If every word in a language had a single, direct equivalent in another language, translation would be a relatively easy process: a dictionary and a knowledge of grammar would be enough.
In fact, if each word simply mapped directly to its equivalent in another language there would be only one correct translation of a text; translation would be an exact science and human translators would be out of a job by now, forced into alternative employment by computers, which are excellent tools for such tasks.
In reality, most words can be translated in a number of different ways depending on the context. And what is true for individual words is doubly true for whole sentences. The translator must grasp the context of the source text, its style and writing conventions, understand the underlying meaning of idioms and metaphors, and posses a cultural sensitivity for meaning and usage. Having done all that, the translator must then construct a new text that accurately reflects the source in both meaning and style.
Since writing in a foreign language is generally even more difficult than reading and comprehending it, it makes sense for translators to translate into the language they know best. In other words, an English translation of a Japanese source text should generally be done by a native English speaker. But even this is not enough. Since the meaning of a text is ultimately whatever the reader takes it to mean, the translated text should be written by a skilled writer to convey the meaning as clearly and as eloquently as possible.
By now it should be obvious why translation is such a difficult job: whether it’s more an art or a craft is debatable, but what it certainly is not is an exact science. And this explains why computers have not yet overtaken humans in this field—despite years of trying (see Translation: Does it compute? below for more on this).
Faithfully Yours: Good vs. Bad Translations
There are two criteria by which we can judge a translation: fidelity and transparency. Fidelity is how closely in meaning the translation reflects the source, while transparency is how well the translation reads to other native speakers. Unfortunately—especially for languages as unrelated as English and Japanese—these two criteria often pull in opposite directions. A translation that is faithful in meaning can often appear somewhat unnatural; it ‘reads like a translation’. However, if the text is rewritten to improve its transparency, the strict fidelity will often be compromised. This is not to say that fidelity and transparency are mutually exclusive—the translator naturally aims for both—but in the end it’s something of a balancing act.
Of course, there are arguments for and against making a translation more or less transparent—some of them quite philosophical: is it ethical that the translator ‘move’ the author closer to the reader (less faithful, more transparent), or should the reader have to make the effort to understand the ‘foreignness’ of the author? The answer depends to some extent on what’s being translated: a novelist might feel insulted if the translator takes too many liberties with his prose—putting too much of themselves into the translated text, as it were; for most commercial writing, however, it hardly makes sense to alienate customers by expecting them to spend time exploring the inherent ‘foreignness’ of the text. They will sooner explore domestic alternatives…
Machine Translation: Does it Compute?
On the face of it, computers might seem the ideal tool to translate electronic texts from one language to another. Computers, after all, excel at the kinds of things that translation seems to require. They will faithfully remember every definition of every word in any dictionary, and if computers can beat all but the finest chess players in the world, then how difficult could it be for them to translate a relatively simple passage of text that any native speaker could understand?
The answer turns out to be “Very”. Translation was an early goal of computer programmers back in the 1950s, yet today the results of computer translation (or machine translation (MT), as it’s usually called) can still cause amusement (or bemusement). Yet translation engines such as Google Translate have got markedly better in the last couple of years, and while they may not be putting actual human translators out of a job just yet, it might be just a matter of time.