Research map / Graph learning / Knowledge and temporal graphs
Path Ranking with Attention to Type Hierarchies
Weiyu Liu, Angel Daruna, Zsolt Kira, Sonia Chernova
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add attention-based type selection over entity type hierarchies | Path-RNN with LogSumExp pooling · WN18RR | MAP%: 50.82 -> 74.51; 51.08 -> 74.51; 67.16 -> 74.51 | Improved |
| Add attention-based type selection over entity type hierarchies | Path-RNN with LogSumExp pooling · FB15k-237 | MAP%: 45.64 -> 54.09; 51.23 -> 54.09; 52.17 -> 54.09 | Improved |
| Use attention pooling over path representations instead of Max, Average, or LogSumExp pooling (ablation) | APR · WN18RR | MAP%: 73.23/80.30/74.51 -> 84.91 | Improved |
| Use attention pooling over path representations instead of Max, Average, or LogSumExp pooling (ablation) | APR · FB15k-237 | MAP%: 52.40/54.72/54.09 -> 57.35 | Improved |
| Learn entity abstraction levels with attention instead of fixing them to the most abstract level (ablation) | APR D · WN18RR | MAP%: 81.17 -> 84.91 | Improved |
| Learn entity abstraction levels with attention instead of fixing them to the most abstract level (ablation) | APR D · FB15k-237 | MAP%: 49.71 -> 57.35 | Improved |
| Learn entity abstraction levels with attention instead of fixing them to the most specific level (ablation) | APR D · WN18RR | MAP%: 82.66 -> 84.91 | Improved |
| Learn entity abstraction levels with attention instead of fixing them to the most specific level (ablation) | APR D · FB15k-237 | MAP%: 54.55 -> 57.35 | Improved |
| Use relation, type, and entity information with direct feature aggregation rather than learned type abstraction | Path-RNN C · WN18RR | MAP%: 67.16 -> 51.08 | Hurt |
| Use APR with learned type abstraction and LogSumExp pooling instead of the strongest prior path-ranking result | WN18RR · WN18RR | MAP%: 67.16 -> 74.51 | Improved |
| Use APR with learned type abstraction and LogSumExp pooling instead of the strongest prior path-ranking result | FB15k-237 · FB15k-237 | MAP%: 52.17 -> 54.09 | Improved |
| Use APR D instead of ComplEx-N3 | Knowledge base fact prediction · WN18RR | MAP%: 40.89 -> 84.91 | Improved |
Closest papers by meaning
- Replacing Paths with Connection-Biased Attention for Knowledge Graph Completion (AAAI 2025 · similarity 0.89)
- Anchoring Path for Inductive Relation Prediction in Knowledge Graphs (AAAI 2024 · similarity 0.87)
- Context-aware Inductive Knowledge Graph Completion with Latent Type Constraints and Subgraph Reasoning (AAAI 2025 · similarity 0.86)
- Knowledge Graph Completion with Relation-Aware Anchor Enhancement (AAAI 2025 · similarity 0.85)
- Towards Global-Topology Relation Graph for Inductive Knowledge Graph Completion (AAAI 2025 · similarity 0.84)
- Knowledge Graphs Can be Learned with Just Intersection Features (ICML 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
