mental_floss magazine
SUBSCRIBE >
GIFT SUBSCRIPTIONS >
DIGITAL SUBSCRIPTIONS >
subscriber services >
The New Scientist Tech blog is reporting that researchers at Carnegie Mellon are developing an exciting new translation device. The gadget uses electrode sensors on your face and neck to essentially lip read your words before relaying them in another language, meaning you don’t even need to whisper to make it work?! The phenomenon is unique because today’s translation devices try to “listen” to your words that you speak out loud, before translating them– you say something, the computer interprets it, then relays it to the other person, making a conversation anything but normal. Instead, the new device tries to fix those gaps in conversation by deciphering and translating the words as you’re mouthing them. The freaky part is that in the future you could have a conversation with someone speaking a different language, and neither of you might ever hear the others’ actual voice! Read more at New Scientist.
Comments (3)
Comment
An actual reading of the piece will uickly set you straight on what nonsense this article is. The underlying research didn’t make anything remotely like this claim.
See http://itre.cis.upenn.edu/~myl/languagelog/archives/003709.html
posted by Jezebel on 10-30-2006 at 8:29 pm
hey Jezebel: I’m always happy to read another view. I’m not sure I understand quite clearly what you’re pointing out, though. I was trying to say that this is a work in progress, and that the device uses the electrodes to understand the words in a new and unique way. The idea sounds fascinating to me. From what I could tell though, the UPENN link posted was basically pointing out the flaws in the system, and how it doesn’t read words accurately yet. I was just pointing out the novelty of the idea. Also, I don’t think the post was disingenous. I think the underlying research does sort of make the claim stated, with the paper you’ve referenced being titled, “Towards Continuous Speech Recognition Using Surface Electromyography.” It seems from a thoroughly non-academics’ view (my own!) to be making that claim. In any case, we’re happy to leave your link up to encourage people to view the evidence for themselves. Diverse opinions are always welcome here.
posted by Mangesh on 10-30-2006 at 9:04 pm
The missing link in the article is the ‘translation’ bit. What the research is trying to do, as the title you quote makes clear, is use electromyography to recognise speech. To recognise speech is not (in itself, anyway) to translate it.
I too find the underlying topic fascinating, and I’m absolutely with you in drawing attention to good research. But let’s not get carried away, nor subscribe to popular press nonsense, about where that research might be leading.
posted by Jezebel on 10-30-2006 at 10:54 pm