Translating the GWOT

An interesting group decision support system is currently evolving to help fight the global war on terror (GWOT). A National Virtual Translation Center connects linguists embedded with troops on the front line with networked translation assistance technologies. Currently this is accomplished with thousands of linguists pouring over the actual documents and audio clips brought from the front in an effort to extrapolate useful information. The new technology will digitize, parse and digest raw intelligence material, to facilitate the analyzing of material, instead of simply translating it.

This center is changing the way the government approaches translation and analysis. According to Kevin Hendzel, a Russian linguist,

“The government is appallingly behind on computer-assisted translation, because they’ve invested all this money on machine translation.”

So, the government is discarding its work on developing the “ultimate spy computer” to replace translators and instead, is developing tools that will help translators perform their jobs more effectively. For example, instead of software that can translate printed Arabic, an analyst would benefit from software that made Arabic writing easier to read. This would greatly improve the efficiency of translating recovered enemy documents. So what does this entail?

Tying together emerging translation-assistance technologies into a large-scale deployment, the “translation web” will be administered from the center and leverage the language skills of people all over the U.S., such as professors, contract translators, government linguists and even retirees; the existing expertise is never lost. This is accomplished by providing linguists remote access to a suite of technologies and tools to digitize, parse and digest raw material. These tools include scanning software, databases that facilitate searches for translated phrases and web-based collaboration methods designed to assist the linguist on the front line. One of the major limitations faced by the center is in the area of optical character recognition.

Until recently, the technology software responsible for recognizing printed characters and rendering them in digital format was limited due to shortsighted development. The software was based on the Roman alphabet and the major European languages; it also followed language-specific rules and only recognized distinct letter symbols. Therefore its reliability when used for non-European languages was relatively poor. Although technology exists to recognize non-European languages, such as Arabic, its reliability requires clearly printed documents; handwritten documents won’t work. A new approach was required to correct this deficiency.

A research group at BBN Technologies in Cambridge, MA, has developed new smart optical recognition software that’s flexible and trainable. For example, where the former software was programmed to look for the distinct shape of the Roman letter ‘R’, the new system looks for a range of shapes that could possibly be an ‘R’, then looks for matches in a list of models created during its training. This new system is much more effective with blurry or handwritten text and is adaptable to a myriad of languages, including Arabic, Chinese and Japanese. How will systems such as this be accessed?

The optical recognition system and all the other translation tools will be tied together by an Internet-based system designed to dispatch projects to analysts in the field, including those on the front lines in deployed locations around the world. This collaboration system will improve the speed and efficiency of translating and analysis, as well as allow linguists to collaborate virtually. However, the challenge to building a web architecture to allow this kind of collaboration requires individual federal agencies to maintain security, not just from hackers, but also from each other. To meet this challenge, the Center is modifying the existing Internet collaboration software to permit the sharing of some databases, while still protecting proprietary information.

The National Virtual Translation Center is a group decision support system designed to give the U.S. an edge in its GWOT. The Center establishes a translation web linking linguists around the world, and provides them with innovative software and database tools to effectively translate documents and voice recordings. Using technology such as flexible optical recognition software and an Internet collaboration system, translators can quickly perform the functions necessary to help protect America.