Table 4.
Concept normalization performance of existing NLP pipelines stand-alone and combined with RECEEM
| Methods | Precision | Recall | F1 |
|---|---|---|---|
| CLAMP | 0.477(597/1251) | 0.205(597/2906) | 0.287 |
| 0.449-0.505 | 0.191-0.22 | 0.269-0.305 | |
| CLAMP+RECEEM | 0.687(1693/2465) | 0.583(1693/2906) | 0.630a |
| 0.661-0.712 | 0.555-0.609 | 0.605-0.655 | |
| cTAKES | 0.286(636/2226) | 0.219(636/2906) | 0.248 |
| 0.266-0.305 | 0.204-0.233 | 0.232-0.264 | |
| cTAKES + RECEEM | 0.469(1312/2800) | 0.451(1312/2906) | 0.460a |
| 0.444-0.492 | 0.426-0.477 | 0.436-0.484 |
CLAMP: Clinical Language Annotation, Modeling, and Processing; cTAKES: clinical Text Analysis and Knowledge Extraction System; RECEEM: reconstruct concepts from coordinated elliptical expressions in medical text.
a
The best performing result in the respective task.