Alright it's been a while since i've posted here and if I want to win that mad scientist award (*hint hint* vote for me) I think i'll need some more topics in here, although some of my topics like the earth is flat and circles don't exist topics really brought up some debate...

So anyway this isn't a theoretical idea or mathematical quandary like some of my previous topics, its an invention that should be possible in the near future. A box with a microphone and a speaker and when you talk into it, it translates what you say into another language.

It uses 3 different technologies that we have had for quite some time but still aren't perfect, and just need some fine tuning. One would be voice recognition, which probably needs the most work, especially after seeing what a terrible job Microsoft did, but I think voice recognition in mac OS is pretty good, and my car also has voice recognition which works pretty good, but still not perfect. So another point would be translation from one language to another. We've had translation software for ages but it still isn't perfect. A bit more fine tuning and it should be decent enough for a prototype of this device. And finally would be electronic speech (i.e. turning words into sound). Once again I think apple's got that down with their new voice in their new operating system: leopard. Their new voice sounds like a real person talking to you, it's amazing.

SO anyway first person to put all of these three into one box and sell it will be a millionaire. I would gladly accept a cheque from the first person to use my idea... just make it out to alex

 

 

 

Reply

There is something similar that is used at multinational meetings and summits, but that has the computer replaced with a real live human translator! It should certainly be possible to do what you're suggesting - I can even do it at the moment with a bit of copy and pasting between an online translator and the text-to-speech program. Automating the process of translation and speech should certainly be possible.

However, like you say, these three technologies need large amounts of work. With an automated system like this you would need to make sure that everything was being translated flawlessly, otherwise some very tense situations could result Speech recognition software needs to make massive improvements to understand more natural speech, and be able to better process grammar and other quirks of speech. The translation software needs to be able to distinguish words with different meanings (such as "bore", "drive" or "line") depending on context, and then sort them out to fit the grammar of the language it's translating into. Not a lot more work needs to be done with the text-to-speech system - they are generally OK, if a little unreal. The technology needed here is much nearer completion than the other two.

Now, to the find the address for the patent office...

 

 

 

Reply

Translating languages isn't just about replacing the words, it's about switching them around also, and that's what will prevent this from ever being made. There's no way some system is going to predict what you will say next on the fly, just so it could translate it. Also, how will it differentiate between literal and non literal statements? I could just go on and on with this, but knowing 4 languages (3 fluently), just trust me, there'll have to be psychic devices before anyone can make this.

Reply

No you don't understand. It wouldn't translate it at the same time you are speaking. Only one sentence at a time. So you would speak a sentence, or a paragraph, and when you pause for more than 5 seconds it will process what you have said and say it out loud.

Reply

Not exactly live, and I doubt that nobody hasn't thought of this before. A couple years of work with a complete team of professionals, and you're good to go on your path to become a millionaire.

Reply

Spending a few millions in the process...

I think it's a relatively nice idea and probably should be made, but I still think this product simply couldn't some pull some things off. Puns for example. "Orange ya gonna ask me?" doesn't have that same feel as "Naranja me preguntaras?" in spanish (of course, the nonsensical latter sounds a million times cooler, but anyway...)

Come to think of it, just about any translation doesn't have the same feel in a different language. My experience in reading Shakespeare in Spanish and de la Barca in English has been totally different from reading it in each author's original medium. At international consortiums and diplomatic meetings - where I'd imagine this type of product would be used most - I'd think a feel of what the speaker is saying would be one of the most important things to transmit. Plus, there'd be the added problem of knowing what not to translate. (For example, Los Angeles - "the angels" - also a city in the US, the product would need to know when to say "the angels" instead of "Los Angeles" when translating to English, in other words, it needs so context-sensitive AI) So, I think that the most amount of investment would need to go into a bit of AI and translation -even though speech recognition still needs a lot of work.

Reply