AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

What Makes Entities Similar? A Similarity Flooding Perspective for Multi-sourced Knowledge Graph Embeddings

Zequn Sun, Jiacheng Huang, Xiaozhou Xu, Qijin Chen, Weijun Ren, Wei Hu

ICML 2023 · Graph learning › Knowledge and temporal graphs · approach: Knowledge graph embeddings · 1 citations (OpenAlex)

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 changedWhereMeasuredEffect
Replace embedding training with similarity flooding over TransE-derived entity compositionsTransE-based EA · DBP15K ZH-EN, JA-EN, FR-EN; OpenEA D-W and D-YHits@1: 0.308 -> 0.315; 0.279 -> 0.372; 0.244 -> 0.347; 0.259 -> 0.294; 0.463 -> 0.503Improved
Replace embedding training with similarity flooding over GCN-derived entity compositionsGCN-based EA · DBP15K ZH-EN, JA-EN, FR-EN; OpenEA D-W and D-YHits@1: 0.413 -> 0.349; 0.399 -> 0.376; 0.373 -> 0.349; 0.364 -> 0.358; 0.465 -> 0.478Mixed
Add self-propagation connections in the aggregation layers (ablation)GCN-Align · DBP15K ZH-EN, JA-EN, FR-EN; OpenEA D-W and D-YHits@1: 0.413 -> 0.441; 0.399 -> 0.446; 0.373 -> 0.414; 0.364 -> 0.378; 0.465 -> 0.495Improved
Add self-propagation connections in the aggregation layers (ablation)AliNet · DBP15K ZH-EN, JA-EN, FR-EN; OpenEA D-W and D-YHits@1: 0.539 -> 0.575; 0.549 -> 0.570; 0.552 -> 0.581; 0.440 -> 0.451; 0.559 -> 0.563Improved
Add self-propagation connections in the aggregation layers (ablation)Dual-AMN · DBP15K ZH-EN, JA-EN, FR-EN; OpenEA D-W and D-YHits@1: 0.731 -> 0.733; 0.726 -> 0.735; 0.756 -> 0.767; 0.683 -> 0.695; 0.767 -> 0.779Improved
Add self-propagation connections in the aggregation layers (ablation)RoadEA · DBP15K ZH-EN, JA-EN, FR-EN; OpenEA D-W and D-YHits@1: 0.570 -> 0.579; 0.569 -> 0.577; 0.578 -> 0.593; 0.495 -> 0.502; 0.212 -> 0.235Improved
Add multilingual entity-name text features to similarity flooding (ablation)TransFlood · DBP15K ZH-EN, JA-EN, FR-ENHits@1: 0.315 -> 0.670; 0.372 -> 0.747; 0.347 -> 0.881Improved
Add multilingual entity-name text features to similarity flooding (ablation)GCNFlood · DBP15K ZH-EN, JA-EN, FR-ENHits@1: 0.349 -> 0.651; 0.376 -> 0.712; 0.349 -> 0.842Improved
Use more GCN layers without self-propagation (ablation)GCN-Align and AliNet · DBP15K ZH-ENHits@1Hurt
Add self-propagation while increasing GCN depth (ablation)GCN-Align and AliNet · DBP15K ZH-ENHits@1Improved
Add self-propagation while increasing GCN depth (ablation)Dual-AMN · DBP15K ZH-ENHits@1No clear effect
Use SPA for AliNet with equal 384-dimensional layer outputs rather than its original dimensions (ablation)AliNet + SPA · DBP15KperformanceNo clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy