HTR - Araucania - XIX manuscript
收藏资源简介：
<strong>General</strong> Ground Truth dataset for Spanish 19th typewritten OCR (XML-ALTO) The archives come from the events of the Occupation of Araucania (1850-1881) in Chile. They are archived in the 'Colección manuscritos' of the Archivo Central Andres Bello - Universidad de Chile. Thereby, it is not possible to publicly distribute the images (.jpg) To use them for segmentation/recognition model training, please contact me : archivo.central@uchile.cl <strong>Methodology</strong> Transcription rules : - xxx for blurred or unreadable characters<br> - ^+letters for superscript letters<br> - ⁋ for new paragraph Using the Kraken OCR engine in finetuning with the Menu_MacFrench template. A template uses the NFKD method. Segmonto ontology <strong>Evaluation</strong> <strong>Name</strong> <strong>Quantity (GT)</strong> <strong>Val_acc</strong> <strong>Test_acc</strong> <strong>CER</strong> <strong>WER</strong> <strong>HTR-Araucania_XIX</strong> 180 0,90354 0,8673 0,05598 0.21423 <strong>HTR-Araucania_XIX_NFKD</strong> 180 0,89872 0,8563 0,06646 0.24963 <strong>Others</strong> JSONL file for NER annotation in <code>ner/</code> (MISC, LOC, PERS, ORG, DATE)
