Portuguese post-OCR Resources for Text Optimisation

Overview Resource for evaluation and develop OCRs and Post-OCR focused on historical Portuguese. How to load Via hugging face from datasets import load_dataset dataset = load_dataset("LIACC/PORTO") Via Zenodo from datasets import load_dataset dataset = load_dataset("parquet", data_files="path/to/porto_dataset/*.parquet") Metadata dataset_info: features: - name: corpus dtype: string - name: book dtype: string - name: date dtype: int32 - name: filename dtype: string - name: transcription dtype: string - name: aadi dtype: string - name: paddleocr dtype: string - name: tesseract dtype: string - name: transkribus dtype: string - name: img dtype: image - name: complex_layout dtype: bool - name: language dtype: string splits: - name: train num_examples: 3782 license: cc-by-4.0 task_categories: - image-to-text - fill-mask - text-generation language: - pt