Explore knowledge graph completion from a practical perspective in this 43-minute conference talk by Danai Koutra at KDD. Dive into the construction, calibration, and evaluation of knowledge graphs in both closed and open world settings. Learn about use cases, human-AI collaboration, and the importance of benchmarks and hyperparameter tuning. Examine the task of link prediction, triple classification, and the comparison between Codex and Freebase. Discover rule-based models, summarization techniques, and inductive approaches. Gain insights into model-independent information and practical applications. Access related papers and the GitHub repository for further exploration.
Revisiting Knowledge Graph Completion From a Practical Perspective - Danai Koutra
Association for Computing Machinery (ACM) via YouTube
Overview
Syllabus
Introduction
Background
Use Cases
Knowledge Graph Construction
Overview
Calibration
Background on Calibration
Multiclass calibration
Negatives
Close World Setting
Results
Open World Assumption
Open World Results
Human AI Collaboration
Recap
Survey
Codecs
Data Collection
Benchmarks
Task of Link Prediction
Hyperparameter Tuning
Triple Classification
Codex vs Freebase
Rules
Summarization
Inductive Summarization
Model Independent Information
RuleBased Model
Summary
Approach
Task Setup
Conclusion
Codex Benchmark
Papers
Github Repository
Questions
Taught by
Association for Computing Machinery (ACM)
