.png){kind=icon}
.png){kind=icon}
Return
Multi-modal knowledge graph entity alignment: a comprehensive survey
.png){kind=icon}
PRE
AI
DOI：10.1007/s00778-026-00986-4.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Multi-modal Knowledge Graph Entity Alignment plays a pivotal role in integrating heterogeneous knowledge from diverse sources. This paper presents a systematic review centered on a fine-grained modular framework. We deconstruct the pipeline of this task into four critical phases: uni-modal encoding, cross-modal interaction, multi-modal fusion, and optimization. A distinguishing contribution of this work is the unification of mathematical notations and definitions across diverse literature, enabling a barrier-free theoretical comparison of mechanisms without the need to consult original texts. Furthermore, we provide a comprehensive analysis of mainstream datasets and performance metrics, identifying the strengths and limitations of current approaches. Finally, we explore open challenges and emerging trends. We hope this survey serves as a foundational reference for the community, offering clear guidance to foster future innovations in bridging the semantic gaps between multi-modal knowledge representations. To promote reproducibility and facilitate future research, we have established an open-source repository that aggregates mainstream MMEA datasets and the source code of representative methods.
Keywords:
Multi-modal knowledge graph
Entity alignment
Cross-modal information interaction
Multi-modal information fusion