Datasets:
Tasks:
Token Classification
Modalities:
Text
Formats:
parquet
Languages:
Dutch
Size:
10M - 100M
Tags:
medical
License:
Update README.md
Browse files
README.md
CHANGED
|
@@ -11,8 +11,10 @@ size_categories:
|
|
| 11 |
- 1M<n<10M
|
| 12 |
---
|
| 13 |
|
| 14 |
-
Using the UMLS and the SNOMED-CT ontology we identified positive term pairs by
|
| 15 |
* making random pairs of descriptions of the same medical concept identifier
|
|
|
|
|
|
|
| 16 |
* using the known positive relationships between CUIs
|
| 17 |
|
| 18 |
We expand this with negative term pairs by random sampling, followed by selecting the top-N most dissimilar terms with (sim<0.25)
|
|
|
|
| 11 |
- 1M<n<10M
|
| 12 |
---
|
| 13 |
|
| 14 |
+
Using the UMLS and the SNOMED-CT ontology we identified *hard* positive term pairs by
|
| 15 |
* making random pairs of descriptions of the same medical concept identifier
|
| 16 |
+
|
| 17 |
+
and *soft* positive term pairs by
|
| 18 |
* using the known positive relationships between CUIs
|
| 19 |
|
| 20 |
We expand this with negative term pairs by random sampling, followed by selecting the top-N most dissimilar terms with (sim<0.25)
|