Explanation of Link Predictions on Knowledge Graphs via Levelwise Filtering and Graph Summarization thumbnail
Pause
Mute
Subtitles not available
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Explanation of Link Predictions on Knowledge Graphs via Levelwise Filtering and Graph Summarization

Published on Jun 17, 202426 Views

Link Prediction methods aim at predicting missing facts in Knowledge Graphs (KGs) as they are inherently incomplete. Several methods rely on Knowledge Graph Embeddings, which are numerical representat

Related categories

Chapter list

Explanation of Link Predictions on Knowledge Graphs via Levelwise Filtering and Graph Summarization00:00
Outline00:01
Context: Knowledge Graphs [7]00:20
Kelpie [10]: Explaining Link Predictions01:05
Kelpie: High-Level Architecture - 101:35
Kelpie: High-Level Architecture - 202:08
Kelpie: High-Level Architecture - 302:46
Goals02:57
Tackling Goals - 103:35
Semantic Pre-Filter03:43
Tackling Goals - 204:40
Quotient Explanation Builder04:58
Compute the quotient graph - Simulation05:26
Build Explanations from the Quotient06:11
Alternative quotients07:12
Solution 2: Bisimulation07:49
Solution 3: Depth-1 Bisimulation08:29
Experimental Setting09:09
Results: Necessary Relevance10:22
Results: Sufficient Relevance11:38
Results: cases of limited effectiveness11:51
Qualitative Evaluation12:49
Conclusions13:59
Thanks for the attention!14:44