Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
pipeline_tag: token-classification
|
| 4 |
+
tags:
|
| 5 |
+
- ner
|
| 6 |
+
- gliner
|
| 7 |
+
- data-use
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# gliner_datause_extended
|
| 11 |
+
|
| 12 |
+
Fine-tune of `urchade/gliner_large-v2.1` for data-use mention extraction
|
| 13 |
+
(dataset / survey / census / registry mentions in economics research papers).
|
| 14 |
+
|
| 15 |
+
## Labels
|
| 16 |
+
- `NAMED_DATA` — a proper name, title, or acronym of a specific data source
|
| 17 |
+
- `DESCRIPTIVE_DATA` — a source described in words but not named
|
| 18 |
+
- `VAGUE_DATA` — generic data wording with no identifiable source
|
| 19 |
+
|
| 20 |
+
## Training
|
| 21 |
+
- base model: `urchade/gliner_large-v2.1`
|
| 22 |
+
- dataset: `rafmacalaba/data-use-mentions-extended` (gliner config)
|
| 23 |
+
- epochs: 5
|
| 24 |
+
- learning rate: 5e-06
|
| 25 |
+
- batch size: 16
|
| 26 |
+
- precision: bf16
|
| 27 |
+
|
| 28 |
+
## Evaluation (holdout)
|
| 29 |
+
|
| 30 |
+
| thr | tp | fp | fn | precision | recall | f0.5 | f1 |
|
| 31 |
+
| --- | --- | --- | --- | --- | --- | --- | --- |
|
| 32 |
+
| 0.10 | 12275 | 8853 | 289 | 0.5810 | 0.9770 | 0.6322 | 0.7287 |
|
| 33 |
+
| 0.20 | 12153 | 6445 | 411 | 0.6535 | 0.9673 | 0.6988 | 0.7800 |
|
| 34 |
+
| 0.30 | 12013 | 5218 | 551 | 0.6972 | 0.9561 | 0.7371 | 0.8064 |
|
| 35 |
+
| 0.40 | 11844 | 4143 | 720 | 0.7409 | 0.9427 | 0.7740 | 0.8297 |
|
| 36 |
+
| 0.50 | 11489 | 3002 | 1075 | 0.7928 | 0.9144 | 0.8145 | 0.8493 |
|
| 37 |
+
| 0.60 | 10542 | 1935 | 2022 | 0.8449 | 0.8391 | 0.8437 | 0.8420 |
|
| 38 |
+
| 0.70 | 8396 | 940 | 4168 | 0.8993 | 0.6683 | 0.8411 | 0.7668 |
|
| 39 |
+
|
| 40 |
+
**Best F0.5**: 0.8437 (thr=0.6)
|
| 41 |
+
**Best F1**: 0.8493 (thr=0.5)
|
| 42 |
+
|