非遗异质图弱连边

... 非遗工坊展示。每个顾客群体都能形成一个细分市场，不同消费群体的需求存在显著的异质性[38]。通过分层运营，展会实现了对全年龄段客群的有效覆盖，避免了传统展会 ...

本文聚焦于内蒙古国家级非物质文化遗产及特色景点，旨在深入探讨知识图谱在研学实践中的应用价值及其初步实现路径。具体而言，通过综合运用数据采集、整理，知识表示及可视化 ...

目的：本文旨在构建历史文化知识图谱，以解决多源异构历史数据的整合与复杂语义关系建模问题，研究以地方志历史活动数据为实证案例，系统探索了领域知识图谱的构建路径。

本文在大数据分析的基础上,提出了利用TF-IDF [2]动态加权模型进行特征提取和时空融合的知识图谱,精准发现濒危非遗技艺. ... 文化遗产数据价值生成.在实现路径上,依托 ...

格兰诺维特的“弱关系优势”理论在此具有启示意义：非遗电商化需构建开放型传承共同体，通过多元主体的协同创新，弥合技艺传承与市场需求间的鸿沟。 徽州墨模雕刻的实践表明， ...

· 图卷积神经网络(GCN)：它可以对图结构的结构信息和节点的属性信息同时学习，共同得到最终的节点特征表示并将节点之间的结构关联性也考虑进模型。 · 时空动态网络(STDN)：其 ...

人工构建的符号网络可通过无标度网络与随机网络组合连接形成，本文采用同配连接和异配连接两种连接方式，生成不同属性的网络模型。随后，通过调整网络的平均负度，深入探究同配 ...

在本研究的图限制博弈框架中，节点的边际贡献取决于其加入后是否实质性改变联盟的可连通结构。对大湾区跨境数据协作而言，城市间连边存在显著异质性，部分内地城市间连通紧密， ...

The rapid development of large language models (LLMs) has provided significant support and opportunities for the advancement of domain-specific LLMs. However, fine-tuning these large models using Intangible Cultural Heritage (ICH) data inevitably faces challenges such as bias, incorrect knowledge inheritance, and catastrophic forgetting. To address these issues, we propose a novel training method that integrates a bidirectional chains of thought and a reward mechanism. This method is built upon ICH-Qwen, a large language model specifically designed for the field of intangible cultural heritage. The proposed method enables the model to not only perform forward reasoning but also enhances the accuracy of the generated answers by utilizing reverse questioning and reverse reasoning to activate the model's latent knowledge. Additionally, a reward mechanism is introduced during training to optimize the decision-making process. This mechanism improves the quality of the model's outputs through structural and content evaluations with different weighting schemes. We conduct comparative experiments on ICH-Qwen, with results demonstrating that our method outperforms 0-shot, step-by-step reasoning, knowledge distillation, and question augmentation methods in terms of accuracy, Bleu-4, and Rouge-L scores on the question-answering task. Furthermore, the paper highlights the effectiveness of combining the bidirectional chains of thought and reward mechanism through ablation experiments. In addition, a series of generalizability experiments are conducted, with results showing that the proposed method yields improvements on various domain-specific datasets and advanced models in areas such as Finance, Wikidata, and StrategyQA. This demonstrates that the method is adaptable to multiple domains and provides a valuable approach for model training in future applications across diverse fields.

The intangible cultural heritage (ICH) of China, a cultural asset transmitted across generations by various ethnic groups, serves as a significant testament to the evolution of human civilization and holds irreplaceable value for the preservation of historical lineage and the enhancement of cultural self-confidence. However, the rapid pace of modernization poses formidable challenges to ICH, including threats damage, disappearance and discontinuity of inheritance. China has the highest number of items on the UNESCO Intangible Cultural Heritage List, which is indicative of the nation's abundant cultural resources and emphasises the pressing need for ICH preservation. In recent years, the rapid advancements in large language modelling have provided a novel technological approach for the preservation and dissemination of ICH. This study utilises a substantial corpus of open-source Chinese ICH data to develop a large language model, ICH-Qwen, for the ICH domain. The model employs natural language understanding and knowledge reasoning capabilities of large language models, augmented with synthetic data and fine-tuning techniques. The experimental results demonstrate the efficacy of ICH-Qwen in executing tasks specific to the ICH domain. It is anticipated that the model will provide intelligent solutions for the protection, inheritance and dissemination of intangible cultural heritage, as well as new theoretical and practical references for the sustainable development of intangible cultural heritage. Furthermore, it is expected that the study will open up new paths for digital humanities research.

In the last decades the rapid development of technologies and methodologies in the field of digitization and 3D modelling has led to an increasing proliferation of 3D technologies in the Cultural Heritage domain. Despite the great potential of 3D digital heritage, the "special effects" of 3D may often overwhelm its importance in research. Projects and consortia of scholars have tried to put order in the different fields of application of these technologies, providing guidelines and proposing workflows. The use of computer graphics as an effective methodology for CH research and communication highlighted the need of transparent provenance data to properly document digital assets and understand the degree of scientific quality and reliability of their outcomes. The building and release of provenance knowledge, consisting in the complete formal documentation of each phase of the process, is therefore of fundamental importance to ensure its repeatability and to guarantee the integration and interoperability of the generated metadata on the Semantic Web. This paper proposes a methodology for documenting the planning and creation of 3D models used in archaeology and Cultural Heritage, by means of an application profile based on the CIDOC CRM ecosystem and other international standards.

This chapter presents the potential of interoperability and standardised data publication for cultural heritage resources, with a focus on community-driven approaches and web standards for usability. The Linked Open Usable Data (LOUD) design principles, which rely on JSON-LD as lingua franca, serve as the foundation. We begin by exploring the significant advances made by the International Image Interoperability Framework (IIIF) in promoting interoperability for image-based resources. The principles and practices of IIIF have paved the way for Linked Art, which expands the use of linked data by demonstrating how it can easily facilitate the integration and sharing of semantic cultural heritage data across portals and institutions. To provide a practical demonstration of the concepts discussed, the chapter highlights the implementation of LUX, the Yale Collections Discovery platform. LUX serves as a compelling case study for the use of linked data at scale, demonstrating the real-world application of automated enrichment in the cultural heritage domain. Rooted in empirical study, the analysis presented in this chapter delves into the broader context of community practices and semantic interoperability. By examining the collaborative efforts and integration of diverse cultural heritage resources, the research sheds light on the potential benefits and challenges associated with LOUD.

We propose a novel graph clustering method guided by additional information on the underlying structure of the clusters (or communities). The problem is formulated as the matching of a graph to a template with smaller dimension, hence matching $n$ vertices of the observed graph (to be clustered) to the $k$ vertices of a template graph, using its edges as support information, and relaxed on the set of orthonormal matrices in order to find a $k$ dimensional embedding. With relevant priors that encode the density of the clusters and their relationships, our method outperforms classical methods, especially for challenging cases.

The classification of Intangible Cultural Heritage (ICH) images in the Mekong Delta poses unique challenges due to limited annotated data, high visual similarity among classes, and domain heterogeneity. In such low-resource settings, conventional deep learning models often suffer from high variance or overfit to spurious correlations, leading to poor generalization. To address these limitations, we propose a robust framework that integrates the hybrid CoAtNet architecture with model soups, a lightweight weight-space ensembling technique that averages checkpoints from a single training trajectory without increasing inference cost. CoAtNet captures both local and global patterns through stage-wise fusion of convolution and self-attention. We apply two ensembling strategies - greedy and uniform soup - to selectively combine diverse checkpoints into a final model. Beyond performance improvements, we analyze the ensembling effect through the lens of bias-variance decomposition. Our findings show that model soups reduces variance by stabilizing predictions across diverse model snapshots, while introducing minimal additional bias. Furthermore, using cross-entropy-based distance metrics and Multidimensional Scaling (MDS), we show that model soups selects geometrically diverse checkpoints, unlike Soft Voting, which blends redundant models centered in output space. Evaluated on the ICH-17 dataset (7,406 images across 17 classes), our approach achieves state-of-the-art results with 72.36% top-1 accuracy and 69.28% macro F1-score, outperforming strong baselines including ResNet-50, DenseNet-121, and ViT. These results underscore that diversity-aware checkpoint averaging provides a principled and efficient way to reduce variance and enhance generalization in culturally rich, data-scarce classification tasks.

Performance artforms like Peking opera face transmission challenges due to the extensive passive listening required to understand their nuance. To create engaging forms of experiencing auditory Intangible Cultural Heritage (ICH), we designed a spatial interaction-based segmented-audio (SISA) Virtual Reality system that transforms passive ICH experiences into active ones. We undertook: (1) a co-design workshop with seven stakeholders to establish design requirements, (2) prototyping with five participants to validate design elements, and (3) user testing with 16 participants exploring Peking Opera. We designed transformations of temporal music into spatial interactions by cutting sounds into short audio segments, applying t-SNE algorithm to cluster audio segments spatially. Users navigate through these sounds by their similarity in audio property. Analysis revealed two distinct interaction patterns (Progressive and Adaptive), and demonstrated SISA's efficacy in facilitating active auditory ICH engagement. Our work illuminates the design process for enriching traditional performance artform using spatially-tuned forms of listening.

In this article, we explain the recent advance of subsampling methods in knowledge graph embedding (KGE) starting from the original one used in word2vec.

Knowledge bases, and their representations in the form of knowledge graphs (KGs), are naturally incomplete. Since scientific and industrial applications have extensively adopted them, there is a high demand for solutions that complete their information. Several recent works tackle this challenge by learning embeddings for entities and relations, then employing them to predict new relations among the entities. Despite their aggrandizement, most of those methods focus only on the local neighbors of a relation to learn the embeddings. As a result, they may fail to capture the KGs' context information by neglecting long-term dependencies and the propagation of entities' semantics. In this manuscript, we propose ÆMP (Attention-based Embeddings from Multiple Patterns), a novel model for learning contextualized representations by: (i) acquiring entities' context information through an attention-enhanced message-passing scheme, which captures the entities' local semantics while focusing on different aspects of their neighborhood; and (ii) capturing the semantic context, by leveraging the paths and their relationships between entities. Our empirical findings draw insights into how attention mechanisms can improve entities' context representation and how combining entities and semantic path contexts improves the general representation of entities and the relation predictions. Experimental results on several large and small knowledge graph benchmarks show that ÆMP either outperforms or competes with state-of-the-art relation prediction methods.

The rise of graph-structured data has driven major advances in Graph Machine Learning (GML), where graph embeddings (GEs) map features from Knowledge Graphs (KGs) into vector spaces, enabling tasks like node classification and link prediction. However, since GEs are derived from explicit topology and features, they may miss crucial implicit knowledge hidden in seemingly sparse datasets, affecting graph structure and their representation. We propose a GML pipeline that integrates a Knowledge Completion (KC) phase to uncover latent dataset semantics before embedding generation. Focusing on transitive relations, we model hidden connections with decay-based inference functions, reshaping graph topology, with consequences on embedding dynamics and aggregation processes in GraphSAGE and Node2Vec. Experiments show that our GML pipeline significantly alters the embedding space geometry, demonstrating that its introduction is not just a simple enrichment but a transformative step that redefines graph representation quality.

High-level automation is increasingly critical in AI, driven by rapid advances in large language models (LLMs) and AI agents. However, LLMs, despite their general reasoning power, struggle significantly in specialized, data-sensitive tasks such as designing Graph Neural Networks (GNNs). This difficulty arises from (1) the inherent knowledge gaps in modeling the intricate, varying relationships between graph properties and suitable architectures and (2) the external noise from misleading descriptive inputs, often resulting in generic or even misleading model suggestions. Achieving proficiency in designing data-aware models -- defined as the meta-level capability to systematically accumulate, interpret, and apply data-specific design knowledge -- remains challenging for existing automated approaches, due to their inefficient construction and application of meta-knowledge. To achieve meta-level proficiency, we propose DesiGNN, a knowledge-centered framework that systematically converts past model design experience into structured, fine-grained knowledge priors well-suited for meta-learning with LLMs. To account for the inherent variability and external noise, DesiGNN aligns empirical property filtering from extensive benchmarks with adaptive elicitation of literature insights via LLMs. By constructing a solid meta-knowledge between unseen graph understanding and known effective architecture patterns, DesiGNN can deliver top-5.77% initial model proposals for unseen datasets within seconds and achieve consistently superior performance with minimal search cost compared to baselines.

Knowledge graph (KG) embedding aims at learning the latent representations for entities and relations of a KG in continuous vector spaces. An empirical observation is that the head (tail) entities connected by the same relation often share similar semantic attributes -- specifically, they often belong to the same category -- no matter how far away they are from each other in the KG; that is, they share global semantic similarities. However, many existing methods derive KG embeddings based on the local information, which fail to effectively capture such global semantic similarities among entities. To address this challenge, we propose a novel approach, which introduces a set of virtual nodes called \textit{\textbf{relational prototype entities}} to represent the prototypes of the head and tail entities connected by the same relations. By enforcing the entities' embeddings close to their associated prototypes' embeddings, our approach can effectively encourage the global semantic similarities of entities -- that can be far away in the KG -- connected by the same relation. Experiments on the entity alignment and KG completion tasks demonstrate that our approach significantly outperforms recent state-of-the-arts.

Pervasive socio-technical networks bring new conceptual and technological challenges to developers and users alike. A central research theme is evaluation of the intensity of relations linking users and how they facilitate communication and the spread of information. These aspects of human relationships have been studied extensively in the social sciences under the framework of the "strength of weak ties" theory proposed by Mark Granovetter.13 Some research has considered whether that theory can be extended to online social networks like Facebook, suggesting interaction data can be used to predict the strength of ties. The approaches being used require handling user-generated data that is often not publicly available due to privacy concerns. Here, we propose an alternative definition of weak and strong ties that requires knowledge of only the topology of the social network (such as who is a friend of whom on Facebook), relying on the fact that online social networks, or OSNs, tend to fragment into communities. We thus suggest classifying as weak ties those edges linking individuals belonging to different communities and strong ties as those connecting users in the same community. We tested this definition on a large network representing part of the Facebook social graph and studied how weak and strong ties affect the information-diffusion process. Our findings suggest individuals in OSNs self-organize to create well-connected communities, while weak ties yield cohesion and optimize the coverage of information spread.

As a social media, online social networks play a vital role in the social information diffusion. However, due to its unique complexity, the mechanism of the diffusion in online social networks is different from the ones in other types of networks and remains unclear to us. Meanwhile, few works have been done to reveal the coupled dynamics of both the structure and the diffusion of online social networks. To this end, in this paper, we propose a model to investigate how the structure is coupled with the diffusion in online social networks from the view of weak ties. Through numerical experiments on large-scale online social networks, we find that in contrast to some previous research results, selecting weak ties preferentially to republish cannot make the information diffuse quickly, while random selection can achieve this goal. However, when we remove the weak ties gradually, the coverage of the information will drop sharply even in the case of random selection. We also give a reasonable explanation for this by extra analysis and experiments. Finally, we conclude that weak ties play a subtle role in the information diffusion in online social networks. On one hand, they act as bridges to connect isolated local communities together and break through the local trapping of the information. On the other hand, selecting them as preferential paths to republish cannot help the information spread further in the network. As a result, weak ties might be of use in the control of the virus spread and the private information diffusion in real-world applications.

Relation ties, defined as the correlation and mutual exclusion between different relations, are critical for distant supervised relation extraction. Existing approaches model this property by greedily learning local dependencies. However, they are essentially limited by failing to capture the global topology structure of relation ties. As a result, they may easily fall into a locally optimal solution. To solve this problem, in this paper, we propose a novel force-directed graph based relation extraction model to comprehensively learn relation ties. Specifically, we first build a graph according to the global co-occurrence of relations. Then, we borrow the idea of Coulomb's Law from physics and introduce the concept of attractive force and repulsive force to this graph to learn correlation and mutual exclusion between relations. Finally, the obtained relation representations are applied as an inter-dependent relation classifier. Experimental results on a large scale benchmark dataset demonstrate that our model is capable of modeling global relation ties and significantly outperforms other baselines. Furthermore, the proposed force-directed graph can be used as a module to augment existing relation extraction systems and improve their performance.

In this paper we study cluster synchronization in networks of oscillators with heterogenous Kuramoto dynamics, where multiple groups of oscillators with identical phases coexist in a connected network. Cluster synchronization is at the basis of several biological and technological processes; yet the underlying mechanisms to enable cluster synchronization of Kuramoto oscillators have remained elusive. In this paper we derive quantitative conditions on the network weights, cluster configuration, and oscillators' natural frequency that ensure asymptotic stability of the cluster synchronization manifold; that is, the ability to recover the desired cluster synchronization configuration following a perturbation of the oscillators' states. Qualitatively, our results show that cluster synchronization is stable when the intra-cluster coupling is sufficiently stronger than the inter-cluster coupling, the natural frequencies of the oscillators in distinct clusters are sufficiently different, or, in the case of two clusters, when the intra-cluster dynamics is homogeneous. We illustrate and validate the effectiveness of our theoretical results via numerical studies.

The structure of a network dramatically affects the spreading phenomena unfolding upon it. The contact distribution of the nodes has long been recognized as the key ingredient in influencing the outbreak events. However, limited knowledge is currently available on the role of the weight of the edges on the persistence of a pathogen. At the same time, recent works showed a strong influence of temporal network dynamics on disease spreading. In this work we provide an analytical understanding, corroborated by numerical simulations, about the conditions for infected stable state in weighted networks. In particular, we reveal the role of heterogeneity of edge weights and of the dynamic assignment of weights on the ties in the network in driving the spread of the epidemic. In this context we show that when weights are dynamically assigned to ties in the network an heterogeneous distribution is able to hamper the diffusion of the disease, contrary to what happens when weights are fixed in time.

Granovetter's weak tie theory of social networks is built around two central hypotheses. The first states that strong social ties carry the large majority of interaction events; the second maintains that weak social ties, although less active, are often relevant for the exchange of especially important information (e.g., about potential new jobs in Granovetter's work). While several empirical studies have provided support for the first hypothesis, the second has been the object of far less scrutiny. A possible reason is that it involves notions relative to the nature and importance of the information that are hard to quantify and measure, especially in large scale studies. Here, we search for empirical validation of both Granovetter's hypotheses. We find clear empirical support for the first. We also provide empirical evidence and a quantitative interpretation for the second. We show that attention, measured as the fraction of interactions devoted to a particular social connection, is high on weak ties --- possibly reflecting the postulated informational purposes of such ties --- but also on very strong ties. Data from online social media and mobile communication reveal network-dependent mixtures of these two effects on the basis of a platform's typical usage. Our results establish a clear relationships between attention, importance, and strength of social links, and could lead to improved algorithms to prioritize social media content.

Plenty of algorithms for link prediction have been proposed and were applied to various real networks. Among these works, the weights of links are rarely taken into account. In this paper, we use local similarity indices to estimate the likelihood of the existence of links in weighted networks, including Common Neighbor, Adamic-Adar Index, Resource Allocation Index, and their weighted versions. In both the unweighted and weighted cases, the resource allocation index performs the best. To our surprise, the weighted indices perform worse, which reminds us of the well-known Weak Tie Theory. Further extensive experimental study shows that the weak ties play a significant role in the link prediction problem, and to emphasize the contribution of weak ties can remarkably enhance the predicting accuracy.

Weak ties play a significant role in the structures and the dynamics of community networks. Based on the susceptible-infected model in contact process, we study numerically how weak ties influence the predictability of epidemic dynamics. We first investigate the effects of different kinds of weak ties on the variabilities of both the arrival time and the prevalence of disease, and find that the bridgeness with small degree can enhance the predictability of epidemic spreading. Once weak ties are settled, compared with the variability of arrival time, the variability of prevalence displays a diametrically opposed changing trend with both the distance of the initial seed to the bridgeness and the degree of the initial seed. More specifically, the further distance and the larger degree of the initial seed can induce the better predictability of arrival time and the worse predictability of prevalence. Moreover, we discuss the effects of weak tie number on the epidemic variability. As community strength becomes very strong, which is caused by the decrease of weak tie number, the epidemic variability will change dramatically. Compared with the case of hub seed and random seed, the bridgenss seed can result in the worst predictability of arrival time and the best predictability of prevalence. These results show that the variability of arrival time always marks a complete reversal trend of that of prevalence, which implies it is impossible to predict epidemic spreading in the early stage of outbreaks accurately.

Heterogeneous information networks (HINs) are widely applied to recommendation systems due to their capability of modeling various auxiliary information with meta-paths. However, existing HIN-based recommendation models usually fuse the information from various meta-paths by simple weighted sum or concatenation, which limits performance improvement because it lacks the capability of interest compositions among meta-paths. In this article, we propose an HIN-based Interest Composition model for Recommendation (HicRec). Specifically, user and item representations are learned with a graph neural network on both the graph structure and features in each meta-path, and a parameter sharing mechanism is utilized here to ensure that the user and item representations are in the same latent space. Then, users' interests in each item from each pair of related meta-paths are calculated by a combination of the user and item representations. The composed user interests are obtained by their single interest from both intra- and inter-meta-paths for recommendation. Extensive experiments are conducted on three real-world datasets and the results demonstrate that our proposed HicRec model outperforms the baselines.

One can point to a variety of historical milestones for gender equality in STEM (science, technology, engineering, and mathematics), however, practical effects are incremental and ongoing. It is important to quantify gender differences in subdomains of scientific work in order to detect potential biases and monitor progress. In this work, we study the relevance of gender in scientific collaboration patterns in the Institute for Operations Research and the Management Sciences (INFORMS), a professional society with sixteen peer-reviewed journals. Using their publication data from 1952 to 2016, we constructed a large temporal bipartite network between authors and publications, and augmented the author nodes with gender labels. We characterized differences in several basic statistics of this network over time, highlighting how they have changed with respect to relevant historical events. We find a steady increase in participation by women (e.g., fraction of authorships by women and of new women authors) starting around 1980. However, women still comprise less than 25% of the INFORMS society and an even smaller fraction of authors with many publications. Moreover, we describe a methodology for quantifying the structural role of an authorship with respect to the overall connectivity of the network, using it to measure subtle differences between authorships by women and by men. Specifically, as measures of structural importance of an authorship, we use effective resistance and contraction importance, two measures related to diffusion throughout a network. As a null model, we propose a degree-preserving temporal and geometric network model with emergent communities. Our results suggest the presence of systematic differences between the collaboration patterns of men and women that cannot be explained by only local statistics.

We investigate some of the properties and extensions of a dynamic innovation network model recently introduced in \citep{koenig07:_effic_stabil_dynam_innov_networ}. In the model, the set of efficient graphs ranges, depending on the cost for maintaining a link, from the complete graph to the (quasi-) star, varying within a well defined class of graphs. However, the interplay between dynamics on the nodes and topology of the network leads to equilibrium networks which are typically not efficient and are characterized, as observed in empirical studies of R&D networks, by sparseness, presence of clusters and heterogeneity of degree. In this paper, we analyze the relation between the growth rate of the knowledge stock of the agents from R&D collaborations and the properties of the adjacency matrix associated with the network of collaborations. By means of computer simulations we further investigate how the equilibrium network is affected by increasing the evaluation time $τ$ over which agents evaluate whether to maintain a link or not. We show that only if $τ$ is long enough, efficient networks can be obtained by the selfish link formation process of agents, otherwise the equilibrium network is inefficient. This work should assist in building a theoretical framework of R&D networks from which policies can be derived that aim at fostering efficient innovation networks.

Quantum walks on translation invariant regular graphs spread quadratically faster than their classical counterparts. The same coherence that gives them this quantum speedup inhibits, or even stops their spread in the presence of disorder. We ask how to create an efficient transport channel from a fixed source site (A) to fixed target site (B) in a disordered 2-dimensional discrete-time quantum walk by cutting some of the links. We show that the somewhat counterintuitive strategy of cutting links along a single line connecting A to B creates such a channel. The efficient transport along the cut is due to topologically protected chiral edge states, which exist even though the bulk Chern number in this system vanishes. We give a realization of the walk as a periodically driven lattice Hamiltonian, and identify the bulk topological invariant responsible for the edge states as the quasienergy winding of this Hamiltonian.

Let $G$ be a graph and $\mathcal {S}$ be a subset of $Z$. A vertex-coloring $\mathcal {S}$-edge-weighting of $G$ is an assignment of weight $s$ by the elements of $\mathcal {S}$ to each edge of $G$ so that adjacent vertices have different sums of incident edges weights. It was proved that every 3-connected bipartite graph admits a vertex-coloring $\{1,2\}$-edge-weighting (Lu, Yu and Zhang, (2011) \cite{LYZ}). In this paper, we show that the following result: if a 3-edge-connected bipartite graph $G$ with minimum degree $δ$ contains a vertex $u\in V(G)$ such that $d_G(u)=δ$ and $G-u$ is connected, then $G$ admits a vertex-coloring $\mathcal {S}$-edge-weighting for $\mathcal {S}\in \{\{0,1\},\{1,2\}\}$. In particular, we show that every 2-connected and 3-edge-connected bipartite graph admits a vertex-coloring $\mathcal {S}$-edge-weighting for $\mathcal {S}\in \{\{0,1\},\{1,2\}\}$. The bound is sharp, since there exists a family of infinite bipartite graphs which are 2-connected and do not admit vertex-coloring $\{1,2\}$-edge-weightings or vertex-coloring $\{0,1\}$-edge-weightings.

We present a novel edge-level ego-network encoding for learning on graphs that can boost Message Passing Graph Neural Networks (MP-GNNs) by providing additional node and edge features or extending message-passing formats. The proposed encoding is sufficient to distinguish Strongly Regular Graphs, a family of challenging 3-WL equivalent graphs. We show theoretically that such encoding is more expressive than node-based sub-graph MP-GNNs. In an empirical evaluation on four benchmarks with 10 graph datasets, our results match or improve previous baselines on expressivity, graph classification, graph regression, and proximity tasks -- while reducing memory usage by 18.1x in certain real-world settings.

Herzog-Hibi-Hreindóttir-Kahle-Rauh introduced the class of closed graph and they proved that the binomial edge ideal $J(G)$ of a graph $G$ has quadratic Gröbner bases if $G$ is closed. In this paper, we introduce the class of weakly closed graph as a generalization of the closed graph and prove that the quotient ring $S/J(G)$ is $F$-pure if $G$ is weakly closed. This fact is a generalization of Ohtani's theorem.

A known failing of many popular random graph models is that the Aldous-Hoover Theorem guarantees these graphs are dense with probability one; that is, the number of edges grows quadratically with the number of nodes. This behavior is considered unrealistic in observed graphs. We define a notion of edge exchangeability for random graphs in contrast to the established notion of infinite exchangeability for random graphs --- which has traditionally relied on exchangeability of nodes (rather than edges) in a graph. We show that, unlike node exchangeability, edge exchangeability encompasses models that are known to provide a projective sequence of random graphs that circumvent the Aldous-Hoover Theorem and exhibit sparsity, i.e., sub-quadratic growth of the number of edges with the number of nodes. We show how edge-exchangeability of graphs relates naturally to existing notions of exchangeability from clustering (a.k.a. partitions) and other familiar combinatorial structures.

In order to have a compact visualization of the order type of a given point set S, we are interested in geometric graphs on S with few edges that unambiguously display the order type of S. We introduce the concept of exit edges, which prevent the order type from changing under continuous motion of vertices. That is, in the geometric graph on S whose edges are the exit edges, in order to change the order type of S, at least one vertex needs to move across an exit edge. Exit edges have a natural dual characterization, which allows us to efficiently compute them and to bound their number.

We define the \emph{visual complexity} of a plane graph drawing to be the number of basic geometric objects needed to represent all its edges. In particular, one object may represent multiple edges (e.g., one needs only one line segment to draw a path with an arbitrary number of edges). Let $n$ denote the number of vertices of a graph. We show that trees can be drawn with $3n/4$ straight-line segments on a polynomial grid, and with $n/2$ straight-line segments on a quasi-polynomial grid. Further, we present an algorithm for drawing planar 3-trees with $(8n-17)/3$ segments on an $O(n)\times O(n^2)$ grid. This algorithm can also be used with a small modification to draw maximal outerplanar graphs with $3n/2$ edges on an $O(n)\times O(n^2)$ grid. We also study the problem of drawing maximal planar graphs with circular arcs and provide an algorithm to draw such graphs using only $(5n - 11)/3$ arcs. This is significantly smaller than the lower bound of $2n$ for line segments for a nontrivial graph class.

Let $G$ be a finite $d$-regular graph with a proper edge coloring. An edge Kempe switch is a new proper edge coloring of $G$ obtained by switching the two colors along some bi-chromatic cycle. We prove that any other edge coloring can be obtained by performing finitely many edge Kempe switches, provided that $G$ is replaced with a suitable finite covering graph. The required covering degree is bounded above by a constant depending only on $d$.

Digital platforms for intangible cultural heritage (ICH) function as vibrant electronic marketplaces, yet they grapple with content overload, high search costs, and under-leveraged social networks of heritage custodians. To address these electronic-commerce challenges, we present GCHS, a custodian-aware, graph-based deep learning model that enhances ICH recommendation by uniting three critical signals: custodians’ social relationships, user interest profiles, and content metadata. Leveraging an attention mechanism, GCHS dynamically prioritizes influential custodians and resharing behaviors to streamline user discovery and engagement. We first characterize ICH-specific propagation patterns, e.g., custodians’ social influence, heterogeneous user interests, and content co-consumption and then encode these factors within a collaborative graph framework. Evaluation on a real-world ICH dataset demonstrates that GCHS delivers improvements in Top-N recommendation accuracy over leading benchmarks and significantly outperforms in terms of next-N sequence prediction. By integrating social, cultural, and transactional dimensions, our approach not only drives more effective digital commerce interactions around heritage content but also supports sustainable cultural dissemination and stakeholder participation. This work advances electronic-commerce research by illustrating how graph-based deep learning can optimize content discovery, personalize user experience, and reinforce community networks in digital heritage ecosystems.

Given that traditional graph structures make it difficult to capture complex interaction information between entities, this study adopts a hypergraph model to represent multimodal and heterogeneous data, to adapt to the complexity of dragon and lion dance movements. This study proposes a new method based on a hypergraph convolutional network (HGCN) for the inheritance and teaching action evaluation of the intangible cultural heritage of the dragon and lion dance. This method constructs the HGCN model, combined with a self-attention mechanism to accurately evaluate action details and promote its inheritance in the digital age. The results show that the HGCN algorithm incorporating the attention mechanism exhibits excellent performance, the accuracy achieves 0.941, the error rate reduces to 0.333, an evaluation efficiency improved by 400%, and user satisfaction increases to 0.900. These results not only validate the efficiency and accuracy of the model but also demonstrate its potential to improve the teaching and inheritance efficiency of the dragon and lion dance. This study not only provides a new technological means for the digitization of intangible cultural heritage in sports but also opens up new paths for the modern teaching and inheritance of traditional sports projects.

The fine-grained mining and construction of semantic associations within multimodal intangible cultural heritage (ICH) resources are crucial for deepening our understanding of their knowledge content and ensuring their systematic protection and transmission in the digital and intelligent era. This paper addresses the urgent need for the digital preservation and transmission of ICH resources. Following a review of current research on Qingyang sachets and ICH, the study introduces an ontology-based approach to constructing a semantic description model for the multimodal digital resources related to Qingyang sachets. By acquiring and processing multimodal resources concerning the craftsmanship and associated customs of Qingyang sachets, the study reorganizes the corresponding textual and visual knowledge. Utilizing knowledge graphs, the research explores multidimensional pathways for delivering knowledge services related to the multimodal digital resources of Qingyang sachets. Empirical research confirms the applicability and feasibility of the proposed semantic association scheme for multimodal ICH digital resources. The findings provide valuable insights for multidimensional organization and integration across scenarios, time periods, and resources within the ICH domain, offering a reference for digital solutions aimed at the systematic protection of ICH.

Purpose This research aims to provide a methodology for constructing a knowledge graph (KG) to map the evolution of Chinese embroidery by associating cultural space and critical incidents within the context of intangible cultural heritage (ICH). Design/methodology/approach This study deconstructs the cultural space for ICH into physical, spiritual and societal dimensions and scrutinizes the structural components governing the evolution of embroidery projects. Utilizing examples from the four famous embroideries and four ethnic minority embroidery styles of China, we construct a KG connecting cultural space and critical incidents, enabling the storage, retrieval and discovery of knowledge about the evolution of embroideries. Findings The societal dimension of the embroidery cultural space contains more diverse information than its physical and spiritual aspects. Critical events in the evolution of embroidery are categorized into 15 types and further distinguished as internal and external stimuli. Internal stimuli reflect the gradual, positive transformation of ICH embroidery driven by developmental needs. External stimuli often induce changes that are not always beneficial but directly impact the evolution of embroidery. Originality/value This research offers fresh insights into ICH evolution by linking cultural space and critical incidents and enabling the exploration of implicit, potentially valuable and ultimately comprehensible information through knowledge visualization.

This paper addresses the challenges of knowledge acquisition and inaccurate relationship construction in constructing an intangible cultural heritage craft knowledge graph. The paper proposes an expert system-based model for constructing an intangible cultural heritage craft knowledge graph and design an incremental weighted rule-based inference algorithm for this knowledge. This model integrates an expert system to acquire intangible cultural heritage craft knowledge from multiple sources. Through a core algorithm, it implements incremental knowledge inference and dynamic knowledge graph construction. Experimental simulations using a specific intangible cultural heritage craft as the research object show that the proposed model achieves an accuracy of 89.6%, a recall of 87.3%, and an F1 value of 88.4%, respectively, representing improvements of 12.5%, 10.8%, and 11.6% compared to traditional methods. For 1,000 pieces of incremental knowledge, the inference time is only 2.3 seconds, significantly shorter than the 8.7 seconds of traditional batch inference algorithms. Research demonstrates that this model can effectively improve the quality and efficiency of constructing an intangible cultural heritage craft knowledge graph, providing strong support for the preservation and inheritance of intangible cultural heritage.

Abstract With the digitization of intangible cultural heritage (ICH), a large number of ICH digital resources have been created and accumulated. In this paper, BERT-CNN-BiLSTM-CRF information recognition model is proposed for obtaining metadata of ICH digital resources. Then a two-stage mapping approach is utilized to construct the knowledge graph of ICH digital resources. That is, metadata mapping to construct knowledge ontology, followed by mapping to knowledge graph through knowledge ontology. After the model performance test and knowledge graph construction, it can be seen that the spatial distribution of national-level ICH in China is mainly concentrated in the east and west regions. The F1 value of the BERT-CNN-BiLSTM-CRF model is 0.922, which is a better performance for the basic information extraction task compared with other models. The knowledge graph visualizes 7 types of entity nodes of ICH projects, digital resources, organizations, things, people, places, and time, which promotes the inheritance of ICH and knowledge sharing.

The rapid development of science and technology is reshaping the ecological pattern of cultural and tourism integration. As a cutting-edge digital technology, knowledge graph shows unique potential in the integration of culture and tourism. Focusing on Inner Mongolia’s national intangible cultural heritage and characteristic scenic spots, this paper aims to deeply explore the application value of knowledge graph in research and study practice and its preliminary realization path. Specifically, by comprehensively using advanced technologies such as data collection and collation, knowledge

A deep learning based named entity recognition method is proposed. Although BiLSTM has a good effect in named entity recognition task, it cannot focus on the local context near the target word in both forward and backward propagation. To address this issue, we propose the utilization of a convolutional neural network, which limits the receptive field through the volume base and encodes the character twice, so that the machine pays more attention to detail information when understanding the text, compared with the BiLSTM method, the F1 value of the proposed model is increased from 0.875 to 0.902, an increase of 0.027. A text classification method based on large pre-training model was proposed for text cleaning, compared with BERTBiLSTM, the F1 value of the proposed model in the Chinese nonposthumous text classification task is increased from 0.85 to 0.87, an increase of 0.02. After extracting information through named entity recognition method, the Chinese intangible cultural heritage knowledge graph is constructed.

No abstract available

No abstract available

161-174

Intangible Cultural Heritage (ICH) witnesses human creativity and wisdom in long histories, composed of a variety of immaterial manifestations. The rapid development of digital technologies accelerates the record of ICH, generating a sheer number of heterogenous data but in a state of fragmentation. To resolve that, existing studies mainly adopt approaches of knowledge graphs (KGs) which can provide rich knowledge representation. However, most KGs are text-based and text-derived, and incapable to give related images and empower downstream multimodal tasks, which is also unbeneficial for the public to establish the visual perception and comprehend ICH completely especially when they do not have the related ICH knowledge. Hence, aimed at that, we propose to, taking the Chinese nation-level ICH list as an example, construct a large-scale and comprehensive Multimodal Knowledge Graph (CICHMKG) combining text and image entities from multiple data sources and give a practical construction framework. Additionally, in this paper, to select representative images for ICH entities, we propose a method composed of the denoising algorithm (CNIFA) and a series of criteria, utilizing global and local visual features of images and textual features of captions. Extensive empirical experiments demonstrate its effectiveness. Lastly, we construct the CICHMKG, consisting of 1,774,005 triples, and visualize it to facilitate the interactions and help the public dive into ICH deeply.

By constructing a knowledge graph based on traditional medical intangible cultural heritage, comprehensive organization and storage of traditional medical intangible cultural heritage can be achieved, thereby achieving the protection of traditional medical culture. Firstly, Python's web crawler technology was used to obtain relevant data on traditional medical intangible cultural heritage, and it was cleaned and organized. Subsequently, using knowledge graph technology, thematic extraction, fusion, and correlation analysis were performed on the data to construct a knowledge graph centered on traditional medicine. In the process of constructing the knowledge graph, ontology modeling and semantic network technology were used to model and represent traditional pharmaceutical related concepts and relationships, achieving structured storage of the knowledge graph. At the same time, using visualization technology, the knowledge graph is presented in a graphical form, enabling the intuitive presentation of the knowledge and value of traditional medical intangible cultural heritage. Ultimately, a knowledge graph based on traditional medical intangible cultural heritage is obtained, which covers a wealth of traditional medical knowledge and related resources, and can help users gain a more comprehensive understanding and learning of traditional medical culture. By storing and visualizing the knowledge of traditional medical intangible cultural heritage in a structured manner, it provides important support and foundation for the protection and inheritance of traditional medicine.

The combination of Artificial Intelligence, Mixed Reality and other high-tech technologies with intangible cultural heritage provides innovative ideas for its inheritance and protection. However, in most intangible cultural heritage virtual reality systems, interaction is more important than knowledge, which brings many challenges to the carding and interactive experience of intangible cultural heritage. Therefore, taking the intangible cultural heritage of Hangluo as an example, we combined knowledge graphs and virtual reality to analyze and mine the data of Hangluo and constructed a relatively complete knowledge graph of Hangluo's cultural resources. And the knowledge graphs were integrated into the virtual reality system to form a virtual interactive system of intangible cultural heritage with strong interaction and an emphasis on knowledge. The system has four modules: providing textile introduction display, textile pattern display, tools and craft display, and knowledge graph visualization. Finally, good test results were collected through the “Millennium Hangluo” experience class, which verified that the virtual interactive system for intangible cultural heritage based on knowledge graph had played a positive role in promoting the inheritance of intangible cultural heritage.

The construction and effective application of an intangible cultural heritage knowledge graph (ICH KG) can realize the knowledge integration, and optimize the ICH knowledge management. However, high complexity and low computational efficiency of ICH KG make its application face challenges. We propose a multi-source knowledge graph embedding (KGE) model named ICHMKGE to convert the ICH KG into the vector representations to improve the computational efficiency of ICH KG and promote the digital sustainable development of ICH. Firstly, we take the Chinese ICH project 24 solar terms as an example and combine multiple official data sources to construct the Chinese ICH KG as a basis for this study. Secondly, as the ICH project is being further explored with the limited coverage of ICH knowledge, entity sparsity poses a serious challenge for ICH KGE. This paper employs the BERT model to encode the complete description information of entities, and establishes connection between triples entities and ontology concepts via cross-view modeling. Finally, the proposed ICHMKGE model is compared with the baseline models, and the experimental results demonstrate that the model exhibits superiority.

No abstract available

No abstract available

With the continuous development of computer algorithms, the analysis of the dynamic Atlas of intangible cultural heritage relies on computers and algorithms to build a systematic framework, which has become an important direction of the research on the living inheritance of intangible cultural heritage. Therefore, an analysis model combining improved spatio-temporal knowledge map and improved multimodal alignment is proposed. The model improves the correlation accuracy of intangible cultural heritage multimodal information through multi-component deep fusion and dynamic optimization. Compared with the three comparison models, the model's cross modal retrieval accuracy rates are 0.95 and 0.94, and the dynamic event matching rates are 0.96 and 0.97, which are excellent in the intangible cultural heritage scene. The model improves the performance of the dynamic map analysis of intangible cultural heritage, and provides a scheme for the accurate interpretation of intangible cultural heritage.

Abstract The development of digital technology promotes the construction of the Intangible cultural heritage (ICH) database but the data is still unorganized and not linked well, which makes the public hard to master the overall knowledge of the ICH. An ICH knowledge graph (KG) can help the public to understand the ICH and facilitate the protection of the ICH. However, a general framework of ICH KG construction is lacking now. In this study, we take the Chinese ICH (nation-level) as an example and propose a framework to build a Chinese ICH KG combining multiple data sources from Baike and the official website, which can extend the scale of the KG. Besides, the data of ICH grows daily, requiring us to design an efficient model to extract the knowledge from the data to update the KG in time. The built KG is based on the triple 〈entity, attribute, attribute value〉 and we introduce the attribute value extraction (AVE) task. However, the public Chinese ICH annotated AVE corpus is lacking. To solve that, we construct a Chinese ICH AVE corpus based on the Distant Supervision (DS) automatically rather than employing traditional manual annotation. Currently, AVE is usually seen as the sequence tagging task. In this paper, we take the ICH AVE as a node classification task and propose an AVE model BGC, combining the BiLSTM and graph attention network, which can fuse and utilize the word-level and character-level information by means of the ICH lexicon generated from the KG. We conduct extensive experiments and compare the proposed model with other state-of-the-art models. Experimental results show that the proposed model is of superiority.

No abstract available

Intangible cultural heritage preservation has been made progressively difficult owing to scattered documentation, loss of indigenous knowledge, and restricted accessibility for younger generations. Current solutions lean towards static digital collections or low-fidelity 3D reconstructions, which do not accurately record contextual interactions and immediate immersive exposures. Conventional solutions like rudimentary multimedia repositories and conventional VR visualization models do not have semantic connectivity, adaptive knowledge migration, and intelligent understanding of cultural context. To solve these problems, a Hybrid Graph Convolutional Network with Transformer-based Attention (HGCT-A) is designed, combining VR technology with graph-based relational learning and multi-head attention to promote cultural knowledge connection, interaction realism, and immersive learning. Experimental testing proves that the HGCT-A model has 97.8% accuracy and 96.4 % F1-score, outperforming traditional models. The comparative evaluation reiterates the superiority of HGCT-A in semantic preservation, dynamic adaptation, and user involvement.

Intangible Cultural Heritage (ICH) represents the customers, behaviors, and expressions that have been handed down through the generations, including folklore, performing arts, rituals, languages, and craftsmanship. Unlike tangible heritage, including objects and monuments, ICH is vulnerable to destruction due to globalization, urbanization, and generational gaps. Traditional methods of preservation, such as manual documentation and oral transmission, struggle to adapt to the digital era, leading to the loss of valuable cultural knowledge. Artificial Intelligence (AI) has become a revolutionary instrument for ICH preservation, offering solutions for digital preservation, restoration, and interactive dissemination. To address these issues, this research proposes a Galactic Swarm Optimized Graph Neural Network (GSO-GNN) for ICH’s improved inheritance and protection. The proposed approach begins with comprehensive ICH preservation of data. A pre-processing phase, including data cleaning and noise reduction, ensures the accuracy and consistency of the dataset and then feature extraction was conducted using Principal Component Analysis (PCA). The GNN is then employed to map complex relationships between cultural elements, preserving their contextual significance. To enhance efficiency, GSO is integrated to refine feature selection, improve model accuracy, and optimize computational resources. The findings show that the proposed method outperforms other techniques in terms of the [Formula: see text]1-score (98%), recall (98.48%), accuracy (99%), and precision (98%). Furthermore, AI-driven dissemination significantly enhances accessibility and engagement for global audiences, contributing to cultural sustainability and protection. By efficiently modeling complex cultural relationships and optimizing computational performance, the GSO-GNN platform offers a practical and scalable way to preserve digital heritage.

As an intangible cultural heritage of China, Cloisonne embodies rich historical and cultural significance alongside unique artistic value. However, rapid modernization has led to a growing disconnection in traditional crafts, posing challenges for both preservation and innovation. This study explores innovative visual image generation methods using generative AI technologies—specifically Stable Diffusion and Midjourney—to design and create cultural product images inspired by Cloisonne. Employing literature review, design practice and expert evaluation, the research covers generative AI tools, concepts of intangible cultural heritage and cultural goods, and new approaches to digitalization and design practice. Experimentally, a Cloisonne knowledge graph was constructed as a theoretical basis, followed by building and training a LoRA model. Stable Diffusion was then used to generate visual images of Cloisonne cultural products, which were further refined with Midjourney. Results demonstrate that these AI tools effectively produce creative and artistic cultural product visuals, revitalizing Cloisonne with fresh vitality and innovation, while advancing the digital preservation and modernization of intangible cultural heritage.

The integration of cultural preservation into smart city initiatives has become increasingly vital as urban systems seek to balance technological advancement with heritage sustainability. This paper presents a computer vision-based framework for recognizing and digitally modeling the intricate skill motions involved in intangible cultural heritage (ICH), such as paper cutting, calligraphy, and embroidery. The system combines pose estimation, action segmentation, and graph-based temporal modeling to capture fine-grained spatial-temporal patterns of craft demonstrations. A dedicated ICH dataset, recorded from authentic workshops, is used to train a hybrid neural network architecture combining 2D CNNs with temporal graph convolutional networks. The recognized motion sequences are reconstructed into digital avatars for use in virtual exhibitions, educational platforms, and cultural simulation systems within smart city infrastructures. Experimental results demonstrate significant improvements over baseline models in recognition accuracy and motion fidelity. This framework provides a foundation for the intelligent integration of ICH into digital urban services, supporting long-term cultural dissemination in future smart societies.

Digital preservation and transmission of intangible cultural heritage (ICH) are important in preserving cultural identity in the face of modernization. Virtual Reality (VR) technology offers an effective medium for experiential, interactive presentation of ICH components with new forms of cultural engagement, particularly among younger audiences. Current systems rely heavily on text or 2D media, which have no experiential depth and emotional impact, resulting in less interested audiences and incomplete cultural comprehension. Traditional methods-i.e., photographic documentation, oral recording, and 2 D archiving-cannot capture the performative and spatial nature of ICH practice. The model adopted is based on a VR-based reconstruction system complemented by the Spatio-Temporal Attention Graph Convolutional Network (STAGCN) algorithm that allows realistic simulation of rituals, crafts, and folk arts. This model guarantees dynamic interaction, high-fidelity visualization, and contextual learning. Comparative performance against baseline CNN and LSTM models demonstrates improved performance of STAGCN model of accuracy at 99.2 %, F1-score at 98.5 %, and pointing towards improved feature representation and cultural authenticity. The excellence comes in the form of spatialtemporal consistency and adaptive attention learning, performing significantly better than static models for user immersion, accuracy, and effectiveness in retaining culture.

No abstract available

The cultural heritage of the Liao dynasty in Chifeng encompasses significant historical and cultural information that requires systematic digital preservation and management. However, heterogeneous data sources across museums, archives, and research institutions lack semantic interoperability, creating barriers for cross-system integration and knowledge discovery. This study proposes a standardized knowledge graph construction method by integrating the CIDOC Conceptual Reference Model version 7.2 with large language models. A unified ontology framework enables semantic consistency across diverse heritage data, while Generative Pre-trained Transformer-based models automatically extract structured triples from unstructured texts through prompt engineering and entity disambiguation, with the resulting knowledge graph implemented in Neo4j graph database. The constructed knowledge graph integrates 106 immovable cultural heritage records from Chifeng City with approximately 20 types of semantic relationships, forming a comprehensive semantic network covering people, places, events, time, and materials. K-means clustering reveals five cultural value themes, including “Nomadic Imperial Power System” and “Multi-Capital Governance Network”, while geospatial mapping identifies a “dual-core and ring-belt” distribution pattern for heritage protection zoning. This research demonstrates how international semantic standards can be integrated with artificial intelligence technologies to enable interoperable cultural heritage knowledge systems, providing practical implications for cross-institutional heritage management and archaeological survey planning.

Abstract With the acceleration of modernisation and changes in the surrounding environment and human factors, the survival of some traditional ethnic sport’s intangible heritage is threatened, and the related protection work needs to be improved urgently, and digital technology provides new technical means and ways for it. In this paper, we construct a knowledge map of traditional sport’s intangible cultural heritage of ethnic minorities based on a graph database, use Cypher query language to achieve the visual presentation of traditional sport’s intangible knowledge and conduct visual knowledge discovery of intangible knowledge from various aspects as well as the use of semantic similarity to calculate the relationship between the non-heritage knowledge and the mining of the law of evolution and then realise the digital protection of non-heritage. Knowledge graph design is carried out for 37 traditional minority sports in X city, and the results show that the visual display of traditional sports NRL knowledge based on a knowledge graph has the feasibility of promoting the protection and inheritance of traditional sports.

Abstract Intangible cultural heritage (ICH) is a precious historical and cultural resource of a country. Protection and inheritance of ICH is important to the sustainable development of national culture. There are many different intangible cultural heritage items in China. With the development of information technology, ICH database resources were built by government departments or public cultural services institutions, but most databases were widely dispersed. Certain traditional database systems are disadvantageous to storage, management and analysis of massive data. At the same time, a large quantity of data has been produced, accompanied by digital intangible cultural heritage development. The public is unable to grasp key knowledge quickly because of the massive and fragmented nature of the data. To solve these problems, we proposed the intangible cultural heritage knowledge graph to assist knowledge management and provide a service to the public. ICH domain ontology was defined with the help of intangible cultural heritage experts and knowledge engineers to regulate the concept, attribute and relationship of ICH knowledge. In this study, massive ICH data were obtained, and domain knowledge was extracted from ICH text data using the Natural Language Processing (NLP) technology. A knowledge base based on domain ontology and instances for Chinese intangible cultural heritage was constructed, and the knowledge graph was developed. The pattern and characteristics behind the intangible cultural heritage were presented based on the ICH knowledge graph. The knowledge graph for ICH could foster support for organization, management and protection of the intangible cultural heritage knowledge. The public can also obtain the ICH knowledge quickly and discover the linked knowledge. The knowledge graph is helpful for the protection and inheritance of intangible cultural heritage.

Macau’s intangible cultural heritage (ICH) exemplifies a unique Sino-Western cultural fusion, wherein the interplay of Eastern and Western traditions complicates conventional analysis of heritage complexity and resilience. To address this challenge, we introduce FusionNet, a multimodal AI framework integrating image-based classification, an attention mechanism, identity embedding, and knowledge graph modeling for context-aware analysis of ICH. FusionNet combines image-based deep learning with an attention mechanism to focus on salient visual features in heritage imagery. This integrated architecture enables a holistic understanding of heritage elements and their adaptability to changing cultural contexts. Applied to Macau’s ICH, FusionNet reveals patterns of cultural resilience, illustrating how traditional practices persist and evolve amid centuries of EastWest influences. Our findings demonstrate the efficacy of fusing visual and knowledge-based modalities for heritage analysis, offering a robust approach for studying and preserving intangible cultural heritage in complex cultural environments. To elucidate how Macau’s intangible cultural heritage (ICH) exhibits “cultural resilience” and the mechanisms of identity (re)construction amid Sino‑Portuguese cultural interweaving; and to propose a computable multimodal framework (FusionNet + cultural‑identity embeddings + knowledge graph) that quantifies and validates these mechanisms. Materials include digital archives and historical texts (e.g., Macau Memory), social‑media text (Weibo plus ~1,000 English TripAdvisor/blog reviews), open heritage images, and structured knowledge bases (China ICH database). Methods comprise an attention‑based image classifier (FusionNet), LDA topic modeling (5‑fold cross‑validation selecting k = 3; mean coherence ≈ 0.59, compared with BERTopic), bilingual sentiment analysis, knowledge‑graph embedding and link prediction (evaluated with MRR, Hits@10), and t‑SNE visualization with clustering (three clusters; average silhouette ≈ 0.47). All implementations are in Python. LDA reveals three stable themes: (A) Chinese traditions (~45%), (B) Lusophone heritage (~30%), and (C) hybrid/local identity (~25%; e.g., Patuá and Macanese cuisine). Sentiment analysis indicates >70% positive evaluations, with ~12–15% negative. On the image side, most categories achieve diagonal accuracy >0.80, with some true‑positive rates reaching 0.95–1.00; Sino‑Portuguese architecture shows interpretable confusion. Knowledge‑graph embeddings and t‑SNE place the “hybrid/local identity” between the Chinese and Portuguese clusters, acting as a bridge (silhouette ≈ 0.47). Overall, multimodal fusion is more robust than multiple baselines on recognition and semantic association tasks, revealing a resilience pathway in which Macau ICH preserves core practices while continually absorbing exogenous elements. The proposed multimodal, knowledge‑driven framework effectively quantifies and explains identity (re)construction and cultural resilience in Macau’s ICH within a Sino‑Portuguese milieu; the “hybrid/local identity” is the key bridging mechanism. Future work can expand cross‑platform data, enhance cross‑modal alignment and knowledge reasoning, and generalize the approach to other multicultural contexts to strengthen external validity.

Named entity recognition (NER) is essential for structuring knowledge in the field of intangible cultural heritage (ICH), supporting applications such as knowledge graph construction and cultural research. However, the lack of annotated datasets for ICH has limited progress in this area. To address this, we present a Chinese dataset specifically designed for NER tasks in the ICH domain, covering key entity categories such as heritage items, inheritors, and material. Additionally, we propose an NER model that integrates RoBERTa for feature representation, the Kolmogorov-Arnold Network (KAN) for extracting complex entity patterns, and a conditional random field (CRF) for sequence labeling. Experimental results demonstrate the model’s effectiveness in capturing the intricate semantic dependencies in ICH texts. The dataset and model contribute to improving entity recognition in the ICH domain, facilitating the structured representation of cultural heritage knowledge. This work provides a valuable resource for further research in information extraction, digital preservation, and cultural heritage studies.

Abstract The digitization of intangible cultural heritage provides the basis for its protection and inheritance. In this paper, we extracted the classes and interrelationships corresponding to each element for the content of Yi Intangible Cultural Heritage, constructed the data model and accomplished the relationship abstraction of different classes. Based on the structure of different data classes, the correlation between spatial change and non-heritage status is explored. The data model is proposed to cluster non-heritage data based on SWC-WMD distance, which improves the similarity calculation based on WMD distance. The association analysis of non-heritage item classes was performed using Yi folk dance as an example, and the overall non-heritage data was analyzed by clustering. In Yi folk dance, the association formed a network correlation graph with N2, N3, and N12 as the three main centers, in which the dance class items associated with N2, N3, and N12 were 9, 9, and 8, respectively. Mining out the relationships among NH items helps to better the overall NH protection.

Compared to traditional dance, intangible cultural heritage dance often involves the isotropic extension of choreographic actions, utilizing both upper and lower limbs. This characteristic choreography style makes the remote joints lack interaction, consequently reducing accuracy in existing human motion prediction methods. Therefore, we propose a human motion prediction method based on the multi-scale hypergraph convolutional network of the intangible cultural heritage dance video. Firstly, this method inputs the 3D human posture sequence from intangible cultural heritage dance videos. The hypergraph is designed according to the synergistic relationship of the human joints in the intangible cultural heritage dance video, which is used to represent the spatial correlation of the 3D human posture. Then, a multi-scale hypergraph convolutional network is constructed, utilizing multi-scale transformation operators to segment the human skeleton into different scales. This network adopts a graph structure to represent the 3D human posture at different scales, which is then used by the single-scalar fusion operator to spatial features in the 3D human posture sequence are extracted by fusing the feature information of the hypergraph and the multi-scale graph. Finally, the Temporal Graph Transformer network is introduced to capture the temporal dependence among adjacent frames within the time domain. This facilitates the extraction of temporal features from the 3D human posture sequence, ultimately enabling the prediction of future 3D human posture sequences. Experiments show that we achieve the best performance in both short-term and long-term human motion prediction when compared to Motion-Mixer and Motion-Attention algorithms on Human3.6M and 3DPW datasets. In addition, ablation experiments show that our method can predict more precise 3D human pose sequences, even in the presence of isotropic extensions of upper and lower limbs in intangible cultural heritage dance videos. This approach effectively addresses the issue of missing segments in intangible cultural heritage dance videos.