Cluster sampling and stratification reduce human annotation costs for knowledge graph accuracy evaluation by up to 80% while maintaining statistical quality.
More specifically, min {MIk ,m} triples are randomly drawn from the k-th sample cluster without replacement
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DB 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Efficient Knowledge Graph Accuracy Evaluation
Cluster sampling and stratification reduce human annotation costs for knowledge graph accuracy evaluation by up to 80% while maintaining statistical quality.