Improving OCR Accuracy on Early Printed Books by combining Pretraining, Voting, and Active Learning.
Christian ReulUwe SpringmannChristoph WickFrank PuppePublished in: CoRR (2018)
Keyphrases
- active learning
- optical character recognition
- annotation effort
- computational cost
- high accuracy
- scanned documents
- prediction accuracy
- error rate
- semi supervised
- multi class
- relevance feedback
- decision trees
- error correction
- document processing
- selective sampling
- random selection
- classification accuracy
- scanned images
- preprocessing