Language segmentation for Optical Character Recognition using Self Organizing Maps


Modern optical character recognition (OCR) systems perform optimally on single-font monolingual texts, and have lower performance on bilingual and multilingual texts. For many OCR tasks it is necessary to accurately recognize characters from bilingual texts such as dictionaries or grammar books. We present a novel approach to segmenting bilingual text… (More)


12 Figures and Tables

Slides referencing similar topics