A Dual-Task Learning Model for Temporal Knowledge Graph Entity Alignment
Jingwei Cheng, Fu Zhang, Xihao Wang
Lecture notes in computer science
Problems Identified (4)
Temporal KG entity alignment: The research problem is to discover equivalent entity pairs across different temporal knowledge graphs for knowledge fusion and integration.
Temporal-relational embedding interference: Including both temporal and relational information in entity embeddings may cause interference and limit temporal-aware entity alignment performance.
Temporal KG entity alignment: The research problem is to discover equivalent entity pairs across different temporal knowledge graphs for knowledge fusion and integration.
Temporal-relational embedding interference: Including both temporal and relational information in entity embeddings may cause interference and limit temporal-aware entity alignment performance.
Proposed Solutions (5)
DTTEA dual-task learning: The paper proposes DTTEA, a dual-task learning model with separate relational and temporal tasks for temporal knowledge graph entity alignment.
Dynamic-loss task embedding optimization: The model updates embeddings for both relational and temporal tasks by optimizing a dynamic loss function.
Dual-task similarity alignment: The model generates coarse-grained and fine-grained similarity matrices using dual-task entity embeddings and a temporal overlap matrix for entity alignment.
DTTEA dual-task learning: The paper proposes DTTEA, a dual-task learning model with separate relational and temporal tasks for temporal knowledge graph entity alignment.
Dynamic-loss task embedding optimization: The model updates embeddings for both relational and temporal tasks by optimizing a dynamic loss function.
Results (3)
Significant SOTA outperformance:
Code availability:
Significant SOTA outperformance:
Research Domain
Temporal knowledge graph entity alignment