Title:
DEEP-LEARNING TRANSFORMER TRANSLATION SYSTEM USING TRIPLE CORPUS
Document Type and Number:
WIPO Patent Application WO/2022/196956
Kind Code:
A1
Abstract:
A deep-learning transformer translation system using a triple corpus creates corrected data through draft translation, reinforcement learning using a deep-learning function, and corrected translation, and economically reflects the emotion, nuance, atmosphere, slang, tone, author intention, context or the like in literature, webtoons, or video subtitles that can be commercially distributed and exported globally, thus having the effect of improving the quality of translation to such a high level that a reader can actually read and consume beyond simple conveyance of meaning. The deep-learning transformer translation system using a triple corpus can automatically produce commercially distributable final versions 24 hours a day without restriction, and reduces the number of or eliminates the need for requests for corrections and checks that are required steps for the completion of a final translation through a process of achieving a singularity (meaning: the final evolution and completion of artificial intelligence).
Inventors:
NAM JI WON (KR)
Application Number:
PCT/KR2022/002275
Publication Date:
September 22, 2022
Filing Date:
February 16, 2022
Export Citation:
Assignee:
INSIGHT VESSEL CO LTD (KR)
International Classes:
G06F40/51; G06F9/451; G06F40/58; G06N3/08
Foreign References:
| KR20200034012A | 2020-03-31 | |||
| KR20170052974A | 2017-05-15 | |||
| JP2000259631A | 2000-09-22 | |||
| KR20200017600A | 2020-02-19 | |||
| US20080288240A1 | 2008-11-20 | |||
| KR102306344B1 | 2021-09-28 |
Attorney, Agent or Firm:
KWAK, Chul Keun (KR)
Download PDF: