Skip to main content

Table 6 DRAGON-AI performance on definition generation task

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

method

model name

accuracy

score

consistency

DRAGON

gpt-3.5-turbo

4.058

3.632

3.735

DRAGON

gpt-4

3.97

3.567

3.689

DRAGON

nous-hermes-13b

3.776

3.389

3.566

curator

human

4.326

4.069

4.13

  1. A comparison of base performance of DRAGON on definition generation when compared with existing editor-provided definitions. Evaluator scores shown for three score categories (accuracy, consistency, and overall score). Evaluators evaluated definitions generated by three different models, alongside existing ontology definitions. Evaluators were not shown the source of definitions until after evaluation
Back to article page

Journal of Biomedical Semantics

ISSN: 2041-1480