Alexa uses Amazon Translate to reach more international customers

Amazon Alexa is available in 15 locales and eight languages. To understand and respond in different languages, Alexa needs to learn new grammar rules, and the content that powers Alexa needs to be translated to new languages. Additionally, Alexa needs to learn about country-specific topics, such as new soccer leagues, regional celebrities, and important historical events.

This post describes how Alexa uses Amazon Translate to understand a multitude of questions in different languages and provide quick and meaningful replies.

What is Amazon Translate?

Amazon Translate is a neural machine translation service that delivers fast, high-quality, and affordable language translation. Neural machine translation is a form of language translation that uses deep learning models to deliver accurate and natural sounding translation. For more information about the languages Amazon Translate supports, see supported Languages.

What is Alexa?

Alexa is Amazon’s cloud-based voice service and is available on hundreds of millions of devices from Amazon and third-party device manufacturers. With Alexa, you can build natural voice experiences that offer your customers a more intuitive way to interact with the technology they use daily.

How does Alexa use Amazon Translate?

Alexa gets new types of questions every day. Teaching Alexa to recognize a single intent might require hand-authoring a list of utterances that invoke that intent. For more information, see Best Practices for Sample Utterances and Custom Slot Type Values.

For example, the following code contains several utterances listed for the intent GetStockPrice:

I want to know the stock price of {company} 
What about {company} stock
I want to know {company} stock 
What is the stock of {company}
How much does {company} stock sell for
What is the stock market quote for {company}
… (several more)

Those lists don’t scale well across languages. In fact, native speakers who have subject matter expertise and are familiar with country-specific expressions often need to localize utterance lists. Not only is this cumbersome, it results in gaps between English and non-English understanding capabilities.

Alexa uses Amazon Translate to address this issue. Instead of manually translating thousands of utterance lists offline, Alexa uses the real-time TranslateText API to translate utterances instantly and on demand. If an utterance in one language doesn’t map to a given intent, Alexa translates the message to English and sends it again for a second attempt. For example, if you don’t include “¿A cuánto cotizan las acciones de Amazon?” to the Spanish utterance list for the GetStockPrice intent, the second attempt uses the translated English phrase “How much does Amazon’s stock sell for?” When the utterance matches a translation, Alexa can identify the intent, process the utterance, and return a meaningful response.

Because Amazon Translate supports every language Alexa is available in, Alexa could roll this enhancement out globally in a matter of days. Alexa’s capability to answer questions increased across all languages. The effect was particularly pronounced in languages that Alexa only recently learned, such as Hindi and Portuguese, because newer languages don’t have a large list of utterances. Amazon Translate enables Alexa to fill this gap and understand a multitude of questions across all languages.

Conclusion

Amazon Translate has helped Alexa become world-wise by enabling Alexa to serve a larger variety of languages and respond to varied questions better. For more information about the capabilities of Amazon Translate, see Amazon Translate documentation.

For more information about privacy protection in Alexa, see Alexa Privacy.


About the authors

Heike Schirmer is a Sr. Manager at Amazon Alexa. She is part of Alexa Information focusing on making Alexa the world’s most knowledgeable assistant.

 

 

 

 

 

View Original Source (aws.amazon.com) Here.

Leave a Reply

Your email address will not be published. Required fields are marked *

Shared by: AWS Machine Learning

Tags: