Impact Statement:Multimodal knowledge graph completion is a key research area in the era of big data. However, existing methods often neglect to maintain consistency between different mod...Show More
Abstract:
Multimodal knowledge graph completion (MKGC) is a popular research topic in recent years. However, existing methods rarely consider the alignment of different entity moda...Show MoreMetadata
Impact Statement:
Multimodal knowledge graph completion is a key research area in the era of big data. However, existing methods often neglect to maintain consistency between different modalities during entity encoding, and they also fail to construct relation representations from a semantic perspective. These two issues limit the effectiveness of multimodal knowledge graph completion models. This paper proposes a novel multimodal knowledge graph completion model that effectively addresses these two challenges. We validated the model on three publicly available datasets, and the results show that the proposed model significantly improves the performance of multimodal knowledge graph completion and outperforms all existing methods.
Abstract:
Multimodal knowledge graph completion (MKGC) is a popular research topic in recent years. However, existing methods rarely consider the alignment of different entity modalities in the process of multimodal fusion, and often lack sufficient attention to the semantic information conveyed by relations, thus resulting in unsatisfactory completion performance. To address these two issues, we propose a new multimodal knowledge graph completion model called C2RS. This model first designs a cross-modal consistency contrastive learning task to align different entity modalities for accurate entity representation. Then, C2RS develops a relation semantic encoding module based on the distributions of knowledge graph triples to extract the semantic information of relations for comprehensive relation representation. Finally, we encode the candidate triples with a triple encoder, and identify the correct entities through a scoring function to complete the multimodal knowledge graph. According to the extensive experiments on three public MKGC datasets, C2RS obviously outperforms the baseline methods. The code of C2RS is available at https://github.com/ADMIS-TONGJI/C2RS.
Published in: IEEE Transactions on Artificial Intelligence ( Early Access )
