Recipe-Controlled Decoder Audit for Structural Knowledge-Graph Completion
Researchers are exploring new methods for knowledge graph completion (KGC) and exploration. One paper proposes a unified taxonomy for post-hoc explanations in KGC to improve reproducibility and evaluation. Another introduces Model Graph Inductive Learning (MGIL) to capture global graph structures for more accurate link prediction. Additionally, a study identifies the 'Initial Exploration Problem' (IEP) that hinders lay users from interacting with unfamiliar knowledge graphs, suggesting new interface designs. Finally, TRACE-KG is presented as a framework for context-enriched knowledge graph generation that bypasses predefined schemas. AI
- CompGCN
- Recipe-Controlled Decoder Audit
- DistMult
- UMLS
- Knowledge-Graph Completion
- WN18RR
- YAGO3-10
- RotatE
- Kinship
- arXiv
- Claire McNamara
- graph neural network
- knowledge graph
- Hits@K
- Mgilane
- Hossein Hajiabolhassan
- Knowledge Graph Completions
- Alessandro Lonardi
- mean reciprocal rank
- TRACE-KG
- Mohammad Sadeq Abolhasani