This dataset is a semantically chunked and span-aligned derivative of PleIAs/Post-OCR-Correction.
Each record contains:
an OCR hypothesis chunk from the original text field;
a corresponding post-OCR correction output chunk from the corrected_text field;
metadata inherited from the PleIAs dataset;
character spans linking each chunk back to the original source document;
alignment diagnostics produced during filtering.… See the full description on the dataset page:
https://huggingface.co/datasets/emanuelaboros/pleias-post-ocr-correction-chonkie-aligned-en.
