The goal of our speech-to-speech translation work is to enable simple voice communication across a language barrier, using a handheld PDA with a microphone. Using the system, an English-speaking operator asks the foreign-language-speaking subject simple questions, such as "Where does it hurt?" or "How long have you had the pain?". The system uses speech recognition to determine which question the English speaker has asked, and then plays out a recorded translation to the subject. The subject replies in his native language, and the system uses speech recognition to turn this speech into text. The system then uses a combination of a grammar and statistical information-extraction techniques to find the directly relevant information in the reply, and render this into English. The system was initially developed for the colloquial Arabic of the Levant region (Lebanon, Jordan, Palestine, Syria). Other languages then followed, including Iraqi Arabic and Indonesian.