可以知道你喜怒哀乐的手机应用

Beyond Verbal Communications Ltd., a voice-recognition software developer here, is rolling out an app promising something Siri can't yet deliver: a readout on how you feel.

语音识别软件开发商Beyond Verbal Communications Ltd.即将推出一款应用软件，有望实现Siri尚且无法实现的一个功能：把你的情绪显示出来。

Called Moodies, it lets a smartphone user speak a few words into the phone's mike to produce, about 20 seconds later, an emotional analysis. Beyond Verbal executives say the app is mostly for self-diagnosis -- and a bit of fun: It pairs a cartoon face with each analysis, and users can share the face on social media.

借助于这款名为“Moodies”的应用，智能手机用户可以朝着手机的麦克风讲话，在大约20秒钟之后生成情绪分析。Beyond Verbal的管理人员说，该应用主要是用于自我诊断，也可以带来一些小小的乐趣：它给每一次分析配上一张卡通脸孔，用户可以把脸孔拿到社交媒体上去分享。

But the app is coming out as the company and other developers -- many clustered in Tel Aviv -- push increasingly sophisticated hardware and software they say can determine a person's emotional state through analysis of his or her voice.

在这款应用面世之际，Beyond Verbal和其他一些开发商――很多都扎堆特拉维夫――正在推出一些越来越尖端、据它们说可以通过分析语音确定一个人情绪状态的硬件和软件。

These companies say the tools can also detect fraud, screen airline passengers and help a call-center technician better deal with an irate customer. And they can be used to keep tabs on employees or screen job applicants. One developer, Tel Aviv-based Nemesysco Ltd., offers what it calls 'honesty maintenance' software aimed at human-resource executives. The firm says that by analyzing a job applicant's voice at an interview, the program can help identify fibs.

这些公司说，这些工具还可以侦测欺诈、检查飞机乘客、帮助呼叫中心技术人员更好地对付发飙的顾客。它们可以用来监视员工或筛选求职者。特拉维夫开发商Nemesysco Ltd.推出以人力资源经理为目标客户的“诚信维护”软件。该公司说，这套程序可以通过分析求职者在面试期间的说话声音来帮助辨别谎言。

That's raising alarm among many voice-analysis experts, who question the accuracy of such on-the-spot interpretations. It's also raising worries among privacy advocates, who say such technology -- especially if it is being rolled out in cheap, easy-to-use smartphone apps -- could be a fresh threat to privacy.

这引起了很多语音分析专家的警惕，他们对这类现场转译的准确性提出了质疑。也引起了很多隐私鼓吹者的忧虑，他们说，这类技术有可能成为一种新的隐私威胁，如果是以便宜好用的智能手机应用的形式推出，威胁就更为严重。

Depending on how the analysis is performed, used and shared, 'there could well be breaches of certain privacy laws,' says Gwendolen Morgan, an associate at Bindmans LLP, a London human-rights law firm.

伦敦人权律师事务所Bindmans LLP合伙人格温德琳・摩根(Gwendolen Morgan)说，这些分析“很有可能违反了某些隐私法规”，具体要看它们是怎样运行以及被使用和共享的。

The new wave of technology is based on so-called layered voice analysis, and it's related to the much broader, more established field of 'speech-to-text' sentiment analysis. Verint Systems Inc., Thomson Reuters PLC and Hewlett-Packard Co. and others have long used speech-to-text technology to record phone calls and break them down into so-called text-based sentiment intelligence by flagging the occurrence of keywords or types of words. Call centers use the data to teach employees to keep customers on the phone or monitor employees for training purposes.

这一波新的技术浪潮基于“深层语音分析”(layered voice analysis)，跟更广泛、更成熟的“语音转文字”(speech-to-text)情绪分析领域有关。Verint Systems Inc.、汤森路透(Thomson Reuters PLC)、惠普(Hewlett-Packard Co.)等公司曾长期使用语音转文字技术来记录通话，并把出现在其中的关键词或某些词语种类标注出来，从而将通话分解为“基于文本的情绪情报”。呼叫中心用这些数据教员工如何让客户保持通话，或为了培训的目的而监测员工。

The new speech-focused tools come as other companies are marketing body-language and facial-recognition sentiment-analysis tools -- including an app for Google Inc.'s Google Glass.

在这些新的语音分析工具面世之际，其他公司也在销售身体语言和面部识别类的情绪分析工具，比如一款针对谷歌公司(Google Inc.)“谷歌眼镜”(Google Glass)的应用。

Voice-recognition and analysis specialists say there is no question that emotions and patterns of speech can be linked. But many say the utility of the analysis can be limited by the extent to which voice samples have been collected -- often requiring years of samples to detect variations caused by emotions.

语音识别与分析专家说，情绪和语音形态可能存在关联，这一点勿庸置疑。但很多人说，这类分析的实用性可能受制于语音样本的采集规模，常常需要采集多年的样本才能侦测到情绪造成的变异。

Andrew Baron, assistant professor of psychology at the University of British Columbia in Vancouver, says information produced by a lot of the commercially focused voice-recognition technology can be 'fuzzy.'

温哥华英属哥伦比亚大学(University of British Columbia)心理学助理教授安德鲁・巴伦(Andrew Baron)说，很多以商业为重的语音识别技术所产生的信息都有可能是“失真”的。

'We simply don't have the technology today, at the level of cognitive neuroscience, to really know the precise content of a person's thought or emotion,' Mr. Baron says.

巴伦说：“想要真正知道一个人思想或情绪的准确内容，我们今天根本就没有在认知神经科学的层面掌握这样的技术。”

Israel has become an epicenter of the new technology largely because of its role as an incubator for security-focused technology. Nemesysco markets to security companies and law-enforcement agencies, as well as insurance companies and other corporate clients.

以色列之所以成为这类新技术的中心，主要缘于它作为安全技术孵化器的角色。Nemesysco的目标客户有安保公司、执法机构，也有保险公司和其他一些企业客户。

Britain's Department of Work and Pension bought some of Nemesysco's software to help it detect benefit fraud in 2007, but it says it discontinued the contract after three years after inconclusive results.

英国就业及退休保障部(Department for Work and Pension)在2007年购入Nemesysco的一部分软件，来帮助它侦测骗取福利的行为。但该部表示，三年过后因为检测结果不确定，便中止了合同。

Nemesysco founder Amir Liberman says his firm has come up with 120 vocal parameters that correlate with human emotions. Using algorithms, Mr. Liberman says the technology can pick up 'patterns and abnormalities' and classify them as related to emotional states like stress, excitement and confusion. He says the technology's benefits far outweigh any perceived infringement on personal privacy.

Nemesysco创始人阿米尔・利伯曼(Amir Liberman)说，他的公司已经确定了120个与人类情感相关的语音参数。利伯曼说，通过算法，这项技术可以甄别出一些“特定的模式和异常情况”，进而确认它们是否与紧张、兴奋、疑惑等情绪状态有关。他说，该技术的好处远远胜过一些人眼中个人隐私受到的侵犯。

EmoSpeech, a startup based in Puebla, Mexico, and Cambridge, Mass., sells its voice-analysis product to call centers in Mexico. It promises an ability to identify four basic emotional states: happiness, anger, impatience and neutrality. The company is targeting call centers in the U.S., says Chief Executive Miriam Reyes.

总部在墨西哥普埃布拉(Puebla)和美国马萨诸塞州坎布里奇(Cambridge)的初创公司EmoSpeech向墨西哥的呼叫中心出售其语音分析产品。它承诺能够辨别出快乐、愤怒、烦躁、中性这四种情绪状态。EmoSpeech首席执行长米丽娅姆・雷耶斯(Miriam Reyes)说，公司正在面向美国的呼叫中心进行推销。

Beyond Verbal, the company behind the Moodies smartphone app, says its software analyzes speech components like timing, energy, frequency and spectral content to produce data that can then be run through its algorithms.

推出Moodies应用的Beyond Verbal公司说，该软件通过分析节奏、能量、频率和频谱内容等语音成分来生成数据，然后用自己的算法来进行加工。

Founded in 2017, Beyond Verbal's main business is selling layered-voice-analysis software to companies in the call-center business. The company recently closed a $2.8 million funding round led by prominent Kazakhstan-based angel investor Kenges Rakishev.

Beyond Verbal成立于2017年，主要业务是向做呼叫中心业务的企业出售深层语音分析软件。最近该公司完成了一轮280万美元的融资，牵头的是哈萨克斯坦知名天使投资家肯格斯・拉基谢夫(Kenges Rakishev)。

Dan Emodi, Beyond Verbal's vice president for marketing and strategic accounts, says the software is based on more than three years of research. Based on user feedback, he says, the smartphone app has an accuracy rate of 80%.

Beyond Verbal负责营销与战略客户的副总裁丹・埃默迪(Dan Emodi)说，该软件基于三年多的研究。他说，从用户反馈判断，这款智能手机应用软件有80%的准确率。

As for any privacy concerns, he says, 'Our responsibility is the people we work with, the partners we choose.' But he does acknowledge there 'is no technological way for us to prevent anyone from taking our product and activating it on somebody else without their knowledge.'

他说，至于隐私方面的忧虑，“我们的负责对象是一起合作的人、我们选择的合作伙伴”。但他也承认，“从技术上讲，我们没有办法阻止任何人拿着我们的产品在别人不知情的情况下在他们身上使用它”。