Production vs. Creativity

And now, to Translator X. With many of our readers inquiring as to the true identity of Translator X (who really does exist, we might add), Pierre Cadieux interviews and loosely translates Translator X in place of our regular editorial.

(Although as a translator and programmer, Translator X like the duckbill platypus, is a mystery of evolution, Pierre would like to state for the record that he may not actually be a vampire.)

Pierre Cadieux: Are you, or have you ever been, a vampire?

Translator X: No.

Pierre Cadieux: Are you, or have you ever been, a translator?

Translator X: Yes.

Pierre Cadieux: Are you, or have you ever been, a programmer?

Translator X: Yes.

Pierre Cadieux: … (before I can start, he does, and I thought I was running this interview)

Translator X: So you believe translation is a production activity…

Pierre Cadieux: Well, yes (I cannot lie, we have discussed this before).

Translator X: (satisfied smile) I agree, translation is certainly not a counter-productive activity.

Pierre Cadieux: Seriously, I mean you have to produce 1500 words or more per day.

Translator X: Yes, we have deliverables and deadlines, do you mean programmers don’t?

Pierre Cadieux: We have deadlines, but they are not daily, more like every few months.

Translator X: And deadlines are often missed, so you mean we are simply more productive.

Pierre Cadieux: That’s not what I mean.

Translator X: Let's get to the bottom of this; let's tackle the gross cliché that translation is a repetitive activity with little creativity. Do you believe translation is creative? Do you believe translation is easier than authoring?

(Now he has me in a corner. I honestly believe that when I write something, finding the ideas and ordering them is the hardest part, finding the words less so. Although I must admit that when I translated my CV, from English back to French, I found it excruciatingly difficult and painful to achieve.)

Pierre Cadieux: If the ideas in the source text are clear, I would say that the authoring is the creative part and that translation is indeed easier (assuming that we are talking about simple translation, not a marketing slogan, a song, poetry, literature, etc.). However, if the source text is unclear (and, as usual, the authors are not available to clarify it), I can see it being quite challenging to come up with a decent translation.

Translator X: So what do you think is the average percentage of sentences that are ambiguous from the point of view of the translator?

Pierre Cadieux: (knowing it happens frequently) I would say somewhere between 10% to 30%.

Translator X: In fact, it happens at least 50% of the time.

Pierre Cadieux: Are you saying that translation is harder than authoring?

Translator X: Absolutely.

At this point, I am still a bit skeptical, but by pure chance, the next day I meet translators Y and Z.

Pierre Cadieux: A translator I will not name was telling me recently that translation is harder than authoring.

Translator Y, Translator Z (in unison): Absolutely.

Translator Y: In fact, I should know. I was a translator, and I became a technical writer because I was tired of having my hands tied behind my back. And the icing on the cake (in Quebec French, he actually said, "la cerise sur le sundae" - the cherry on the sundae) is that it pays better.

Pierre Cadieux: Why do you say you had your hands tied?

Translator Y: Because you are always constrained by the source text and, since it is often ambiguous, you constantly have to make choices of interpretation that will invariably disagree with someone else's interpretation. The result is that you start translating defensively.

Pierre Cadieux: So you're basically saying that translation is harder than authoring and that it pays less. Why is that?

Translator X: Maybe because you only notice bad translations. A successful translation is invisible…

Source Text Ambiguity

Pierre Cadieux: A lot of the difficulties in translation seem related to the ambiguity of the source text. Why do translators perceive 50% ambiguity on average? Is it because translators must go deeper into the meaning of the text than the average reader, or because the text is of poor quality to start with?

Translator X: A bit of both. When a text is ambiguous, the average reader will simply interpret it relative to his frame of reference; he will simply read what he wants to read. He may not even notice that there is more than one possible meaning. The translator, on the other hand, must try and figure out the true intent of the author and this is often a problem, in particular in the technology domain.

A great deal of technical documentation is written in English that is by nature a highly ambiguous language. Compounding the problem, many of these technical documents are produced by writers who are not trained technical writers and, in fact, are not even native English speakers. These documents are then directly passed on to the translators who have to make due.

There is a double irony here: native speakers and review are both required for translation, while neither seem required for authoring.

Pierre Cadieux: Are you saying that there is no quality assurance before translation? Why would that be?

Translator X: For the same reason that the people authoring the documents are not trained English writers. It is simply because in the high technology area, many people believe that the quality of language is not important: "We build choose-your-device; we don't do English."

Pierre Cadieux: I guess it’s a question of focus and process maturity. I once worked in a young medical software company. It was hard to convince them to test their software, but it was virtually impossible to get them to check their documentation (even though at that time, a major U.S. airline had just been embarrassed in the news because one of its manuals for airline pilots had derogatory comments about Spanish-speaking pilots). Hardware companies sometimes see software as a necessary evil, software companies may see documentation as a necessary evil, and many may see testing and translation as grudgingly necessary and even more evil!

Do you believe the lack of quality assurance in documentation can also be attributed to reticence on the part of the authors?

Translator X: I think so, particularly since in many cases, the results of the revision would be "stop writing."

Pierre Cadieux: One interesting approach to solve this problem is called "First Translation Edit." Rather than review the English originals, they are first translated into German and then thoroughly reviewed. The results of this review are then used to improve the originals.

Translator X: German is a good choice since it is one of the least ambiguous languages, but why not just review the English original in the first place?

Pierre Cadieux: Probably to avoid some political or territorial issues. Nevertheless, one could argue that if the objective is to translate into many languages, the best review can be achieved by going through the translation process once, precisely because the translator is best suited to identify all ambiguous points, as you already pointed out.

Translation Ambiguity

Pierre Cadieux: It seems ambiguity is an integral part of translation.

Translator X: Absolutely. A translated text is always more ambiguous than the source. Whether consciously or unconsciously, the translator always introduces ambiguity. Because source text is ambiguous, because the authors are usually not available.

Pierre Cadieux: Could you share a couple of examples?

Translator X: Sure, here are two real cases from my personal experience:

(1) "The tank must have a liter capacity."

Does this mean the tank is graduated with liter measurements, that its capacity is one liter, or that a number is simply missing?

(2) "When there is a fire, the lamp will go."

One can guess that the indicator lamp will change states, but will it go on or off? Many people would be tempted to guess that "go" means "go on." However, in this actual case, “go” referred to an AC-connected smoke detector: the indicator lamp is always on, going off when smoke is detected.

Source text is full of gray areas, and many of them put the translator on the spot. Sometimes, as in the above case, the translator's decision may involve public security, making it that much more crucial. The translator can either call the author or introduce ambiguity. Authors are often unknown, often unreachable and often defensive when reached. Sometimes their answer is: "Don’t try to understand it, just translate it," which is the most infuriating cliché of them all - as if one could translate without understanding!

Consequences of Translation Ambiguity

Chunking Creates Problems

Translator X: Another frequent problem from English to French is gender. Consider the following:

"The dentist works on her patient's tooth."

Here we know that the dentist is a woman, but what about the patient? There is no neutral gender in French, so the translator must assign a gender to the patient. And it should be consistent with whatever gender the patient actually is, which may only be revealed a few pages before or after.

Pierre Cadieux: Which highlights the problem with the modern trend towards "chunking" in content management systems. If the above text were part of a chunk that did not identify the gender of the patient, the translator would have no way of knowing.

Translator X: Absolutely. In fact, you can now see several corporate Web sites using this "chunking" approach because, in the French version of the sites, the gender and verb tenses are regularly inconsistent or simply wrong. Language has a continuity that chunking breaks.

Translation Is Not Reversible

Translator X: Another source of ambiguity is that different languages have different levels of nuance for different things. For example, French has two words: "efficace" which means "efficient" in English and the French "efficient" (a “false friend”) which means “cost-efficient” in English. Conversely, a single French word like "aimer" can mean "love," "like" or "fond of" in English. The translator must decide which one it is!

And, as any translator knows, translation is not reversible. The other day, a customer of mine wanted to invert a Translation Memory, i.e. to use it in the reverse direction, from target to source: this simply will not work.

Pierre Cadieux: That is one of the main differences between Open Lexicon Interchange Format (OLIF, which is unidirectional) and Term Base eXchange (TBX, which requires bi-directionality).

Translator X: My experience with terminology has been mostly unidirectional. I get very concerned when non-translators assume that terminology can easily be bi-directional and completely horrified when this assumption is generalized to multi-directionality.

English as a Pivot Language

Translator X: One thing I see too much of is English as a pivot language, i.e. authoring in a variety of languages, then translating to English first, then a second translation from English to other languages. This is very dangerous and often results in problems. First of all, since each translation step introduces ambiguity, going through a pivot language degrades the original (like cloning a clone). Secondly, English is a very ambiguous language because it is highly polysemous (many words have multiple meanings); it is thus a poor choice as pivot.

Pierre Cadieux: There might be ways to make it work; say, using controlled English as a pivot and enforcing a very rigorous review or "First Translation Edit."

Translator X: I would prefer using another less ambiguous language (e.g. German) as pivot, with a solid review, of course.

(And there we left it… a bit ambiguous. To be continued, somewhere in Transylvania…)