Following an agreement with the BnF (French National Library) and in accordance with copyrights, we propose for the competition a dataset that is a subpart of the corpus collected in the context of the AmeliOCR project, led by the L3i laboratory (University of La Rochelle, France) and the BnF.
The dataset is multilingual (50% French, 50% English, with maybe few expressions borrowed from other languages such as Latin). It consists in texts published in monographs and periodicals during the last four centuries. It accounts more than 12 million characters, and includes both noisy OCR-ed texts and the corresponding Gold-Standard (GS) which has been aligned at the character level.
The documents come from different collections (e.g. BnF, British Library) supported by various projects (e.g. Europeana Newspapers, IMPACT, Gutenberg, Perseus, Wikisource and Bank of wisdom).
Error rates, and thus difficulty, vary according to the nature of the documents. In both English and French parts of the dataset, historical newspapers, which have been reported to be especially challenging, represent more than 25% of the data.
Important notice, in the archive :
- Even ids (document numbers) were exclusively used for Task 1) Detection
- Odd ids (document numbers) were exclusively used for Task 2) Correction
- IMPACT, European Commission’s 7th Framework Program, grant agreement 215064
- EU Competitiveness and Innovation Framework Programme grant ENP 297380
Origins and copyrights related to every text are detailed in the full version of the dataset.