Research map / Graph learning / Knowledge and temporal graphs
Commonsense Knowledge Base Completion with Structural and Semantic Context
Chaitanya Malaviya, Chandra Bhagavatula, Antoine Bosselut, Yejin Choi
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 |
|---|---|---|---|
| Use BERT node representations instead of ConvTransE's learned node embeddings | ConvTransE · CN-100K, subgraph sampling | MRR: 18.68 -> 49.56 | Improved |
| Use BERT node representations instead of ConvTransE's learned node embeddings | ConvTransE · ATOMIC, subgraph sampling | MRR: 12.94 -> 12.33 | Hurt |
| Add graph convolutional node representations | ConvTransE · CN-100K, subgraph sampling | MRR: 18.68 -> 29.80 | Improved |
| Add graph convolutional node representations | ConvTransE · ATOMIC, subgraph sampling | MRR: 12.94 -> 13.12 | Improved |
| Add semantic-similarity links (ablation) | GCN + ConvTransE · CN-100K, subgraph sampling | MRR: 29.80 -> 30.03 | Improved |
| Add semantic-similarity links (ablation) | GCN + ConvTransE · ATOMIC, subgraph sampling | MRR: 13.12 -> 13.88 | Improved |
| Fuse GCN representations with BERT representations (ablation) | ConvTransE · CN-100K, subgraph sampling | MRR: 49.56 -> 50.38 | Improved |
| Fuse GCN representations with BERT representations (ablation) | ConvTransE · ATOMIC, subgraph sampling | MRR: 12.33 -> 10.8 | Hurt |
| Add semantic-similarity links to the fused GCN+BERT model (ablation) | GCN + BERT + ConvTransE · CN-100K, subgraph sampling | MRR: 50.38 -> 51.11 | Improved |
| Add semantic-similarity links to the fused GCN+BERT model (ablation) | GCN + BERT + ConvTransE · ATOMIC, subgraph sampling | MRR: 10.8 -> 10.33 | Hurt |
| Use graph embeddings alongside BERT representations (ablation) | ConvTransE · CN-100K, full graph training | MRR: 52.25 -> 50.80 | Hurt |
| Add semantic-similarity links to GCN+BERT (ablation) | ConvTransE · CN-100K, full graph training | MRR: 50.80 -> 50.17 | Hurt |
Closest papers by meaning
- Amplifying Diversity and Quality in Commonsense Knowledge Graph Completion (Student Abstract) (AAAI 2024 · similarity 0.90)
- Towards Global-Topology Relation Graph for Inductive Knowledge Graph Completion (AAAI 2025 · similarity 0.88)
- S²DN: Learning to Denoise Unconvincing Knowledge for Inductive Knowledge Graph Completion (AAAI 2025 · similarity 0.88)
- Replacing Paths with Connection-Biased Attention for Knowledge Graph Completion (AAAI 2025 · similarity 0.88)
- Think-on-Graph 2.0: Deep and Faithful Large Language Model Reasoning with Knowledge-guided Retrieval Augmented Generation (ICLR 2025 · similarity 0.87)
- Structure Is All You Need: Structural Representation Learning on Hyper-Relational Knowledge Graphs (ICML 2025 · similarity 0.87)
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.
