thedrifter
07-25-07, 07:21 AM
Darpa, NIST Evaluate Military Translation Computers
By K.C. Jones
InformationWeek
Tue Jul 24, 12:44 PM ET
It's tough for military personnel to tell if a civilian has bad or good intentions if they don't speak the same language.
So, as several technology companies work to improve translation software, the National Institute of Standards and Technology also is doing its part. NIST announced this week that its researchers are evaluating instant two-way translation systems for the Spoken Language Communication and Translation System for Tactical Use (Transtac) at the Defense Advanced Research Projects Agency.
Darpa and NIST are focused on bidirectional translations of English and Arabic spoken in Iraq. NIST, U.S. Marines, and Iraqi Arabic speakers just completed lab and outdoor tests on prototypes in Gaithersburg, Md. The evaluations included controlled background noise from speakers, generators, garage doors, running vehicles, and radio broadcasts to mimic noise in real-world situations.
Participants acted out 10 scenarios, including conversations at traffic checkpoints and neighborhood interviews. Those testing the devices carried them in back packs and other hands-free configurations. Lab participants couldn't see their laptop screens as they recorded the conversations. Iraqis who understand English wore earphones that blocked out the English language portions and relayed the system's Arabic interpretations.
Craig Schlenoff, project leader of the NIST evaluation project, said the evaluations showed improvements to the translation systems and provide information about which technologies are most promising.
"Effective two-way translation devices would represent a major advance in field translators," he said in a prepared statement. "Although American forces in Iraq currently have the use of phrase-based translators, the devices can only translate English into pre-recorded Arabic phrases. They cannot translate Iraqi Arabic into English."
Darpa hopes to eventually provide American forces with palm-sized translators for increased convenience, ease, and safety in war zones and potentially hostile environments. Darpa also wants to position itself to develop automatic translator systems within 90 days of receiving a request for that language.
As part of Transtac, IBM has created bidirectional translation systems called Multilingual Automatic Speech-to-Speech Translator software in rugged compact boxes -- the size of a few composition notebooks stacked on top of each other -- for communications between military personnel and citizens in Iraq. The company announced the deployment of 35 Mastor laptops to the Army Medical Department, the U.S. Special Operations Command, and the Marines for testing and training last year.
Scientists have worked with engineers, manufacturing, and military experts at IBM's T.J. Watson Research Center since 2001 to combine automatic speech recognition, natural language understanding, and speech synthesis for a system that can overcome speech recognition errors and grammatically incorrect phrases common in conversations. Mastor's vocabulary exceeds 50,000 English words and 100,000 in Iraqi Arabic.
IBM is eyeing public and private sector market opportunities for language translation in several fields, including aerospace and defense, medicine, law enforcement, banking, travel, and government customer service.
Ellie
By K.C. Jones
InformationWeek
Tue Jul 24, 12:44 PM ET
It's tough for military personnel to tell if a civilian has bad or good intentions if they don't speak the same language.
So, as several technology companies work to improve translation software, the National Institute of Standards and Technology also is doing its part. NIST announced this week that its researchers are evaluating instant two-way translation systems for the Spoken Language Communication and Translation System for Tactical Use (Transtac) at the Defense Advanced Research Projects Agency.
Darpa and NIST are focused on bidirectional translations of English and Arabic spoken in Iraq. NIST, U.S. Marines, and Iraqi Arabic speakers just completed lab and outdoor tests on prototypes in Gaithersburg, Md. The evaluations included controlled background noise from speakers, generators, garage doors, running vehicles, and radio broadcasts to mimic noise in real-world situations.
Participants acted out 10 scenarios, including conversations at traffic checkpoints and neighborhood interviews. Those testing the devices carried them in back packs and other hands-free configurations. Lab participants couldn't see their laptop screens as they recorded the conversations. Iraqis who understand English wore earphones that blocked out the English language portions and relayed the system's Arabic interpretations.
Craig Schlenoff, project leader of the NIST evaluation project, said the evaluations showed improvements to the translation systems and provide information about which technologies are most promising.
"Effective two-way translation devices would represent a major advance in field translators," he said in a prepared statement. "Although American forces in Iraq currently have the use of phrase-based translators, the devices can only translate English into pre-recorded Arabic phrases. They cannot translate Iraqi Arabic into English."
Darpa hopes to eventually provide American forces with palm-sized translators for increased convenience, ease, and safety in war zones and potentially hostile environments. Darpa also wants to position itself to develop automatic translator systems within 90 days of receiving a request for that language.
As part of Transtac, IBM has created bidirectional translation systems called Multilingual Automatic Speech-to-Speech Translator software in rugged compact boxes -- the size of a few composition notebooks stacked on top of each other -- for communications between military personnel and citizens in Iraq. The company announced the deployment of 35 Mastor laptops to the Army Medical Department, the U.S. Special Operations Command, and the Marines for testing and training last year.
Scientists have worked with engineers, manufacturing, and military experts at IBM's T.J. Watson Research Center since 2001 to combine automatic speech recognition, natural language understanding, and speech synthesis for a system that can overcome speech recognition errors and grammatically incorrect phrases common in conversations. Mastor's vocabulary exceeds 50,000 English words and 100,000 in Iraqi Arabic.
IBM is eyeing public and private sector market opportunities for language translation in several fields, including aerospace and defense, medicine, law enforcement, banking, travel, and government customer service.
Ellie