Lightweight finetuned text-to-text transformer for text cleaning/correction/normalization

Hello, I’m looking for models that I can use to reformat badly parsed text (e.g. punctuation, extra or missing spaces). I’m looking for T5 type of models, I already tested some fine-tuned models but they do not work well for my task. Anyone if aware of such fine-tuned text-to-text models on the Hub? Or any suggestions of other models/papers that could perform this task without needing more fine-tuning? Thanks.

1 Like