metadata
language: is
tag: text2text-generation
pipeline_tag: text2text-generation
widget:
- text: ék var að borðaði maturinn min
inference:
parameters:
max_length: 512
license: cc-by-sa-4.0
This is a model for correcting spelling and grammar errors in Icelandic text. It is based on the pretrained ByT5 model (https://arxiv.org/abs/2105.13626) and finetuned on Icelandic error correction data along with synthetic error data. The model is trained using the HuggingFace and PyTorch libraries.
The model is trained to correct a single sentence at a time, but may work on longer context.
The model performs well on correcting a variety of common issues in Icelandic text.
This README will be updated soon along with citation reference.