File size: 4,782 Bytes
609a9a8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
---
language: de
license: mit
tags:
- flair
- token-classification
- sequence-tagger-model
base_model: dbmdz/bert-base-historic-multilingual-64k-td-cased
widget:
- text: In Teltsch und Jarmeritz wurden die abgegebenen Stimmen für Genossen Krapka
    ungiltig erklärt , weil sie keinen Wohnort aufwiesen .
---

# Fine-tuned Flair Model on German NewsEye NER Dataset (HIPE-2022)

This Flair model was fine-tuned on the
[German NewsEye](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-newseye.md)
NER Dataset using hmBERT 64k as backbone LM.

The NewsEye dataset is comprised of diachronic historical newspaper material published between 1850 and 1950
in French, German, Finnish, and Swedish.
More information can be found [here](https://dl.acm.org/doi/abs/10.1145/3404835.3463255).

The following NEs were annotated: `PER`, `LOC`, `ORG` and `HumanProd`.

# Results

We performed a hyper-parameter search over the following parameters with 5 different seeds per configuration:

* Batch Sizes: `[4, 8]`
* Learning Rates: `[3e-05, 5e-05]`

And report micro F1-score on development set:

| Configuration     | Seed 1       | Seed 2          | Seed 3       | Seed 4       | Seed 5       | Average         |
|-------------------|--------------|-----------------|--------------|--------------|--------------|-----------------|
| `bs8-e10-lr3e-05` | [0.3931][1]  | [**0.4248**][2] | [0.4127][3]  | [0.3938][4]  | [0.4187][5]  | 0.4086 ± 0.0145 |
| `bs4-e10-lr3e-05` | [0.338][6]   | [0.4183][7]     | [0.4041][8]  | [0.4384][9]  | [0.3974][10] | 0.3992 ± 0.0377 |
| `bs8-e10-lr5e-05` | [0.3861][11] | [0.3757][12]    | [0.3764][13] | [0.4099][14] | [0.3593][15] | 0.3815 ± 0.0186 |
| `bs4-e10-lr5e-05` | [0.3813][16] | [0.0][17]       | [0.3339][18] | [0.2489][19] | [0.2931][20] | 0.2514 ± 0.1489 |

[1]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-1
[2]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2
[3]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-3
[4]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-4
[5]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-5
[6]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-1
[7]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2
[8]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-3
[9]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-4
[10]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-5
[11]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-1
[12]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-2
[13]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-3
[14]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-4
[15]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-5
[16]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-1
[17]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-2
[18]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-3
[19]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-4
[20]: https://hf.co/stefan-it/hmbench-newseye-de-hmbert_64k-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-5

The [training log](training.log) and TensorBoard logs (not available for hmBERT Base model) are also uploaded to the model hub.

More information about fine-tuning can be found [here](https://github.com/stefan-it/hmBench).

# Acknowledgements

We thank [Luisa März](https://github.com/LuisaMaerz), [Katharina Schmid](https://github.com/schmika) and
[Erion Çano](https://github.com/erionc) for their fruitful discussions about Historic Language Models.

Research supported with Cloud TPUs from Google's [TPU Research Cloud](https://sites.research.google/trc/about/) (TRC).
Many Thanks for providing access to the TPUs ❤️