elenanereiss
commited on
Commit
•
8d94ca4
1
Parent(s):
5ecfef0
Update README.md
Browse files
README.md
CHANGED
@@ -4,24 +4,49 @@ license: cc-by-4.0
|
|
4 |
tags:
|
5 |
- named-entity-recognition, legal, ner
|
6 |
datasets:
|
7 |
-
-
|
8 |
metrics:
|
9 |
- precision
|
10 |
- recall
|
11 |
- f1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
---
|
13 |
|
14 |
-
## Model description
|
15 |
|
16 |
-
|
17 |
|
18 |
-
##
|
19 |
|
20 |
-
|
|
|
|
|
|
|
21 |
|
22 |
-
##
|
23 |
|
24 |
-
|
25 |
|
26 |
## Training procedure
|
27 |
|
@@ -31,11 +56,10 @@ The following hyperparameters were used during training:
|
|
31 |
- learning_rate: 1e-05
|
32 |
- train_batch_size: 12
|
33 |
- eval_batch_size: 16
|
34 |
-
-
|
35 |
- num_epochs: 3
|
36 |
|
37 |
-
|
38 |
-
### Results
|
39 |
|
40 |
```
|
41 |
eval_loss = 0.020239440724253654
|
@@ -60,4 +84,21 @@ test_samples_per_second = 59.849
|
|
60 |
test_steps_per_second = 3.748
|
61 |
```
|
62 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
|
|
|
4 |
tags:
|
5 |
- named-entity-recognition, legal, ner
|
6 |
datasets:
|
7 |
+
- german-ler
|
8 |
metrics:
|
9 |
- precision
|
10 |
- recall
|
11 |
- f1
|
12 |
+
model-index:
|
13 |
+
- name: elenanereiss/bert-german-ler
|
14 |
+
results:
|
15 |
+
- task:
|
16 |
+
name: Token Classification
|
17 |
+
type: token-classification
|
18 |
+
dataset:
|
19 |
+
name: german-ler
|
20 |
+
type: german-ler
|
21 |
+
args: german-ler
|
22 |
+
metrics:
|
23 |
+
- name: F1
|
24 |
+
type: f1
|
25 |
+
value: 0.9546215361725869
|
26 |
+
- name: Precision
|
27 |
+
type: precision
|
28 |
+
value: 0.9449558173784978
|
29 |
+
- name: Recall
|
30 |
+
type: recall
|
31 |
+
value: 0.9644870349492672
|
32 |
+
pipeline_tag: token-classification
|
33 |
+
widget:
|
34 |
+
- text: "Herr W. verstieß gegen § 36 Abs. 7 IfSG."
|
35 |
---
|
36 |
|
|
|
37 |
|
38 |
+
# bert-german-ler
|
39 |
|
40 |
+
## Model description
|
41 |
|
42 |
+
This model is a fine-tuned version of [bert-base-german-cased](https://huggingface.co/bert-base-german-cased) on the
|
43 |
+
[German LER Dataset](https://huggingface.co/datasets/elenanereiss/german-ler).
|
44 |
+
Model fine-tuning is done via [T-NER](https://github.com/asahi417/tner)'s hyper-parameter search (see the repository
|
45 |
+
for more detail). It achieves the following results on the test set:
|
46 |
|
47 |
+
## Intended uses & limitations
|
48 |
|
49 |
+
to do
|
50 |
|
51 |
## Training procedure
|
52 |
|
|
|
56 |
- learning_rate: 1e-05
|
57 |
- train_batch_size: 12
|
58 |
- eval_batch_size: 16
|
59 |
+
- max_seq_length: 512
|
60 |
- num_epochs: 3
|
61 |
|
62 |
+
## Results
|
|
|
63 |
|
64 |
```
|
65 |
eval_loss = 0.020239440724253654
|
|
|
84 |
test_steps_per_second = 3.748
|
85 |
```
|
86 |
|
87 |
+
### Usage
|
88 |
+
to do
|
89 |
+
|
90 |
+
### Reference
|
91 |
+
```
|
92 |
+
@misc{https://doi.org/10.48550/arxiv.2003.13016,
|
93 |
+
doi = {10.48550/ARXIV.2003.13016},
|
94 |
+
url = {https://arxiv.org/abs/2003.13016},
|
95 |
+
author = {Leitner, Elena and Rehm, Georg and Moreno-Schneider, Julián},
|
96 |
+
keywords = {Computation and Language (cs.CL), Information Retrieval (cs.IR), FOS: Computer and information sciences, FOS: Computer and information sciences},
|
97 |
+
title = {A Dataset of German Legal Documents for Named Entity Recognition},
|
98 |
+
publisher = {arXiv},
|
99 |
+
year = {2020},
|
100 |
+
copyright = {arXiv.org perpetual, non-exclusive license}
|
101 |
+
}
|
102 |
+
|
103 |
+
```
|
104 |
|