Go to contents


Korean researchers develop 9-language voice recognition technology

Korean researchers develop 9-language voice recognition technology

Posted April. 19, 2017 07:18,   

Updated April. 19, 2017 07:25


Korean researchers have developed a multi-language voice recognition technology that recognizes nine languages automatically and convert them into words. This is expected to become a source technology for the development of various services including automatic interpretation, artificial intelligence assistants, real time movie captions and unmanned call centers.

The Electronics and Telecommunications Research Institute said Tuesday that its research team led by Kim Sang-hoon developed a language voice recognition device for nine languages, with a ramped up performance.

Voice recognition technologies are already used on smartphones including in Apple's Siri and Samsung's Bixby. Given that a 90-percent recognition capability is still inconvenient for many people, the latest technology is crucial as it has reduced errors further by increasing recognition capability to 95 percent. In addition, it supports nine different languages including Korean, English, Japanese, Chinese, French, Spanish, German, Russian and Arabic. When it comes to daily conversation, this technology enables constant recognition to convert in real time voices into words. "In the case of Korean language, it excels voice recognition technologies of world prominent firms, while the technology for other languages is on an equal footing," the research team said.

The researchers used Ai technology of "deep learning" to automatically analyze voice data (big data) to raise performance levels. The device will be released in two types of "client server," which can be used with Internet connection, and an "embedded" type, which can be installed in mobile devices like smartphones. The former has great performance ability, while the latter is convenient as it can be used when there is no Internet connection.

The institute started test services of this technology in its previous automatic interpretation program called "Genie Talk." It plans further to make an interpretation service for the upcoming winter Olympics by working with Hancom, the official sponsor of the Pyeongchang Winter Olympics' automatic interpretation and translation service. It also plans to increase the number of applicable languages to 14 by 2020 when the Tokyo Olympics is held, and ultimately hopes to support at least 20 languages in total.