Translation memory manager vs. machine translation
To those of us working in the translation industry the two technologies couldn’t be more different in their purpose and usage. But we often find that the two are easily conflated in the minds of people outside our field, including current and prospective clients.
Here’s the way we explain the distinction on our company site:
The use of translation memory management software (TMM) is not to be confused with machine translation (MT) such as automated online translation applications that translate at the level of vocabulary and grammar – and produce results that typically read like amusing gibberish. TMM assists human translators by allowing them to break a document’s text into segments (sentences, headings or clauses) and pair each source language segment with a translated version. The application manages these human-translated segments so that the same native translator (or another native translator sharing the same database) can call up and recycle the segments as appropriate when working on updates of the same document or similar documents. The result is as natural and native as any other human translation. The use of computer-assisted translation tools saves translation professionals time, so it saves translation clients money. And there are other benefits. You get better standardization and consistency within and between documents.
Yes, sir, it’s still human translation
I think that clients easily grasp the fact that computer-assisted translation (CAT) tools save time and reduce costs. But they don’t readily understand that we’re still talking about human translation, that TMM software is just a tool in a translation team’s toolbox, or that it helps the team edit the client’s information assets with increased care, consistency and clarity.
They understand “translation” but misunderstand “translation memory”
Even if laypeople aren’t completely hip to professional translation software, more and more businesses (large, medium and small) these days understand the need for quality translation. Haven’t we all noticed the widespread localization (a.k.a. localisation) of just about every business website lately? The web is soooo worldwide that there are no “foreign languages” anymore, just local languages. More and more people outside the translation industry are coming around to the understanding that their organizations are situated in a multicultural/multilingual world that requires “global communication,” as clichéd as that already sounds. Members of our profession are valued as intercultural interpreters.
Humans have human sense, and software applications don’t
I guess we need first to explain to clients what human translators do that automated translation software can’t.
The other day I was chatting with a client who shared with me a rolling-on-the-floor laughable document that some speech-to-text transcription software had produced. Instances like this remind us, to our great relief, that human labor is still indispensable even in 2008. In translation work we can’t produce quality documents without being humans endowed with great writing skills and a completely native mastery of the target language and its overarching culture.
A professional translator makes sense of the meaning of each part of a document by understanding the whole document. Likewise, she understands the whole context by following the intended meanings of the parts. Like a back-and-forth interpretive dance. It’s likely that she’ll have to “read between the lines” and rely on a broader understanding of, say, marketing objectives that aren’t even stated in the document.
A software application alone – no matter how fancy the algorithms you program into it – can’t perform such feats. And it seems doubtful whether it’ll be possible in our lifetimes, if ever. Only a human can follow the sequence of thoughts and intended meanings throughout an entire document.
Professional copywriters have the good sense to translate entire phrases as equivalents to phrases in the source copy even if the equivalence is in the impact of the words rather than their literal meaning. And they have the good sense to leave some words untranslated for effect. No software is savvy enough to make skillful decisions like that.
Software and its limitations
When a client hears the phrase “translation software,” does he or she imagine a skilled copywriter working with tools? Or does he or she imagine machine translation (sans human)? An unscientific polling of project managers leads me to expect that most people imagine the latter. And no wonder. We all know the free web applications based on SYSTRAN technologies such as AltaVista Babel Fish, don’t we?
Have you heard of the way biologists use the word “translation”? (I found this in Wikipedia.) Not sure I get it, but,
In translation, messenger RNA (mRNA) is decoded to produce a specific polypeptide according to the rules specified by the genetic code.
The process seems to follow predictable laws.
Linguistic translation is completely different, isn’t it? Unlike mRNA, text requires shared subjective understandings. Linguistic translation requires a certain cultural accuracy and pure savvy that no computer algorithm can possibly approach. When Walter Mondale in 1984 asked Gary Hart “Where’s the beef?” he wasn’t trying to locate some meat. He was trying to win the Democratic nomination. How would Babel Fish translate that debate into Spanish, German, or French? When we express something in our native language from a foreign source we don’t just decode words and sentence structures and match them up according to a language pair dictionary. If we did we couldn’t expect culturally appropriate and comprehensible language to flow out automatically. Right?
Machine translation creates output that often reads like absurdist poetry rather than meaningful prose. And often very funny absurd poetry at that.
Today I entered this sentence into Google Translate:
“We knew it from the word go.”
English speaking natives will understand that the idiom “from the word go” means “from the beginning.” I had Google translate it into “Japanese,” and got:
“Wareware ni iku to iu kotoba wo shiteru kara da.”
Japanese readers will know that this is ungrammatical (just a clause with no subject) and meaningless. It reads like:
“Because (I/we/they/he/she/you/it) know the word(s) of us to go..”
If you replaced some characters you could make it grammatical, but idioms are expressions of culture. Interpreting an idiomatic phrase in a literal way, as free online translation does, results in a loss of meaning (and style, too).
Okay, okay… This is not news to any of us. We all know what MT produces. But this is what clients fear we might use on their documents.
Explaining CAT tools to laypersons
Is it any wonder that the idea of using translation tools gives people a negative feeling? Inside the translation industry we know the complete difference – like apples and orangutans – between Trados and Babel Fish. But do clients get it? Can they get it?
This is where I’m stuck. I’m in translation marketing. I’m not a linguist. I’d love to get some comments to this post. (See Comment box, below.) There must be a useful, quick and clear way of explaining this to clients. I know basically how translation memory apps work by watching Japanese translators work in our office, but I’m not sure whether my explanations to client prospects are persuasive when the question comes up. The clearest difference I see is that CAT tools organize entire documents of text that has been translated by human translators. Entire documents, not word-by-word dictionary translations. These tools simply help translators organize segments of human-translated text inside the document. And…erm… Is there any more concise way to explain it to a lay client?
Of course the smart option for huge translation agencies such as WorldLingo is to offer both machine translation and human translation. Wicked smart! Or… Does that only cause more confusion??
:: :: :: :: ::
By the way, if you own or manage a translation agency, please visit our new page for translation agencies!