|
--- |
|
license: apache-2.0 |
|
language: |
|
- en |
|
- de |
|
- fr |
|
- fi |
|
- sv |
|
- nl |
|
- nb |
|
- nn |
|
- 'no' |
|
--- |
|
|
|
# hmTEAMS |
|
|
|
[](https://github.com/stefan-it/hmTEAMS) |
|
|
|
Historic Multilingual and Monolingual [TEAMS](https://aclanthology.org/2021.findings-acl.219/) Models. |
|
The following languages are covered: |
|
|
|
* English (British Library Corpus - Books) |
|
* German (Europeana Newspaper) |
|
* French (Europeana Newspaper) |
|
* Finnish (Europeana Newspaper, Digilib) |
|
* Swedish (Europeana Newspaper, Digilib) |
|
* Dutch (Delpher Corpus) |
|
* Norwegian (NCC Corpus) |
|
|
|
# Architecture |
|
|
|
We pretrain a "Training ELECTRA Augmented with Multi-word Selection" |
|
([TEAMS](https://aclanthology.org/2021.findings-acl.219/)) model: |
|
|
|
 |
|
|
|
# Results |
|
|
|
We perform experiments on various historic NER datasets, such as HIPE-2022 or ICDAR Europeana. |
|
All details incl. hyper-parameters can be found [here](https://github.com/stefan-it/hmTEAMS/tree/main/bench). |
|
|
|
## Small Benchmark |
|
|
|
We test our pretrained language models on various datasets from HIPE-2020, HIPE-2022 and Europeana. |
|
The following table shows an overview of used datasets. |
|
|
|
| Language | Dataset | Additional Dataset | |
|
|----------|--------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------| |
|
| English | [AjMC](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-ajmc.md) | - | |
|
| German | [AjMC](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-ajmc.md) | - | |
|
| French | [AjMC](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-ajmc.md) | [ICDAR-Europeana](https://github.com/stefan-it/historic-domain-adaptation-icdar) | |
|
| Finnish | [NewsEye](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-newseye.md) | - | |
|
| Swedish | [NewsEye](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-newseye.md) | - | |
|
| Dutch | [ICDAR-Europeana](https://github.com/stefan-it/historic-domain-adaptation-icdar) | - | |
|
|
|
# Results |
|
|
|
| Model | English AjMC | German AjMC | French AjMC | Finnish NewsEye | Swedish NewsEye | Dutch ICDAR | French ICDAR | Avg. | |
|
|----------------------------------------------------------------------------------------|--------------|--------------|--------------|-----------------|-----------------|--------------|--------------|-----------| |
|
| hmBERT (32k) [Schweter et al.](https://ceur-ws.org/Vol-3180/paper-87.pdf) | 85.36 ± 0.94 | 89.08 ± 0.09 | 85.10 ± 0.60 | 77.28 ± 0.37 | 82.85 ± 0.83 | 82.11 ± 0.61 | 77.21 ± 0.16 | 82.71 | |
|
| hmTEAMS (Ours) | 86.41 ± 0.36 | 88.64 ± 0.42 | 85.41 ± 0.67 | 79.27 ± 1.88 | 82.78 ± 0.60 | 88.21 ± 0.39 | 78.03 ± 0.39 | **84.11** | |
|
|
|
# Release |
|
|
|
Our pretrained hmTEAMS model can be obtained from the Hugging Face Model Hub: |
|
|
|
* [hmTEAMS Discriminator (**this model**)](https://huggingface.co/hmteams/teams-base-historic-multilingual-discriminator) |
|
* [hmTEAMS Generator](https://huggingface.co/hmteams/teams-base-historic-multilingual-generator) |
|
|
|
# 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 ❤️ |
