Retrodigitalization of zbMATH Open volumes (1-529): Mathpix 2024 OCR Results

Dataset representing the result of applying Mathpix OCR API to scanned documents from zbMATH Open volumes. Configurations of the API call: image_uri = "data:image/tif;base64," + \ base64.b64encode( open(file_path, "rb").read() ).decode() OCR_result = requests.post("https://api.mathpix.com/v3/text", data=json.dumps({'src': image_uri, "formats": ["text", "data", "html", "latex_normal"], "data_options": { "include_asciimath": True, "include_latex": True }, "include_line_data": True, "include_detected_alphabets": True, }), headers={"app_id": user, "app_key": key, "Content-type": "application/json"}) The file OCR_representation.tar contains 252,209 JSON files (one per page), organized by volumes (each volume is a separate folder). The numbering of files within folders corresponds to the numbering of the original scans, keeping the same sequence as the scanned pages in the books. Mathpix errors occurred while processing 2,235 scans (see Errors.xlsx). Some volumes are missing (as they were not scanned, see Missing_volumes.txt), 480 in total were processed. How to get from the file to the corresponding zbMATH Open webpage (example): OCR_representation/016/00000431.json -> https://zbmath.org/0016.43101 For more information, see DOI 10.4171/NEWS/117/11