Posted by: Steve Vitek | August 14, 2011

Is Further Hamsterization of Freelance Translators Inevitable?

Hamsterization is a term that was first used by Dean Starkman in a Columbia Journalism Review article titled “The Hamster Wheel“. The hamsterization of journalism has created a number of new models for “news production”, such as The Huffington Post blog, which publishes contributions of authors who mostly work for free or for peanuts. I understand that authors contributing to this kind of hamster factory are typically paid either nothing or about 30 dollars per story. However, hamster farms can be eventually very profitable to the owner of the hamster factory. This was proven once again by Ariana Huffington, the great American liberal who is so passionate in her bestselling books about the plight of the poor in America, when she sold her hamster factory for 315 million dollars, only to be sued by angry bloggers who for some reason feel that “[they] have been turned into modern day slaves on Ariana Huffington’s plantation”. Since the average hard cover book sells for 29.95 dollars these days, being a bestselling liberal who is very concerned about the plight of hamsters who are forced to work for nothing or for peanuts is an occupation that can pay really well.

Translation memory tools (CATs) such as Trados, which determine “matches” and “fuzzy matches” in new translations based on existing translations, have been a major driving force behind the hamsterization of translation work and of the function of a freelance translator. One of the unstated but obvious purposes of these CATs is to lower the cost of translation by defining these “matches” or “fuzzy matches” with software so that the translator could then be then paid nothing or peanuts. Translation agencies now routinely incorporate “a sliding payment scale” defining how much translators will be paid for matches determined by software into agreements that translators who want to work for them must sign first. The same agencies, on the other hand, are under no obligation to pass the savings which are based on income lost by the translator on to their customers who may or may not be aware of the arrangement.

A recent trend in hamsterization of translation is the concept of using translators as post-editors of machine translations. It is an idea that on surface makes a lot of sense. Wouldn’t it be cheaper to have huge chunks of text first translated by software and then have human translators (in the role of obedient hamsters) edit the text at a low hourly rate so that it would actually make sense? The owners of these hamster factories should be able to make a lot of money in this manner, although probably not as much as great American liberals like Ariana Huffington and many others.

Many people believe that this is what will eventually happen to human hamsters who still call themselves translators. After all, journalism has all but disappeared from our media under the onslaught of hamsterization on news in print and in broadcasting. News segments now have to be short and targeted only to areas that attract advertisers, while topics that could displease advertisers must be avoided at all cost. Given how the Internet has decimated the once profitable business of news making, it really does not matter whether the news is important, relevant or useful, as long it makes money.

What does it matter whether a translation is the result of a creative activity of highly educated individuals who are passionate about their work if the result of texts that have been edited by CATs, or translated by software and then edited by a human being, look almost just like a real translation?

Depending on the language, the subject and the style of the text, some machine translations of some texts are probably suitable for post-editing by human hamsters, at least to the extent that the product of this hamster wheel could probably look almost like a real translation.

Fortunately for me, hamsterization has not really have had much influence on my field of translation, namely the translation of patents. This is due to a number of reasons (I was not really prescient almost 25 years ago when I decided to specialize in this particular field. Well, maybe I was just a little bit prescient).

One reason for this is that the text of published patents is usually available only in the form of PDF files which cannot be easily processed with memory translation tools, at least not without a laborious OCR conversion. Another reason is that a lot of money is invested in patents, since a lot of money can be made or lost with a patent. Therefore, the translations must be accurate, which is why the rates paid for translation of patents tend to be somewhat higher.

In fact, most patents can now be easily translated from and into a number of  languages for free using the machine translation function on a number of websites, such as the World Intellectual Property Organization (WIPO) website. Will my clients eventually rely on free translations of patents in foreign languages, which could be eventually massaged into a shape that looks more like a real translation by what is left of “freelance translators” or by “multilingual paralegals”?

Personally, I doubt it. Especially with languages like Japanese, the results of machine translation are still very inadequate. And they are likely to stay inadequate for a very long time, possibly forever. For instance, when you translate a long Japanese claim which may contain five, six, or seven hundred words in a single sentence, you have to determine first based on Japanese grammar what is the “wadai” (話題 = topic) of the monster sentence, which is not the same thing as the subject, although it can be the subject in some cases.

There is no way to translate a claim like that correctly unless you understand what the text really means. For some reason, designers of machine translation packages don’t even seem to understand that the part that you have to start with to translate a long claim is almost always in the last few characters of the monster sentence because based on Japanese grammar, the qualifying part is always in front of what is being qualified regardless of the length of the sentence.

However, all machine translations of long Japanese claims that I have seen, and I have seen many, always start from the beginning of the Japanese text (because the “topic” is often in the beginning of the sentence in Japanese), which makes it impossible for a human hamster to edit the machine translation product. Instead, it must be rewritten by the same human hamster, formerly thought of as a translator, which will of course take much longer.

Although further hamsterization of freelance translators is probably inevitable in some languages and some fields, I think that freelance patent translators are likely to enjoy a relative job security for a long time, especially in “exotic” languages which are very much in demand now, such as Chinese, Japanese and Korean.

About these ads

Responses

  1. I couldn’t agree more. I often tell people that Japanese-English patent translation (my field too) will be one of the last to succumb to MT, largely for the reasons you spelled out in your post, but it’s reassuring sometimes to read that someone else feels the same way.

  2. Hamsters of the world unite!

    I wonder, do you use translation memory tools like Trados?

  3. I use a CAT tool (AppleTrans) for some of my work, but I’ve never been tempted to get Trados.
    Also, I’ve never offered, or been requested to give, discounts for repetitions or fuzzy matches.

  4. There is only one reason claims for U.S. patents are one “sentence” long: tradition. European countries also use one sentence but there is no rule dictating that. Again, tradition.

    But you can look up American lawyers who want to end the one sentence rule since it is cumbersome and shorter sentences still give the same discription.

    Once the one sentence rule is gone, then all parts of patents will be translated with MT and some human editing.

  5. When I first joined the ATA and NCTA, I had the uninformed idea that they were organizations with members primarily interested in the intracies.of language per se and translating. Then clique-ishness and Trados came on the scene. If machine translating cannot differentiate between President Wade and just plain “wade,” or choose between “Bewitched, Bothered and Bewildered,” then we have proof sufficient, so to say.

  6. Do you remember George Kirby? He was NCTA president at one point and he was also a member of the ATA board. He died about two years ago.

    He once told me that the way he saw it, the main purpose of the ATA was to organize the ATA conference every year.

    He used to refer to a certain ATA president who shall remain unnamed as “Estrogen Alert”.

  7. But what MT is capable of today is not what it will be capable of in a few years.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

Follow

Get every new post delivered to your Inbox.

Join 1,136 other followers

%d bloggers like this: