stefan-it commited on
Commit
3046172
1 Parent(s): dc89aad

readme: add initial version of model card

Browse files

Hi,

this PR add the initial version of model card.

Files changed (1) hide show
  1. README.md +73 -0
README.md ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: de
3
+ license: mit
4
+ tags:
5
+ - flair
6
+ - token-classification
7
+ - sequence-tagger-model
8
+ base_model: hmteams/teams-base-historic-multilingual-discriminator
9
+ widget:
10
+ - text: In Teltsch und Jarmeritz wurden die abgegebenen Stimmen für Genossen Krapka
11
+ ungiltig erklärt , weil sie keinen Wohnort aufwiesen .
12
+ ---
13
+
14
+ # Fine-tuned Flair Model on German NewsEye NER Dataset (HIPE-2022)
15
+
16
+ This Flair model was fine-tuned on the
17
+ [German NewsEye](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-newseye.md)
18
+ NER Dataset using hmTEAMS as backbone LM.
19
+
20
+ The NewsEye dataset is comprised of diachronic historical newspaper material published between 1850 and 1950
21
+ in French, German, Finnish, and Swedish.
22
+ More information can be found [here](https://dl.acm.org/doi/abs/10.1145/3404835.3463255).
23
+
24
+ The following NEs were annotated: `PER`, `LOC`, `ORG` and `HumanProd`.
25
+
26
+ # Results
27
+
28
+ We performed a hyper-parameter search over the following parameters with 5 different seeds per configuration:
29
+
30
+ * Batch Sizes: `[8, 4]`
31
+ * Learning Rates: `[3e-05, 5e-05]`
32
+
33
+ And report micro F1-score on development set:
34
+
35
+ | Configuration | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Avg. |
36
+ |-----------------|-------------|--------------|--------------|--------------|--------------|--------------|
37
+ | bs8-e10-lr5e-05 | [0.4156][1] | [0.3964][2] | [0.4141][3] | [0.4106][4] | [0.3922][5] | 40.58 ± 0.96 |
38
+ | bs4-e10-lr3e-05 | [0.416][6] | [0.3886][7] | [0.3998][8] | [0.4309][9] | [0.3884][10] | 40.47 ± 1.65 |
39
+ | bs8-e10-lr3e-05 | [0.407][11] | [0.3797][12] | [0.4035][13] | [0.4078][14] | [0.3891][15] | 39.74 ± 1.11 |
40
+ | bs4-e10-lr5e-05 | [0.396][16] | [0.3856][17] | [0.4044][18] | [0.3894][19] | [0.3974][20] | 39.46 ± 0.65 |
41
+
42
+ [1]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-1
43
+ [2]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-2
44
+ [3]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-3
45
+ [4]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-4
46
+ [5]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-5
47
+ [6]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-1
48
+ [7]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2
49
+ [8]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-3
50
+ [9]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-4
51
+ [10]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-5
52
+ [11]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-1
53
+ [12]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2
54
+ [13]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-3
55
+ [14]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-4
56
+ [15]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-5
57
+ [16]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-1
58
+ [17]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-2
59
+ [18]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-3
60
+ [19]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-4
61
+ [20]: https://hf.co/stefan-it/hmbench-newseye-de-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-5
62
+
63
+ The [training log](training.log) and TensorBoard logs (only for hmByT5 and hmTEAMS based models) are also uploaded to the model hub.
64
+
65
+ More information about fine-tuning can be found [here](https://github.com/stefan-it/hmBench).
66
+
67
+ # Acknowledgements
68
+
69
+ We thank [Luisa März](https://github.com/LuisaMaerz), [Katharina Schmid](https://github.com/schmika) and
70
+ [Erion Çano](https://github.com/erionc) for their fruitful discussions about Historic Language Models.
71
+
72
+ Research supported with Cloud TPUs from Google's [TPU Research Cloud](https://sites.research.google/trc/about/) (TRC).
73
+ Many Thanks for providing access to the TPUs ❤️