On Devanagari document processing

Devnagari document processing system discussed here makes use of various knowledge sources at all levels. Extraction of test zone from a document is a preprocessing stage which uses document layout knowledge represented syntactically. The test zone is then segmented into lines, lines into words and words into characters. Since Devnagari characters is a complex composition of symbols, various algorithms are used to further segment the character into its constituent symbols instead of treating the character as a unit. The symbols are then recognized using various features which are extracted and saved during training phase. The recognized symbols are composed back and sent for validation through a partitioned dictionary.