|
--- |
|
license: apache-2.0 |
|
--- |
|
|
|
# Swedish OCR Correction |
|
|
|
This model is an updated version of https://huggingface.co/viklofg/swedish-ocr-correction |
|
|
|
The model has been trained to correct OCR predictions by Abbyy, Tesseract, and a combination of those on newspaper from 1818-2018 (see [A Two-OCR Engine Method for Digitized Swedish Newspapers](https://ecp.ep.liu.se/index.php/clarin/article/view/8) ). |
|
|
|
Please check the [original model](https://huggingface.co/viklofg/swedish-ocr-correction) for more information. |
|
|
|
This new model has been trained much longer and manages to outperform the previous one using the same train-test split. |
|
|
|
| Model | CER | WER | |
|
| - | - | - | |
|
| Original OCR | 3.01 | 13.23 | |
|
| viklofg | 1.92 | 7.41 | |
|
| KBLab | 1.57 | 6.23 | |
|
|
|
|
