Mashtots: A system for optical recognition of handwritten symbols and letters

The system is aimed at optical recognition (OCR) of handwritten Armenian, Russian and English letters and symbols. The system is used for OCR of formalized documents filled-in handwriting. Such forms are used for social and medical insurance, census, elections and other purposes. The system is used for documents of Medical Institutions Data Analyses System (MIDAS) now operational in Armenia.

Main functions of the system

• Development of document forms for scanner data entry
• Registration of Armenian, Russian and English handwritten letters and forming a database
• Big volume of data entry and ensuring correctness of entered data
• Morphological and syntax analyses of recognized data.

The structure of the system

The functional diagram of the system

Hardware and software used in the system

• The Mashtots system uses TWAIN protocol which is a standard and ensures friendly interface between the user and the system.
• OCR of the system is tested with Fujitsu and HP family scanners.
• The system uses Client/Server technology. It is possible to use Workstation for OCR and analyses and the Server - for forming a database.
• The system makes it possible to organize in an optimal way the process of data entry and forming the database with application of different types of scanners, workstations and servers.
• The system recognizes documents with mixed alphabet, numbers and special marks.

Minimal hardware/software needed for the system:

• PC Pentium 133 and higher
• Windows 95/98/2000, NT 4.0 (Service Pack 3 and higher)
• 32 MB RAM
• 50 Mb HDD for installation of software
• CD-ROM for installation of software
• LPT port
• Fujitsu or HP6350 type scanner.

The system may be used with other scanners after corresponding testing.

The system recognizes symbols regardless their size and may contain vocabularies for morphological or syntax analyses.

When filling documents according to the rules, OCR error rate doesn’t exceed 0,6%. Adding of vocabularies and analyses will ensure 100% of OCR correctness. Here an example of filled OCR Armenian document is shown.