1 paper
Fonseca Cacho, Jorge Ramón
OCR post-processing needs OCR text aligned word by word with its ground truth. Data sets rarely provide that mapping, and a monotone aligner can lose its place when zoning errors p…