Automated transcription and translation systems

Research importance: Automated tools are crucial because agencies lack the manpower to monitor, translate, transcribe and analyze the huge volume of communications being collected as part of anti- terrorism efforts.

Potential benefits: Speech-recognition technologies could play a key role by helping officials transcribe and translate broadcast media, telephone conversations and in-person communications. Portable translation devices could also assist intelligence analysts and soldiers in the field.

Current state of the art: Systems remain imperfect, with the best conversational speech transcription systems getting one of every three words wrong. Translation devices used in Afghanistan today are little more than voice-activated phrase books tethered to a limited number of foreign phrases and unable to translate responses.

Research and development success factors: The goals are to reduce word error rates to less than 10 percent and develop handheld devices capable of translating limited two-way communications between English speakers and speakers of Pashto, Arabic, Dari Farsi or Mandarin Chinese.

Challenges: The main challenge is developing solutions that can overcome different language dialects, accents and noisy environments. Current portable computer platforms are not powerful enough to support advanced translation capabilities. Also, recruiting people who can speak certain languages to assist in system development is difficult.