Table 9 Results for DASH and our symptom alignment method using different embeddings with and without context (accuracy score)
From: Explanatory argumentation in natural language for correct and incorrect medical diagnoses
Model | Accuracy |
|---|---|
DASH | 0.37 |
BERT + no context | 0.39 |
SciBERT + no context | 0.39 |
UMLSBERT + no context | 0.44 |
S-PubMedBERT no context | 0.53 |
BERT + context | 0.53 |
SciBERT + context | 0.57 |
UMLSBERT + context | 0.59 |
S-PubMedBERT + context | 0.70 |