Amazon Productizes NLP via Alexa, Puts ‘Live Translation’ Into Echo

亚马逊通过Alexa实现NLP产品化,将“实时翻译”放入Echo

2020-12-15 19:40 slator

本文共275个字,阅读需3分钟

阅读模式 切换至中文

On December 14, 2020, Amazon launched “Live Translation” for Alexa on Echo smart speakers that have their locale set to “English (US).” The new feature allows Alexa to serve as interpreter and translator for both sides of a two-person convo. It will initially support six language pairs: English and Spanish, French, German, Italian, Brazilian Portuguese, or Hindi. How it works according to Amazon: Echo user asks Alexa to start a session (e.g., “Alexa, translate into Italian”). After the beep, start speaking in either English or Italian. Alexa will then detect which language is spoken and translate each side of the conversation. Users of Echo Show can also read the translation off the touchscreen. To end, simply say, “Alexa, stop.” The same announcement said the new feature was built on existing Amazon tech, such as Amazon Translate as well as Alexa’s automatic speech recognition (ASR) and text-to-speech feature, with machine learning models “designed and optimized” for translating conversational speech. Overseeing Alexa Live Translation is Shirin Saleem, who serves as senior manager for translation at Alexa AI. A data scientist with an MS from Carnegie Mellon, she has worked at Amazon Alexa since 2013, serving in various roles mainly focused on machine learning. Before Amazon, there was Google, which announced it was offering “Interpreter Mode” for Google Nest in December 2019. Interpreter Mode had already been made available for smartphones via Google Assistant back in February 2019. And, in an adjacent, B2B use case from back in August 2019, Slator covered Baidu’s productization of speech-to-speech translation, Baidu Translate. (The latest news on the Baidu page: “AI simultaneous interpreting beta is open” for Chinese-English.)
2020年12月14号,亚马逊在Echo智能音箱上推出了Alexa的“实时翻译”,这些音箱的区域设置为“英语(美国)”。 这项新功使得Alexa能在两人对话中为双方担任口译员和笔译员。它最初将支持六组语言:英语和西班牙语,英语和法语,英语和德语,英语和意大利语,英语和巴西葡萄牙语或英语和印地语。 根据亚马逊介绍Alexa的工作原理是:Echo用户向Alexa发起一个会话请求(例如,“Alexa,翻译成意大利语”)。哔一声后,开始用英语或意大利语讲话。Alexa随后会检测出用户所说的是哪种语言,并对对话的双方都进行翻译。Echo Show的用户还可以通过触摸屏阅读翻译。结束时,简单地说一句:“Alexa,停下来。” 同一声明中也指出,这一新功能建立在亚马逊现有技术基础上,如Amazon Translate技术以及Alexa的自动语音识别(ASR)技术和文本转语音功能技术,通过“设计和优化”的机器学习模型来进行翻译对话语音。 负责Alexa实时翻译的是Shirin Saleem,她是Alexa AI的高级翻译经理。她是一名数据科学家,拥有卡耐基梅隆大学的理学硕士学位,自2013年以来一直在亚马逊Alexa工作,在多个主要专注于机器学习的岗位上任职。 在亚马逊之前,谷歌曾于2019年12月宣布为谷歌Nest提供“口译员模式”。早在2019年2月,Google Assistant就已经为智能手机提供了口译模式。 此外,类似的有,在2019年8月的一个B2B使用案例中,Slator涵盖了百度语音转语音翻译的产品,即百度翻译。(百度页面最新消息“AI中文转英文同声传译测试第二版开放”。)

以上中文文本为机器翻译,存在不同程度偏差和错误,请理解并参考英文原文阅读。

阅读原文