Around the 47-minute mark, you suggest that people can get an idea of the quality of a translation by running the translated version through AI again to translate it back into its original language. This is a potentially bad piece of advice - you should only do this experiment if you actually understand both languages! The problem is that the errors made by AI when translating into the second language may be made again (or undone, so to speak) when the AI translates the text back into the first language. So you might think, "ok, the back-translated version sounds about right, so the translation must be okay" but in reality the translation may have various grammar and mistranslation issues. I recently had this situation in my dayjob (as a translator) when I received a piece with various mistranslations, unidiomatic sentences etc. by someone who didn't speak my target language. I wanted to make him understand the problems with the translation, so I thought, "let me quickly run this through the same AI and hopefully the output will sound as odd and confusing as the translated version" - but it didn't. The AI mistranslated the text again and recreated the original version (with just minor differences, as shown on your slide in the video). I suppose it has to do with how the engine works behind the scenes. It doesn't actually understand the text but rather looks for the translation that has the highest possibility of being correct. When it makes the wrong choice in one direction, it's likely to make the same mistake in the other direction. So authors who are trying to get a cheap translation should always - always! - let a native speaker look at it and not just rely on such a back-translation experiment. (You'd think that this doesn't need to be said but I've encountered people who wanted to selfpublish and get their books translated cheaply but had absolutely unrealistic expectations.) This was a very interesting webinar. Thank you! I also just had a quick look at Nuanxed's site, and it seems that they still care about quality, not quantity, despite the emphasis on affordability, so this is refreshing to see. (There are too many big translation agencies whose focus nowadays is only on quantity.) Kudos for that! 🙂
Around the 47-minute mark, you suggest that people can get an idea of the quality of a translation by running the translated version through AI again to translate it back into its original language. This is a potentially bad piece of advice - you should only do this experiment if you actually understand both languages! The problem is that the errors made by AI when translating into the second language may be made again (or undone, so to speak) when the AI translates the text back into the first language. So you might think, "ok, the back-translated version sounds about right, so the translation must be okay" but in reality the translation may have various grammar and mistranslation issues.
I recently had this situation in my dayjob (as a translator) when I received a piece with various mistranslations, unidiomatic sentences etc. by someone who didn't speak my target language. I wanted to make him understand the problems with the translation, so I thought, "let me quickly run this through the same AI and hopefully the output will sound as odd and confusing as the translated version" - but it didn't. The AI mistranslated the text again and recreated the original version (with just minor differences, as shown on your slide in the video).
I suppose it has to do with how the engine works behind the scenes. It doesn't actually understand the text but rather looks for the translation that has the highest possibility of being correct. When it makes the wrong choice in one direction, it's likely to make the same mistake in the other direction. So authors who are trying to get a cheap translation should always - always! - let a native speaker look at it and not just rely on such a back-translation experiment. (You'd think that this doesn't need to be said but I've encountered people who wanted to selfpublish and get their books translated cheaply but had absolutely unrealistic expectations.)
This was a very interesting webinar. Thank you! I also just had a quick look at Nuanxed's site, and it seems that they still care about quality, not quantity, despite the emphasis on affordability, so this is refreshing to see. (There are too many big translation agencies whose focus nowadays is only on quantity.) Kudos for that! 🙂
Fascinating conversation--thank you!