It will come as no surprise - it is old hat to state that the killer app - quite literally - for augemented sensory reality through technology is the battlefield. The severity of the contest and the willingness to pay is unrivalled. So it will come as no surprise that the US Army is investigating the use of a Universal Translator Star Trek style. The solution to the comprehension problem is years off so basically the idea is to use a limited form of universal speech generation app.

Essentially these are just digital phrase books that actually speak out the required phrase in a number of languages, but using an approach not too unlike what I would think a child is doing to learn language you can do better.
You would work your way out from a sort of functional language model - using simple substitution rules for templated sentences to specifiy a large number of sentences that are precise enough. In the same fashoin you could have some comprehension, by using these sentences in an interrogation style conversation taking not much more than affirmative or negative cues from the responses you're given in a native tongue.

The problem of correctly converting speech to meaning is hard enough, but if you can limit the responder to use a limited functional grammar - that could even be made an unambiguous closed subset of the open ambiguous grammar of the local language, you could conceivable have a valuable interchange with the speakers of this other language.

This is not too unlike the approach in artificial languages like esperanto, except that the use of technology affords additional simplicity in that one could specify a functional subset of any natural language of interest, instead of forcing everybody to adapt to a shared functional language subset.

This seems to me to be a case for humanistic intelligence. The downconversion to a subset of your own language is close to effortless. - The translation of the downconverted language to a subset of another natural language could be feasible for a computer, and would make communication a lot simpler.

We all do this when we try to speak some language of last resort in a foreign country. Working out from stock phrases using a simple subset of techniques for constructing meaning in the foreign language, we try to make ourselves understood. This may be a personal experience of mine, but I always find the approach stumbles when I'm answered, in the signal processing phase - when I receive auditory input in response that is alien to me.
If I could have just a babelfish quality translation of words said to me in French with me when traveling I would be infinitely better off than I am today if forced to understand French.
A device to offer that kind of information should be possible today - even in a mobile device - certainly a personal computer of fairly recent model is sufficient, and the most powerful palmtops are only something like 6-7 years behind stationary devices in computing power.

So in short what I want on my PocketPC is MS Talk - It is a personal dictation program in any of the supported languages, and it offers translation of the quality Systran has been able to offer since the mid '90s of the dictated text. Very few sentences would make any sense at all - but just a fast translation of stock phrases, and common terms would help me book hotel rooms, find the train station etc.