At the BDÜ conference I went to just a few presentations -- like most of you, I tend to get more benefit from talking to colleagues than from the actual sessions. Still, I tried to fit as many sessions on translation technology into my schedule as I could. With some I was frankly a little disappointed, especially the CAT-tools-are-really-cool- or -are-really-bad presentations. However, some were very stimulating, in particular a podium discussion titled "Who killed creativity? Do we really need TM?" This discussion was coordinated by Terry Oliver (who has been mentioned in this newsletter a number of times), and he was joined on the panel by some highly distinguished and experienced translators. Here is the write-up of this session from the proceedings:

To deny the blessings of technology would be absurd in an age like ours which benefits so much from the accuracy of laser surgery, the convenience of microwaved foods and the speed of broadband access. But in all these cases we still need the skill of the surgeon to guide the laser scalpel, the experience of the cook to set the heat and timer on the oven, and the discernment of the individual to sift, sort and select the information presented.

Voice recognition and computer-assisted translation tools can no doubt offer the same advantages of accuracy, speed and consistency. However, while it is irrefutable that translation memory (TM) software can improve terminological consistency and reduce confusion by standardizing forms of expression, the downside is often a less fluent, a stilted, or even a syntactically flawed translation. For certain types of text this signals a significant deterioration in quality of the end result.

Translation buyers and translators themselves need to be aware that, in many instances, the quicker and less considered the initial translation, the more extensive and time-consuming the process of self-editing becomes. TM tools are merely a means to an end and not an end in themselves. As such, they do not necessarily lead to overall cost savings on the scale that is frequently anticipated.

Like many of the other technological aids that surround us today, TM tools relieve us of much of the burden of our work, but they can never relieve us of our professional responsibility for producing an end product that is in every way as fit for purpose as it is possible for it to be.

The debate

Why is it that some people, once they start using TM, stop creating translations that "sing"? Does the TM approach improve process quantity at the expense of translation quality? Does the man-machine interface actually affect the thought processes of those using it? Is there anything we can do to prevent translation quality suffering? Are there certain kinds of texts that TM tools should not be used for on principle? Are people forgetting that TM tools are merely a means to an end, not an end in themselves? These and related questions will be discussed by a group of experienced translators who work in widely differing fields and who range from the highly skeptical to the convinced user of TM tools.

Most of the members of the panel spoke quite strongly against the use of translation memory technology (they actually spoke about "translation memory tools" -- but more on that later) in their particular fields of translation. Their arguments were so passionate that several folks approached me afterward to ask why I hadn't stood up to say something in defense of translation memory. The answer is that there wasn't much that could be said.

We so often make the mistake -- and I am as guilty of it as anyone -- of assuming that "translation" needs to be approached in a certain way. Yes, we know that there is literary translation, and we (I) by implication exclude that from our deliberations on how to translate. However, we (I) tend to lump all functional translations together in one big, well, lump. The fact is that there are so many fine and shady lines between the areas of "functional" and "literary" translations that this distinction may very well not make sense. For instance, panel member Chris Durban said that her financial translations need to be not only informative but also interesting and fun to read or the reader loses interest and puts the unread report aside. And there are many other examples where the repeatability and consistency that translation memory technology promises (and delivers) is not only undesirable but actually needs to be avoided.

But then, of course, there are many, many areas of translation where these are highly desirable traits. (There are also areas where a rough machine translation is all that is needed if it communicates the core concepts -- think of Microsoft's machine-translated knowledge base and the usability approach to translation.) The fact is that the majority of translators (and this will be especially true for subscribers of this newsletter) benefit (or could benefit) from certain aspects of translation memory technology, provided that they learn how to use it adequately. Yet it's also a fact that there are many translators who feel that translation memory technology clips their wings (to stay with the "singing" metaphor from the introduction to the panel).

Here is the only thing I would like to add to this discussion (and what I should perhaps have added during the discussion in Berlin): We are not doing ourselves a favor by calling tools like Trados, Transit, Swordfish, Fortis and the many others "translation memory tools." Granted, translation memory is a feature of these tools, but so is terminology management, quality assurance, file management/conversion/protection/accessibility, access to any former translation, and on and on. Very few tools are "translation memory tools" -- in fact, I can think of only a couple: Snowball and AnyMem. Most of the others are tools that provide a wide range of features to aid the translation process, creating a complete translation environment. I would also suggest that those who -- possibly rightly -- assume that translation memory in its core functionality (re-using perfect and fuzzy matches) hinders rather than helps them, would benefit from some of the other features of a translation environment tool.

Interestingly, there was a good discussion on the ATA's Language Technology Division's email list about the merits or drawbacks of the use of the term TEnT (as in "translation environment tool"). But I gotta say that, especially after once again seeing the misconceptions created by terms like "translation memory tool" at the Berlin conference, I think "TEnT" is rather helpful.

Oh, and speaking of conferences, it would be fun to see some of you in Warsaw in a couple of weeks!