Speech recognition software is sci-fi heavenComment on this story
London - Microsoft now present us with something so magical, so wondrous, that we forgive them almost anything.
A video has popped up on YouTube featuring its chief research officer, Rick Rashid, demonstrating a system whereby his English speech is translated on the fly by computer into a synthesised version of his voice speaking Mandarin Chinese.
As each sentence is translated automatically for the crowd they go wild, envisaging a time where Douglas Adams' fictional Babel fish earpiece becomes a reality.
The company has used deep neural networks to improve the error rate of existing speech recognition by around 15 percent - hugely significant.
Then, using translation software to convert the text, and speech synthesis to modulate the output, on-the-fly translation is miraculously achieved. It's far from perfect; you wouldn't want to use it to settle border disputes between leaders at international summits. But as an onstage spectacle? It's sci-fi heaven. - The Independent