Table 2.
Specific applications of resources in the lexicon construction and evaluations.
| Application | Aa | Bb | Cc | Dd | Ee | A+E | EHRf corpus | MedLex | UMLSg | SNOMED-CTh | ||
| Lexicon construction | ||||||||||||
|
|
Fine-tuning the BERTi model | ✓ |
|
|
|
✓ | ✓ |
|
|
|
||
|
|
Dictionary entry collection |
|
|
|
|
|
|
✓ |
|
|
|
|
|
|
Dictionary concept enrichment |
|
|
|
|
|
|
|
✓ |
|
|
|
|
|
Semantic type selection |
|
|
|
|
|
|
|
|
✓ |
|
|
| FHj system development | ||||||||||||
|
|
Development of rule-based FH system | ✓ | ✓ |
|
|
|
|
|
|
|
|
|
|
|
Fine-tuning deep learning–based FH system | ✓ | ✓ |
|
|
|
|
|
|
|
|
|
| Evaluation | ||||||||||||
|
|
Dictionary coverage |
|
|
✓ |
|
|
|
|
|
✓ | ✓ | |
|
|
Challenge task 1 |
|
|
✓ | ✓ |
|
|
|
|
|
|
|
|
|
Challenge task 2 |
|
|
✓ | ✓ |
|
|
|
|
|
|
|
aA: Training set (BioCreative).
bB: Training set (N2C2).
cC: Testing set (BioCreative).
dD: Testing set (N2C2).
eE: Training set (I2B2/2010).
fEHR: electronic health record.
gUMLS: Unified Medical Language System.
hSNOMED-CT: Systematized Nomenclature of Medicine Clinical Terms.
iBERT: bidirectional encoder representations from transformers.
jFH: family history.