谷歌语言交互新突破 能更逼真模拟人声

 　　Google's DeepMind have revealed a new speech synthesis generator¹ that will be used to help computer voices, like Siri and Cortana, sound more human.

　　谷歌旗下的人工智能公司DeepMind近日研制出了一种新型语音合成系统, 该技术可以让如Siri和Cortana这样的计算机合成语音听起来更接近真实人声。

　　Named WaveNet, the model works with raw audio waveforms to make our robotic assistants sound, err², less robotic.

　　这项名为WaveNet的技术通过研究原始音频波形，使机器人助手的声音听起来不那么像机器人。

　　WaveNet doesn't control what the computer is saying, instead it uses AI to make it sound more like a person, adding breathing noises, emotion and different emphasis into senteneces.

　　WaveNet并不会控制计算机的说话内容，它只会应用人工智能技术在句子中添加呼吸声、情感和各种重音，从而使计算机语音听起来更像真人。

　　Generating speech with computers is called text-to-speech (TTS) and up until now has worked by piecing together short pre-recorded syllables³ and sound fragments to form words.

　　用计算机合成语音的技术叫做“从文本到语音（TTS）”，现存的工作原理是将提前录制好的短音节和声音碎片合成语言。

　　谷歌语言交互新突破 能更逼真模拟人声

　　As the words are taken from a database of speech fragments, it's very difficult to modify the voice, so adding things like intonation⁴ and emphasis is almost impossible.

　　由于语言是从语音碎片数据库中提取出来的，声音很难修饰，所以几乎不可能添加声调和重音等因素。

　　This is why robotic voices often sound monotonous⁵ and decidedly different from humans.

　　这就是为什么机器人语音听起来很生硬，明显和人声不同。

　　WaveNet however overcomes this problem, by using its neural⁶ network models to build an audio signal from the ground up, one sample at a time.

　　然而WaveNet克服了这个难关，利用神经元网络模型从头建立一个音频信号，每次生成一个样本。

　　During training the DeepMind team gave WaveNet real waveforms recorded from human speakers to learn from.

　　培训期间，DeepMind团队让WaveNet学习了一些真实记录的人类语音波形。

　　Using a type of AI called a neural network, the program then learns from these, much in the same way a human brain does.

　　通过一种叫做神经元网络的人工智能技术，这个系统可以像人类的大脑一样对这些波形进行学习。

　　The result was that the WaveNet learned the characteristics of different voices, could make non-speech sounds, such as breathing and mouth movements, and say the same thing in different voices.

　　所以WaveNet学习了不同声音的特点，可以发出非语言声音，比如呼吸声和嘴部活动的声音，并且可以用不同的声音说同样的内容。

　　Despite the exciting advancement⁷, the system still requires a huge amount of processing power, which means it will be a while before the technology appears in the likes of Siri.

　　虽然这个系统有激动人心的进步，但是它需要很强大的处理能力，这意味着这项技术并不能很快应用到Siri当中。

　　Google's machine learning unit DeepMind is based in the UK and have previously⁸ made headlines when their computer beat the Go World champion earlier this year.

　　Google旗下的机器学习技术企业DeepMind总部设在英国，今年早些时候，他们的计算机因打败了围棋世界冠军而上了头条。

1 generator   

n.发电机，发生器

参考例句：

• All the while the giant generator poured out its power.巨大的发电机一刻不停地发出电力。
• This is an alternating current generator.这是一台交流发电机。

2 err   

vi.犯错误，出差错

参考例句：

• He did not err by a hair's breadth in his calculation.他的计算结果一丝不差。
• The arrows err not from their aim.箭无虚发。

3 syllables   

n.音节( syllable的名词复数 )

参考例句：

• a word with two syllables 双音节单词
• 'No. But I'll swear it was a name of two syllables.' “想不起。不过我可以发誓，它有两个音节。” 来自英汉文学 - 双城记

4 intonation   

n.语调，声调；发声

参考例句：

• The teacher checks for pronunciation and intonation.老师在检查发音和语调。
• Questions are spoken with a rising intonation.疑问句是以升调说出来的。

5 monotonous   

adj.单调的，一成不变的，使人厌倦的

参考例句：

• She thought life in the small town was monotonous.她觉得小镇上的生活单调而乏味。
• His articles are fixed in form and monotonous in content.他的文章千篇一律，一个调调儿。

6 neural   

adj.神经的，神经系统的

参考例句：

• The neural network can preferably solve the non- linear problem.利用神经网络建模可以较好地解决非线性问题。
• The information transmission in neural system depends on neurotransmitters.信息传递的神经途径有赖于神经递质。

7 advancement   

n.前进，促进，提升

参考例句：

• His new contribution to the advancement of physiology was well appreciated.他对生理学发展的新贡献获得高度赞赏。
• The aim of a university should be the advancement of learning.大学的目标应是促进学术。

8 previously   

adv.以前，先前（地）

参考例句：

• The bicycle tyre blew out at a previously damaged point.自行车胎在以前损坏过的地方又爆开了。
• Let me digress for a moment and explain what had happened previously.让我岔开一会儿，解释原先发生了什么。