神奇的变声器：尖叫和怒吼秒变温柔平和的声音 Company creates AI voice filter that turns angry screams into calm speech

在呼叫中心上班，免不了经常面对客户的怒吼和指责，长此以往确实会令人崩溃。为了减轻呼叫中心员工的压力，日本一家公司研发出了一款人工智能变声器，可以检测出愤怒的尖叫声并将其转化为语调温柔的声音。

A Japanese company has developed an artificial intelligence filter that can detect angry screams and translate them into calm speech in order to reduce the stress experienced by call center operators.
一家日本公司研发出一款人工智能变声器，可以检测出愤怒的尖叫声并将其转化为平和的话语，从而减轻呼叫中心操作员感受到的压力。

Large tech companies have long been aware of the stress their call center operators are exposed to every single day, with many customers taking out their anger and frustrations on them. Some of them have even introduced stress management programs that include relaxation techniques, meditation, yoga, or therapy meant to irritability and anxiety. However, one Japanese company may have come up with a much more efficient solution – using AI to completely take screaming and aggressive speech out of the equation. SoftBank claims to have spent three years creating a voice filter that detects screaming and automatically translates it into calm speech.
长期以来，大型科技公司一直都很清楚他们的呼叫中心操作员每天承受的压力，许多客户都把自己的怒火和沮丧发泄到他们身上。一些公司甚至引入了包括放松技巧、冥想、瑜伽或减少易怒和焦虑情绪的理疗在内的压力管理方案。不过，日本软银公司或许找到了一个有效得多的解决方案——运用人工智能将尖叫和攻击性的声音“消除”。该公司声称耗时三年研发出了一款可以检测出尖叫声并将其自动转化为平和话语的变声器。

SoftBank’s voice filter consists of two stages, one in which the AI identifies an angry voice and extracts key points of speech, and a second where it uses acoustic tools to transform it into a more natural, even polite tone. Interestingly, the filter doesn’t change any of the words spoken by the person, but significantly softens the intonation. The call center operator will still hear any insults uttered, only in a soft tone, which should help reduce their stress and anxiety.
软银公司的变声器分两步进行，第一步，人工智能识别出愤怒的声音并提取出讲话要点，第二步，用声学工具将其转化为语调更自然、甚至更礼貌的声音。有趣的是，变声器不会改变说话者说出的任何字词，但是会让这些话的语调变温柔很多。呼叫中心操作员仍然可以听到对方说的任何侮辱性话语，只是听到的语调很柔和，这应该有助于减轻呼叫中心操作员的压力和焦虑情绪。

To train the AI, SoftBank’s engineers asked 10 actors to record at least 100 common phrases including screams, accusations, threats, and demands for an apology. In total, more than 10,000 pieces of voice data were used to train the AI filter.
为了训练人工智能，软银公司的工程师请10名演员录制了至少100个常用短语，包括尖叫、指责、威胁和要求道歉，训练过程中总共使用了超1万条声音数据。

It’s unclear when SoftBank plans to implement the new scream-filtering AI into its call centers, but it will be interesting to see how well it works and how it impacts the emotional wellbeing of operators.
目前尚不清楚软银公司计划何时将这款新的智能尖叫变声器应用到呼叫中心，但是到时候看到其效果以及对呼叫中心操作员情感健康的影响应该会很有趣。

英文来源：Oddity Central

翻译&编辑：丹妮

审校：齐磊、董静