It would be valuable to add training data to the Tesseract OCR engine to support OCR for Ethiopic, Bengali and Gujarati.