Multi-modal entity alignment in hyperbolic space

Guo, Hao; Tang, Jiuyang; Zeng, Weixin; Zhao, Xiang; Liu, Li

doi:10.1016/j.neucom.2021.03.132

Cited by 39 publications

(21 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, HMEA (Guo et al. 2021 ) aligns entities with multiple forms by mapping multi-modal representations into hyperbolic space. Although many researchers have worked on multi-modal knowledge fusion, it is still a critical task.…”

Section: Technical Challengesmentioning

confidence: 99%

Knowledge Graphs: Opportunities and Challenges

et al. 2023

View full text Add to dashboard Cite

With the explosive growth of artificial intelligence (AI) and big data, it has become vitally important to organize and represent the enormous volume of knowledge appropriately. As graph data, knowledge graphs accumulate and convey knowledge of the real world. It has been well-recognized that knowledge graphs effectively represent complex information; hence, they rapidly gain the attention of academia and industry in recent years. Thus to develop a deeper understanding of knowledge graphs, this paper presents a systematic overview of this field. Specifically, we focus on the opportunities and challenges of knowledge graphs. We first review the opportunities of knowledge graphs in terms of two aspects: (1) AI systems built upon knowledge graphs; (2) potential application fields of knowledge graphs. Then, we thoroughly discuss severe technical challenges in this field, such as knowledge graph embeddings, knowledge acquisition, knowledge graph completion, knowledge fusion, and knowledge reasoning. We expect that this survey will shed new light on future research and the development of knowledge graphs.

show abstract

Section: Technical Challengesmentioning

confidence: 99%

Knowledge Graphs: Opportunities and Challenges

et al. 2023

View full text Add to dashboard Cite

show abstract

“…It integrated multiple representations of entities via common space learning. HMEA [6] adopted the hyperbolic graph convolutional networks (HGCNs) to learn structural and visual embeddings of entities separately, then merged them in the hyperbolic space by a weighted Mobius addition. EVA [7] employed GCNs [18] to learn structural representations for entities, and used feed-forward networks to learn embeddings from image, relation and attribute features, respectively.…”

Section: Related Workmentioning

confidence: 99%

“…Most existing models for EA leverage graph structures and/or side information of entities such as name and attributes along with KG embedding techniques to achieve alignment [1,2]. Several recent methods enrich entity representations by incorporating images, a natural component of entity profiles in many KGs such as DBpedia [3] and Wikidata [4], to address EA in a multimodal view [5][6][7].…”

Section: Introductionmentioning

confidence: 99%

Probing the Impacts of Visual Context in Multimodal Entity Alignment

Wang

Shi

Yang³

et al. 2023

Data Sci. Eng.

View full text Add to dashboard Cite

We study the problem of multimodal embedding-based entity alignment (EA) between different knowledge graphs. Recent works have attempted to incorporate images (visual context) to address EA in a multimodal view. While the benefits of multimodal information have been observed, its negative impacts are non-negligible as injecting images without constraints brings much noise. It also remains unknown under what circumstances or to what extent visual context is truly helpful to the task. In this work, we propose to learn entity representations from graph structures and visual context, and combine feature similarities to find alignments at the output level. On top of this, we explore a mechanism which utilizes classification techniques and entity types to remove potentially un-helpful images (visual noises) during alignment learning and inference. We conduct extensive experiments to examine this mechanism and provide thorough analysis about impacts of the visual modality on EA.

show abstract

“…Compared with traditional entity alignment methods, methods based on GCNs not only require relatively less human involvement in the process of feature construction, but also such methods can be extended to large knowledge graphs. Methods based on GCNs [6,[20][21][22][23][24][25][26][27][28][29][30][31] usually combine embedding methods to embed the data to be processed into a unified vector space, and the method based on embedding has been well applied and developed [32,33]. JAPE [6] sets two embedding modules, namely Structure Embedding (SE) and Attribute Embedding (AE), which jointly embed the structures of two knowledge graphs into a unified vector space, and then use the attribute correlation in the knowledge graph to further improvement.…”

Section: Entity Alignment Based On Graph Convolutional Networkmentioning

confidence: 99%

“…This paper uses the DBP15K dataset and uses Hits@k(k = 1, 10), a widely used metric in experiments [18,22]. A Hits@k score (higher is better) is computed by measuring the proportion of correctly aligned entities ranked in the top k list.…”

Section: Metricsmentioning

confidence: 99%

Cross-Language Entity Alignment for Joint Entity Screening and Dual Relation Graph

Zhang

Wang

2022

Preprint

View full text Add to dashboard Cite

In knowledge graphs of different languages, entity alignment is usuallyinterfered by problems such as structural heterogeneity and languagedifference. The emergence of these problems seriously affects the taskof entity alignment. Existing methods continuously improve alignmentstrategies, but often ignore the impact of triples with the same headand tail entity names on the entity alignment task. This paper takes thehead entity in the triple as the central entity, and proposes an entityalignment method that effectively utilizes neighbor entities with thesame entity name as the central entity. We take advantage of the multi-order properties of Graph Convolutional Neural Networks to processentities, relationships and attributes in the source and target knowl-edge graphs through knowledge structure embedding and entity nameembedding. We define the primal graph and the dual relation graphrespectively, the former is composed of two knowledge graphs together,and the latter is constructed by the association of relations in triplesto strengthen the role of relations. We use the primal graph and thedual relation graph to process the entities and relationships in knowledgegraphs. We compare the output vectors obtained by the two meth-ods, and then select the optimal output vector, combine the entityID to remove the neighbor entities that are completely equivalent tothe central entity, and finally obtain the final alignment result throughneighbor sampling and the similarity calculation of the correspondingentity. Experimental results on real datasets show that our methodsignificantly helps improve the task of cross-lingual entity alignment.

show abstract

Multi-modal entity alignment in hyperbolic space

Cited by 39 publications

References 18 publications

Knowledge Graphs: Opportunities and Challenges

Knowledge Graphs: Opportunities and Challenges

Probing the Impacts of Visual Context in Multimodal Entity Alignment

Cross-Language Entity Alignment for Joint Entity Screening and Dual Relation Graph

Contact Info

Product

Resources

About