Naver introduced Papago in the form of a mobile app in August 2016 in beta testing to showcase its machine learning-based translation algorithms to users, collect more data and receive broader feedback.
The translation program was initially offered only as a mobile application and could handle only 200 words per translation query, limiting users to translating short sentences on mobile devices.
Now launching Papago as an official service, the smart translation program’s word handling limit has been raised to 500 words -- the same word limit for Google Translate -- while a PC version of the app has been introduced as well.
Like the latest version of Google Translate, Papago relies on a new, AI-based computing algorithm called neural machine translation to deliver more accurate and natural translations.
Similar to the human brain, NMT simultaneously considers the different ways to translate a sentence and arrives at the most appropriate outcome based on context learned by studying vast database of translated materials.
It fundamentally differs from the conventional phrase-based translation system that interprets individual words and phrases independently within a sentence and rearranges them, often resulting in awkward mistranslations.
Looking ahead, Naver is preparing to add a number of new languages to Papago — Vietnamese, Taiwanese, Indonesian and Thai — by the end of the year to expand its user base.
Papago currently supports translations between Korean, English and simplified Chinese via the NMT method. It also offers translations between Korean and Japanese, Spanish and French, though they are based on the phrase-based translation system for now.
By Sohn Ji-young (firstname.lastname@example.org)