What is Alibaba AI Real-time Translation? Alibaba AI Real-time Translation Review, Features

On May 18, 2018, Alibaba officially launched the real-time translation function in cross-border e-commerce to solve the language barrier between buyers and sellers. It is reported that this is a real-time translation AI (artificial intelligence) product in the global e-commerce field. With a daily translation capacity of 750 million times, it far exceeds Facebook and Amazon.

About Alibaba AI Real-time Translation

The AI ​​real-time translation tool developed by Alibaba provides translation services in 21 popular languages ​​around the world, including 43 language directions. For example, if a product is published in English, it can be automatically translated into 16 languages ​​in real time. In 2017, a total of 300 billion translation services were provided to global customers. As of April 2018, Alibaba AI real-time translation has provided services for up to 700 million commodity scenarios, supporting everything from website building to commodity search, as well as a variety of optimization tools. For example, product titles, details, comments, etc. can all be translated online in real time to meet the information reading needs of global users when browsing commodity services in different countries.

For languages ​​with rich morphology such as Russian, the algorithm model created by the technical team based on the neural network algorithm can better solve problems such as "word form" errors. The distributed neural network translation model training platform can learn hundreds of millions of corresponding translation sentences within two days, and then learn a new language.

use

Alibaba's real-time translation system can translate a sentence in just 200 milliseconds, a time difference that humans cannot perceive. In addition to being fast, this translation AI is like a human, and can automatically correct errors based on contextual semantics and grammatical understanding. It is even smarter than humans, and can detect in one second that a user is communicating in a non-native language, and automatically remind the user to turn on the real-time translation function.

Luo Weihua, a technical expert at Alibaba DAMO Academy, introduced that this real-time translation system for dialogue communication uses the industry's new neural network model and integrates multiple AI technologies of DAMO Academy, such as machine translation, semantic understanding, automatic language recognition and speech recognition. When the system was launched, it only supported four languages: Chinese, English, Russian and Spanish. However, they expect to expand the translation service of small languages ​​such as Thai, Arabic, Vietnamese and Turkish in the second half of the year. In addition to real-time text translation, Alibaba also plans to officially launch the real-time voice translation and automatic generation of bilingual subtitles for live video in the second half of 2018. When launched, cross-border merchants can communicate directly through video.

Specific applications of AI scenarios in intelligent translation in the field of cross-border e-commerce

"In the cross-border e-commerce scenario, the biggest gap is not distance, but language barriers," said Shi Yangbin, head of the intelligent translation solutions direction of the Machine Intelligence Technology Laboratory of Alibaba DAMO Academy. In today's globalized world, cross-border e-commerce business is spread all over the world, but due to language communication barriers, many sellers are unable to accurately capture the information that the other party wants to express when communicating with buyers.

1. Language complexity of cross-border e-commerce business

"The language problem is a big problem in the cross-border e-commerce field," Shi Yangbin pointed out. As shown in the above figure, the 18 major cross-border e-commerce countries are distributed in different regions, and the sellers in the above figure can be understood as different original products, including English, Russian, Spanish and other original products. These different sellers sell goods to buyers from 18 different countries on the left. Furthermore, the e-commerce platforms are below, including Alibaba's internal and other e-commerce platforms in the fields of logistics, payment, and marketing. Different fields require different languages, which shows the complexity of the language business of cross-border e-commerce.

2. Application of intelligent translation in the entire chain of cross-border e-commerce

(1) When sellers conduct cross-border business, they need to use multilingual localized websites or apps, which involves diversification of website copywriting, international standard scanning, localization testing, and localization of network regulations and website information.

(2) After the seller has a localized site, the next step is to attract new customers to the APP or website, which involves the diversification of traffic-generating keywords and titles, the diversification of creative copywriting, and the diversification of categories.

(3) After sellers direct users to their own websites, they need to consider how to increase the conversion rate of site visits, which involves multiple searches, multiple categories, etc.

(4) When users successfully find the products they want to buy, you need to consider how to improve the user's purchase conversion rate. For example, the conversion rate from D to O in the above figure requires diversification of all product details, including titles, comments, and UGC content.

(5) The final successful order by the user also involves the logistics payment link, which includes the payment logistics and the resolution of problems in the customs inspection process. When the goods arrive in the hands of the buyer, the seller needs to consider how to increase the retention and repurchase rate, including multiple real-time communication and intelligent customer service.