Is machine translation better than humans?
*MT (Machine Translation), did you know that machine translation is a technology that is so old that it even existed in the 1950s? However, since there are so many mistranslations, I didn't use it very often in the past. However, as time goes by, the results of machine translation get better, and in recent years, the development of AI technology has greatly improved the quality of machine translation statements. So is machine translation now better than the results of human translation?
Limitations of machine translation
Of course, AI has helped advance machine translation, but it cannot be said that the quality is unconditionally better than human translation. As an example, let's translate the original text of the article below using a translator in a search engine.
Original text example: Last year King Seth Frankoff was a starter and struggled with 4 innings, 3 hits, 2 strikeouts, and no run, then became a candidate for the 4th round Lee Young-haGa climbed in 3 innings, 3 hits, 2 strikeouts 1 goal (visa book)It was recorded.
MT results: Last year, Seth Frankoff, the king of many wins, started and pitched 4 innings, 3 hits, 2 strikeouts, and noruns. Lee Young-ha, a candidate for the 4th starting pitcher, took the mound and recorded 3 hits, 2 strikeouts and 1 run in3 innings.
If you actually copy and paste it into the translator, the translation will be completed in less than a second. You can see mistranslations everywhere in the completed translation.
First, 'Last year'The meaning of 'is not clearLast year'It doesn't mean To 'Last Year' It's been translated. Also, the notation for English and Korean is different 'Lee Young-ha' The positions of the first and last name are reversed, and the third line's (Visa book)has been removed from the content. It's a sports article that must be strictly factual, so if it were translated like this, a big error would occur, right?
At the end of the day, we need a human touch.
These mistranslation shows that there are still translations that machines can't do. No matter how good a translator is at learning, it can't fully understand a situation as well as a human being. Even machine-translated texts require a process of improvement through human hands.
In this way, the process of a human directly modifying a machine translation MTPE (Machine Translation Post Editing), Machine translation post-editingIt's called.
MT results: Last year, Seth Frankoff, the king of many wins, started and pitched 4 innings, 3 hits, 2 strikeouts, and noruns. Lee Young-ha, a candidate for the 4th starting pitcher, took the mound and recorded 3 hits, 2 strikeouts and 1 run in3 innings.
MTPE: Seth Frankoff, Last year's most wins by Apitcher, started and pitched 4 innings, 3 hits, 2 strikeouts, and noruns. Young-ha Lee, a candidate for the 4th starting pitcher, took the mound and recorded 3 hits, 2 strikeouts and 1 run (unearnedrun) in 3 innings.
By entrusting the areas that machines are good at and translating only the necessary parts directly, humans can fill in the deficiencies of machine translation and make better translations.
Two ways to make more accurate translations with MTPE
First, it's easier to write the original text. Writing the original text with easy-to-understand words and concise sentences can produce better MT results. If the original text already exists, proofreading the original text beforehand can reduce the hassle of post-editing.
The second is to utilize **TM/TB. You can reduce the scope of post-editing by applying pre-built TM/TB. Consistent translations are possible by borrowing the memory of computers that are smarter than humans.
Although machines cannot completely replace humans, the field of MTPE is expanding as it has replaced a large part of it.
Recently, various translation industry workers, such as translation companies, translators, and agencies, are working through MTPE. This is because I think translating with MTPE is more productive than translating by humans all at once. They are even translating at a lower unit price.
LETR WORKS (LETR WORKS) is a translation tool that supports MTPE to help users improve translations more easily. Machine-translated sentences can be easily edited within the editor. Enterprise plans also support TM/TB functionality. Try Letterworks for free!
(TM/TB is a separate product, not a feature included in the plan.)
Experience MTPE and LETR WORKS (Letter Works), where translations become more accurate.
*MT (Machine Translation): https://en.wikipedia.org/wiki/Machine_translation
**TM/TB: https://www.letr.ai/blog/insight-230118
Reference: http://www.sportsseoul.com/news/read/738495
Also see other articles by Letterworks
How to make translation easier with TM/TB
Large-scale translation projects, why collaboration apps are needed
Are you still copying? Why I don't have a favorite translator
