Wikipedia
MANTRA-Rajbhasha is a machine assisted translation tool, developed by C-DAC which translates from English to Hindi. The system is based on the MANTRA Technology developed by the Applied Artificial Intelligence Group (AAI) of C-DAC, Pune.
MANTRA translates the English text into Hindi in a specified domain of Personal Administration, specifically Gazette Notifications, Office Orders, Office Memorandums and Circulars. The utility of MANTRA has to be viewed in the context of the Indian environment, its socio-economic and demographic conditions, its historical background and above all the transitory phase India is passing through presently. Translating from one language to another, particularly of a different family, covers a very broad spectrum. The strategy adopted here is - NOT WORD-TO-WORD... NOT RULE TO RULE.... but a LEXICAL TREE-TO-LEXICAL TREE. MANTRA uses the Lexicalized Tree Adjoining Grammar (LTAG) formalism to represent the English as well as the Hindi grammar.
Sponsored by the Department of Official Language (DOL), Ministry of Home Affairs, Government of India, Mantra-Rajbhasha has been developed as personal computer, Intranet and Internet versions. MANTRA-Rajbhasha has been designed to assist the various Departments and Ministries of the Government of India to procure quick and standard Hindi translations.
MANTRA is a recipient of the Computerworld Smithsonian Award and a part of the '1999 Innovation Collection' in the National Museum for American History. In MANTRA-Rajbhasha, the English document is initially sent for pre processing followed by parsing and generation during which the English Lexicon, Hindi Lexicon and Transfer Lexicon are used. The generator generates the Hindi output that undergoes post processing (if required) and leading to the translated Hindi document.
The MANTRA-Rajbhasha Internet version has been designed and developed, using the thin client architecture where the complete translation process is carried out at the server side. This enables even low-end systems with Internet connectivity to avail the facility to translate the documents from remote locations. The translated documents are stored in the user inbox for later retrieval.
Besides the common translation tools such as add word, multiple output, online thesaurus and dictionary, the system is enriched with added functionalities for report generation, splitting a large document into smaller ones for distribution to more than one translator/vetter and merging the split translated documents into one.
The MANTRA Technology is being expanded to translate English texts into other Indian official languages such as Gujarati, Bengali, and Telugu. The accuracy rate of translation in these languages is considerably high. Besides, it is also being expanded to translate Hindi text into English in the specified domain of personnel administration, as well as other domains like Banking, Transportation and Agriculture.
The Salient Features of MANTRA are:
- Format Retention ( Font sizes, alignment, styles)
- Input to MANTRA can be through:
- .rtf (Rich Text Format) or .htm file created using standard DTP packages
- Speech recognition program
- Optical Character Recognition Package
- Pre-processing Tools:
- Phrase Marker
- Proper noun, Dates and other domain specific identifiers
- Spell and Grammar checker
- Uses Tree Adjoining Grammar (TAG) for Parsing and Generation
- Custom modifications to the Earley's style bottom-up parsing algorithm to speed up the parser
- User friendly selection tool for multiple outputs
- On-line word addition and grammar creation, updation facility