Zero-Shot Relational Learning for Multimodal Knowledge Graphs
This addresses a key challenge in knowledge graph completion for AI applications by enabling inference on new relations without training data, though it is incremental as it builds on existing multimodal and relational learning methods.
The paper tackles the problem of zero-shot relational learning in multimodal knowledge graphs, where there is no training data for new relations, and proposes a novel framework that integrates multimodal information and graph structures, achieving superior performance on three benchmarks.
Relational learning is an essential task in the domain of knowledge representation, particularly in knowledge graph completion (KGC). While relational learning in traditional single-modal settings has been extensively studied, exploring it within a multimodal KGC context presents distinct challenges and opportunities. One of the major challenges is inference on newly discovered relations without any associated training data. This zero-shot relational learning scenario poses unique requirements for multimodal KGC, i.e., utilizing multimodality to facilitate relational learning.However, existing works fail to support the leverage of multimodal information and leave the problem unexplored. In this paper, we propose a novel end-to-end framework, consisting of three components, i.e., multimodal learner, structure consolidator, and relation embedding generator, to integrate diverse multimodal information and knowledge graph structures to facilitate the zero-shot relational learning. Evaluation results on three multimodal knowledge graphs demonstrate the superior performance of our proposed method.