Machine Translation of Battlefield Messages

Effective communication is vital to winning battles. This is particularly true on the digital battlefield of the next century. Communication must be multilingual; allowing input from or output to either coalition partners or indigenous populations who are not English speakers. Language Translation and Understanding (LT&U) technologies can contribute to improved real-time situational awareness, decision making, and real-time force synchronization through automatic translation of messages between multi-national friendly forces, automatic translation of intercepted hostile messages, and multilingual interfaces to critical information. LT&U technology is improving but there are still several severe technical challenges that have resulted in barriers to the use of this technology. In general, previous language processing systems require massive manual effort by computational linguists for each input language, for example, to write a grammar for the language, to handcraft syntactic and semantic properties of each word, and to write formal semantic rules. As a result, translation technology tends to be specific to a given pair of source and target languages, e.g., French to English. In addition, for language translation and understanding, the technology is too brittle in the face of unexpected input, such as input errors, words not in the lexicon, or words used in an unknown way. Finally, to create large-scale, language-specific lexicons and parsers has hitherto required huge labor efforts that are repeated anew for each new language. We propose to develop a language translation prototype that addresses the technical challenges above through the use of (1) Language processing modules that are easily extensible to more than one language or language pair, (2) Procedures for text and speech understanding in the face of unexpected input, and (3) Techniques for rapid creation of large-scale, language-specific dictionaries and grammars. Through this project, not only will technology for language translation and understanding be advanced, but also critical functionality for battlefield communications and battle management will be significantly enhanced. This work has enormous potential for improving milit