A PDA-Like device with a full-duplex sound device... The speaking party wears a headset and the audio is input to voice rocognition software. The voice recognition software then outputs the text data to a translator. The translator then outputs to a speech synthesis application, such as festival. the audio is out put to a speaker or headphones of the second party, or whatever.
I don't beleive I know of any portable devices designed for this purpose.
Of course you have to buy 2 to be able to hold a conversation in realtime.. That's where my fictional company gets ya.
If someone wants to steal my idea, they are more than welcome to. I only ask that you call the product "Maple Syrup" and sell it in a plastic bottle shaped like a woman.
Note the date and time. This writup is more legally binding than a patent.