metadata
language: fo
tag: text2text-generation
pipeline_tag: text2text-generation
widget:
- text: l/ú veit eg tað várar í P'oroyum
inference:
parameters:
max_length: 512
Model Card for Model ID
OCR post processing for Faroese.
Model Details
This model is finetuned using a ByT5 model (base) trained on Icelandic OCR post-processing data: https://huggingface.co/atlijas/byt5-is-ocr-post-processing-modern-texts The Faroese training data was created by extracting authentic errors from OCR-ed Faroese texts and applied to a corpus of Faroese, along with random character noise.