Flitto, which is a startup company in South Korea, is exporting translation data for Chinese and Japanese to Baidu, which is the biggest search portal in China. These data will be used to enhance ‘Baidu translate’, which is a translation service for Chinese. Flitto is planning to expand its global partnership based on its platform that covers both AI (Artificial Intelligence) translation and human translation.
Flitto announced on the 25th that it signed off on a contract with ‘Baidu NLP’ during middle of June and that it will be providing Japanese-Chinese corpus. Baidu NLP is a department that is responsible for Baidu translate service which is being used the most in China. Data that is exported will be used to enhance Baidu translate service. Flitto and Baidu NLP will continue their close cooperation through Flitto’s Chinese branch.
It was important that Flitto has enormous amount of translation data with good quality. Translation data that it has accumulated for 5 years is 80 million pairs. It is almost impossible to easily collect translation data due to copyright because one has to consider a translator’s copyright on translations and a requestor’s copyright on originals. This is why there are not too many companies that supply translation data.
Portal sites such as Google and Naver manage participating translation sites and collect translation data after receiving agreements from users. Motivation to participate in translation is weak without compensation. Flitto manages collective intelligence-based online translation platform that pays translators who will provide translations when users upload originals. Through money transaction, it was able to solve both copyright problems and problems regarding quality of translation.
Flitto recently added AI translation function that is applied with NMT (Neural Machine Translation). When users, who want free translation service, are not satisfied with AI translation service, they can immediately request actual translators to translate their original documents. Through this, Flitto continuously receives difficult translations such as new words, technical words, slang, and regional dialect that are hard for machines to understand.
“Baidu is very satisfied with our corpus that Baidu did not correct single corpus.” said CEO Lee Jung-soo of Flitto. “We have implemented a platform that covers human translation and machine translation and it is possible for us to continuously secure data with good quality.”
Value of use of language data is becoming bigger as language data is a necessary element for development of AI which is one of the major keys for Industry 4.0. Entire processing process of translation data is a necessary element to implement search and interpretation and translation technologies. By supplying corpus to Baidu, Baidu will accelerate its cooperation with global companies and is currently supplying its data to variety of foreign companies such as Microsoft and NTT Docomo. Besides its Chinese branch, it is looking to establish Japanese branch in order to work with Japanese companies. It also provides translation service for Arabic and is receiving positive reactions from Saudi Arabia and Morocco from the Middle East.
“Flitto will be the company that will be at the front line of Industry 4.0 and produce language data that is necessary to develop AI.” said CEO Lee. “We are going to develop a platform that continuously provides data with good quality by connecting variety of translation and AI services from all over the world.”
Staff Reporter Oh, Daeseok | ods@etnews.com