Interview with Jost Zetzsche | The Big Wave

Interview with Jost Zetzsche

The Big Wave’s Luigi Muzii interviews one of the most popular and prominent figures in the translation industry, translation technology pundit Jost Zetsche.

A native of Germany, Jost joined the translation industry in 1997.

He is the co-founder of International Writers’ Group, the editor of The Tool Box, a monthly newsletter for translators published since January 2004, and co-author of Found in Translation: How Language Shapes Our Lives and Transforms the World with Nataly Kelly.

Jost can be also reached on Twitter.

LM: What is the major change you noticed in the translation industry and in translation technology since the first edition of the Translator’s Tool Box newsletter some thirteen years ago, back in the 20th century?

JZ: It’s nearly impossible to limit myself to just one, but here are some of the changes that I just talked about in my last newsletter:

  • The size of our markets has dramatically changed — both in a good sense: we can and do work for clients around the world — and in a bad way — competition from everywhere has also risen dramatically.
  • The size of some clients, the complexity of some projects, and the natural market forces of a maturing industry have led to the growth and formation of some very large language service providers who all use their own technology.
  • Technology has become a very integral part of our everyday work lives, making it more and more unlikely to “opt out” (though I have a strong feeling that those who have chosen not to use any technology beyond a word processor and email and are making a good living so far will still be able to do this for a long time to come).
  • The ever-growing presence of machine translation has led to a different set of expectations for many customers.
  • Workflow models like crowdsourcing have come out into the open (they’ve been around for a long time!), accompanied by the participation of untrained translators in certain translation projects.
  • Portals like ProZ have furthered a segmentation of our industry into lower-wage and higher-wage workers.

While some of this might not look too bright at first sight, I think that our industry is doing quite well (keep reading).

LM: In your opinion, what is the major change in the attitude, approach, and profile of the typical translator in the last thirteen years or so?

JZ: I think there is a more visible separation between translators who are refusing to work for language service providers (LSP) and those who work mostly or exclusively for LSPs. There are advantages and disadvantages with both approaches: Those who directly contract with clients typically ask for significantly higher rates (but also do a lot more legwork to get the jobs). Those who primarily work for LSPs are not being paid as much but can almost exclusively focus on translating (plus they work on a whole range of projects: from the birth-certificate-like mini project to the next big thing from Google or GM, companies that don’t contract with individual translators). This separation is not a problem per se, but I sometimes worry about the deteriorating attitude of each side toward the “other” — a trend that I would like to see reverse.

The typical professional translator is also a lot more tech-savvy than she was 13 years ago. And while there are always exceptions to that rule, I’m glad to see that some of our efforts (along with the general demands of the market) have been successful. It’s no longer “cool” to be technically challenged. That in itself is an achievement.

It troubles me, though, that translators tend to look at that list of changes I mentioned and end up with a very negative outlook on the future of the profession. Especially the advances in machine translation have taken the wind out of the sails of many translators. Too many of us have read and begun to believe the many articles in popular media that speculate on the immediate demise of the “human translator.” Here’s the truth: Translators are here to stay. They have been around as long as humans have communicated, and that’s how much longer they will survive. The way we do our job will change — it has already changed in many ways — but the core competencies that make us translators will be in demand in the immediate, mid-term, and long-term future.

The book that Nataly Kelly and I co-wrote, Found in Translation: How Language Shapes Our Lives and Transforms the World, which portrays the relevance of translation and interpretation in every area of everyone’s life, really was geared toward the non-translator — our potential clients — to make them excited about translation. But I’m starting to think it might also be good (and sweet!) medicine for those translators who are depressed about our industry.

LM: The Translator’s Tool Box has still the old-fashion vest of a newsletter. Have you ever thought of reshaping it for a blog, a Facebook page, or a kind of social app? Why?

JZ: Right when RSS started to appear many years ago, I announced that I would be switching to a format that would allow my newsletters to be downloaded as RSS feeds. This obviously never happened, and I’m glad that it didn’t. You’re right that most people today write in the form of blogs or other social media, but I know that many of my subscribers can’t wait to receive the new edition of the newsletter in their inboxes every four weeks. Maybe it’s because newsletters have become the exception rather than the norm. At any rate, it’s a model that works for my subscribers, my advertisers, and me. I have found an extension of sorts in my Twitter account where I often continue discussions that I started in the newsletter. And even though only about a quarter of my subscribers follow me on Twitter, this has still been a good additional outlet. I also offer access to the archives of the newsletters from the last few years or subscribers to the Premium edition.

LM: In an article for the Northwest Translators and Interpreter’s Society newsletter in 2004, you wrote that what distinguishes us from machine translation systems is that machine translation systems might know language rules, lexicons, and technology, but since they don’t truly understand language in context, they often fail us. In the light of the development of translation technology, and especially machine translation, in your opinion, what role does language knowledge play?

JZ: For the translator, language knowledge is naturally the most important asset he or she has. And I would even go so far as saying that a sub-par translator has less of a role today than 10 or 15 years ago. While chances are that a translation of a translator with sub-par language knowledge will still be better than machine translation output, it makes for a difficult argument to pay a lot more for the services of that person than for something by Systran or Google Translate.

It is our excellence as language specialists that makes the difference today and will continue to differentiate us in the future.

LM: Some observe that technology for the professional translator has not developed as fast as it could, and that very few professional translators make full and savvy use of TEnTs, while an always increasing number of translators is using online machine translation systems for a first “dirty” translation pass if no translation memory hit is found. Do you think these systems are to blame for commodification of translation?

JZ: No, I think the commodification of translation is not happening because of translation technology but because of processes. What I’m talking about are processes where translation projects are offered to an almost anonymous group of translators and the one who responds first and/or offers the lowest price gets the job. Such processes can work if there are enough safety measures involved (companies like Facebook have shown that even with a largely voluntary group of translators you can produce good products), but often they don’t.

You’re right that the practice of gist translation by integrated MT systems that need to be corrected is becoming more common. But though I have no hard data to prove it, my feeling is that it’s not nearly as widely used as the mushrooming of those features in translation environment tools might suggest. Most professional translators know that this is more likely than not a waste of time. There are some isolated projects where I have found it useful (for example, short software strings), but aside from that it often takes a lot more time to correct than to translate from scratch, plus the quality of the results is often better when doing it manually. (And this is where it’s really different than working with fuzzy TM matches.)

The one area where I see potential for the use of machine translation in a productive manner is in the closer combination of translation memory, termbases, and MT. Some tools already offer to repair fuzzy TM matches with MT, and there are a lot more possibilities out there. I have addressed some of those in this short article.

LM: According to Sharon O’Brien of Dublin City University, we are – or should be – moving toward a dynamic quality evaluation model for translation. In your opinion, does it still make sense talking about quality in translation without reshaping the basic methodology, with a view to translation technology?

JZ: Just recently I had a talk with a person who was closely involved with EN15038, the European standard that Sharon also mentions. The talk quickly became a little more heated than I ever expected or wanted it to become. I think translation quality will remain a contentious topic of discussion, maybe more so than as a matter of implementation.

Though this cannot be applied to many other text types, the approach to quality in the machine-translated Microsoft knowledgebase (see an example right here) is one of my favorite examples of a clever and telling implementation to measure quality. At the end of the MS article, you’ll find questions about the quality of the machine-translated article. These are the same questions that you find at the end of humanly translated KB articles (see this one here). A translator who compares the translation quality of the two articles will immediately have a visceral response: one has “good quality” and the other seems to scream out its “poor, machine-translated quality.” But the users? They find them both (virtually) equally helpful, according to the response data Microsoft collects.

So, yes, the perception of quality needs to be a lot more dynamic. There is certainly room for quality metrics and standards, but we need to accept that these don’t apply to everything. And some of the translation buyers have long figured that out.

Oh, and I would be remiss if I failed to mention that I’ve always been impressed with the name of the journal that Sharon’s article appeared in: JoSTrans — it’s not often that I can see my name in such prominent display.

LM: In 1950, Alan Turing, the father of artificial intelligence, wrote a single essay, asking the question, “Can machines think?”. He had actually no very convincing arguments to support his views, and computers have not evolved quite as Turing expected them to. And yet we now often talk of “training” a machine translation system. Do you think translators education should be rethought to accommodate the widest possible use of machine translation systems?

JZ: If you are asking whether translator education should involve some of the upcoming uses of machine translation in our translation processes that I mentioned above, the answer is yes, of course. The largest hurdle that I see in that is that teachers probably would have to overcome their own bias — and we know that this can take a long time considering how long it took for other kinds of translation technology to be widely taught in universities and colleges.

If you are asking whether university translation programs need to adapt to accommodate machine translation developers’ potential need for linguists, the answer is not quite as clear-cut.

There is no doubt that linguists are needed to help in the advancement and fine-tuning (maybe a better word than “training”) of machine translation systems. This is something that even the most hard-core statistical machine translation developers have started to realize. I’m not so sure, however, whether this is necessarily compatible with the profile of a translator in the classic sense. However, I do think that someone who has a dual degree in translation and computer science is going to be very attractive to MT developers.