Many people who see nothing wrong with machine translating have criticized a few translators for their “elitist” attitude. I disagree with some of their arguments, but I think they are right to denounce translation that is done by decoding fragments of stuff they do not know about and gluing them together. When words and meanings are lost in translation, labeling that work as an English version of the original content is wrong.
Since the late trend is to discuss about machine translation, I thought I might as well chime in.
Note: This does not reflect the thoughts of any of the staff members that work for Heretic.
When I joined the English LN community back in 2010 – the LN community being Baka-Tsuki – so I could get to know more about these books, I decided to put my skills to use. Back then, I was just a 13-year-old high school student, and I had very poor knowledge of both English and Chinese. I mean, it was a language I learned mostly by playing Runescape and watching Harry Potter. I still decided to give it a try, because it sounded like a fun experience.
My early translations were disgustingly horrible, as you would expect. I kept the original sentence structure because I didn’t bother giving flow to my sentences, and made some basic mistakes such as keeping the original punctuation, translating literally, attempting to simply do word-to-word translation, etc.
Still, I have not once being tempted to do machine translating. For one, I went websites such as Google Translate with the intent of using it as a tool similar to a dictionary. I quickly noticed however that it filters out all subtlety the text brought and makes some basic mistakes as well.
Another reason why I did not fall in the habit of machine-translating is that it was highly, highly discouraged by my peers at that time. The “Preview” tag for machine translations can also be read as “Garbage” for many. Falling into the habit of using machines to translate meant was an absolute no-go, especially as a new translator. The way to go was learn, translate, learn by translating, learn, translate, etc.
In any case, the re-apparition of machine translation started with EnigmaticAxiom’s work on Madan no Ou to Vanadis. It has been hyped before the translations were posted by the new guard of LN lovers that were thirsty for specific types of novels, novels that include a certain amount of fanservice (more than just panty shots), a diverse harem of strong girls (more than just tsunderes), a main character that is rather strong or powers up quite fast, etc. So they tried Enigmatic’s work and found out that his work was a readable machine translation. Regardless of translation accuracy, readers loved his work, since it could be read like an actual novel. Way better than reading novel summaries, which were (and may still be as of today) the trend.
Looking back, EnigmaticAxiom’s case was a Pandora’s box. Had it been blocked totally with a tolerance-zero approach on machine translations, you likely won’t see as much machine translation today. Yes, machine-translating would pierce through eventually, but certainly not as fast. The simple act of allowing it has opened up the idea that machine translation was viable. And finally months later, with the popularity of webnovels, machine translating has become an “acceptable evil” that is necessary to make Japanese content accessible to readers.
Ever since machine translation became an issue for BT to deal with, my opinion on them have swung back and forth between “to be tolerated” and “to be eradicated”. In my case, I more or less reached a conclusion on that topic.
The major problem is obviously the accuracy of the English content. Some groups actually send translation checkers, which make them a notch above the rest of pure machine translators. In theory, there should be no problem. However, a good amount of work is necessary to change words, sentences or even huge blocks of text because of a misinterpretation error by the machine translator. For a bad machine translation, it is simply faster for the TL checker to translate the damn thing by himself or herself. Re-structuring a sentence is, I think, the most painful thing to do. If something needs to be corrected, you must break the chain, replace the good part and find a way to regroup the pieces together. It is demanding, and probably even harder if you feel that connotation was lost in the translation. In any case, I would find a translation like this acceptable, though TL checking is not popular and it sounds somewhat inefficient.
On the other hand, machine translations with no form of checking whatsoever before making an interpretation can be the source of major mistakes. One example that I have in head is when a translator screwed up a line, and ended up interpreting what should be contents of a dream into something in reality. Things like these messes up a ton lot with the translation, and no one would notice this problem. The machine translator in charge would also modify future paragraphs so the meanings match. For instance, if I were to mistakenly assume A-chan, an evil girl who appears in MC’s dream as his girlfriend, as MC’s actual girlfriend, the way conversations between the two are translated would be greatly affected. An issue like this could go on and on if no one with access to the source were to read the English version.
As for those who do machine translations and who don’t know how to create a coherent sentence structure…Sorry, but they should not even be translating. Luckily, I don’t remember many examples of translations like these, but as the standards get lower and lower, stuff like that might happen more.
I believe that if the light novel community could draw a line separating what is fine and what is not, things would go better. And maybe it’s time to discard a certain amount of (bad) content so the norm for English LNs doesn’t get worse.