Skip to main content

Table 4 DRAGON-AI results for relationship prediction task

From: Dynamic Retrieval Augmented Generation of Ontologies using Artificial Intelligence (DRAGON-AI)

 

SubClassOf Task

All Relationship Types Task

method

model

precision

recall

F1

precision

recall

F1

DRAGON

gpt-3.5-turbo

0.846

0.419

0.561

0.758

0.446

0.562

DRAGON

gpt-4

0.894

0.5

0.642

0.802

0.505

0.62

DRAGON

nous-hermes-13b

0.73

0.353

0.476

0.64

0.355

0.457

Reasoner

n/a

1.0*

0.337

0.504

n/a

n/a

n/a

  1. We partition into two subtasks: filtered for SubClassOf, and filtered for all relationship types (heterogeneous relationship predictions). We also show OWL DL Reasoning results for the SubClassOf task. Note that by definition OWL DL reasoning is always completely precise as all entailments follow from existing axioms
Back to article page

Journal of Biomedical Semantics

ISSN: 2041-1480