Why artificial intelligence translation (machine translation) is inconvenient for video subtitling
Getting started
If you've ever had trouble combining phrases into a sentence when subtitling a video with machine translation, pay attention to this note. I would like to introduce an application along with Letterworks' unique text assembly function.
- Sentence assembly: AI sentence paraphrase recognition function that corrects missegmented sections when parsing sentences in content
What happens when video captions are turned to AI
What happens when videos created with hard work are automated with AI transcription? As shown in the example below, you can see that when Hangul CC subtitles are turned on on on YouTube, the words are displayed separately according to the firing speed.
Source: Gicon Studio YouTube
There is a disadvantage that it is difficult to grasp the full context when checking the content with only video subtitles without sound when the voice is transcribed with a time difference according to the timing of the utterance. On the other hand, if the subtitle falls smoothly into a sentence, you can understand the speaker's lines at a glance, which is convenient. This kind of phrase separation is a common phenomenon when an AI program is used for subtitling.
Text assembly, paraphrase recognition function
The Letterworks team also often faced a phenomenon where subtitles were separated when translating various video content such as dramas, entertainment, lectures, and YouTube. If the original text was cut off from the transcription before translation, the work of matching the sentences had no choice but to take a lot of time.
In order to reduce the inconvenience that occurs in the business, Letterworks has introduced a text assembly function directly. As shown below, you can see the completed assembled text parsed in the [original text] area on the left.
Afterwards, I looked at the same section in the video as a translation project. Based on the original text, well-assembled sentences with a high degree of perfection were recommended even in the translated text.
Sentence assembly was applied to parsing not only English and Korean, but also various foreign languages such as Chinese and Japanese, making it possible to improve the quality of machine translation to the fullest. A well-finished original text leads to a high-quality translation.
Letterworks' research to improve the efficiency and quality of video translation work will continue, so stay tuned!
Good content to watch together
🔗 Video translation overview #1 6 steps to complete video translation
🔗 Video translation overview #2 The more complicated it is, the more people you work with, the more important security is.
🔗 Video translation overview #3 How are SDH, CC, and Subtitle used separately in the industry?
🔗Video translation overview #4 How does Subtitling for the Deaf (SDH) work?
🔗Video translation preview #6 Translation that makes use of character speech and speech
Editor l Kim Bomi
bomi.kim@twigfarm.net
