection timestamps, inheritor authorization signatures, and data modification logs, and achieves heterogeneous data synchronization between different intangible cultural heritage databases through cross-chain interoperability protocols.

5. The method according to claim 1, characterized in that, In step S4, the intangible cultural heritage knowledge graph extracts intangible cultural heritage entity labels through the BERT pre-trained model and constructs a semantic association network containing more than 32,000 nodes. The node types include intangible cultural heritage projects, inheritors, techniques and processes, and cultural symbols.

6. A method for dynamic value assessment and application promotion of digital resources of intangible cultural heritage, characterized in that, The management method according to any one of claims 1-5 further includes the following steps: S6. Dynamic Assessment of Resource Value: The resource value coefficient is calculated using a self-developed formula for assessing the value of intangible cultural heritage digital resources. V = V·(1-λ·T) + μ·U + ν·I Where V is the dynamic value coefficient, V is the basic value, λ is the inheritance attenuation coefficient, T is the digital storage duration, μ is the application frequency weight, U is the application scenario adaptability, and ν is the cultural dissemination gain coefficient. S7. Intelligent Application Push: Based on user profile data and the value coefficient of step S6, customized intangible cultural heritage resource data interfaces are pushed to different users through collaborative filtering algorithms.

7. The method according to claim 6, characterized in that, In step S6, the basic value V is determined by combining expert scoring with big data statistics. The scoring dimensions include historical and cultural value, uniqueness of skills, and urgency of inheritance.

8. The method according to claim 6, characterized in that, The user profile data in step S7 includes browsing history, search keywords, and application scenario tags, and LBS technology is used to achieve precise delivery of regional intangible cultural heritage resources.

9. The method according to claim 1, characterized in that, During the dynamic update process in step S5, an autoregressive integral moving average model is used to complete the missing skill process data, ensuring the integrity of the intangible cultural heritage skill lineage.

10. The method according to claim 1, characterized in that, It also includes data security safeguards: adopting a consent-based access control mechanism for digital resources of intangible cultural heritage, generating access consent tokens, and allowing access to sensitive data only when the token is verified to be valid.

Images