Posted by: patenttranslator | November 6, 2015

What really matters is what is happening in our heads

Translators are surrounded by nonsensical and complicated propagandistic notions defined using specialized terminology that has been generated by “the translation industry” in the last two decades or so to emphasize the indispensability of this particular industry for … the very survival of humankind, I suppose.

The industry no longer refers to translation agencies as translation agencies. They renamed themselves about a decade ago and henceforth are to be known solely by the acronym LSP which nobody outside of “the translation industry” understands. As far as I can tell, there is only one possibility: the great minds of so-called translation industry created this acronym, which stands for Language Services Provider, to hide the fact that they are in fact mere brokers of language services and not the actual providers of language services, who are the translators that work for these agencies.

The interesting thing is how quickly some translators have adopted this nonsensical acronym and seemingly delight in throwing it around in online discussions without giving a second thought to its meaning. When you use the right acronyms, you belong to the right club.

A lot of terminology that is used by translation agencies, and then willingly and eagerly adopted by translators, makes no sense to Mad Patent Translator. (Could it be that the madness clouding my mind is ordinary insanity rather than divine madness? Perish the thought).

The notion and the phrase “editing of machine translations”, for example, seems more than inaccurate to me. Completely insane would be a much better way to put it.

How can you abuse language by calling retranslation “editing”? Here is a sample of a machine translation of a Japanese patent dealing with mechanical engineering techniques that I am translating this morning:“In the cylinder 1 is cylindrical, in portrait cylinder for generating ice to the inner wall. Depending also Ri the inner wall goodness in and the child to cool the peripheral portion of the out outer, upper and lower tubular outlet and face it fixed the junction marks La Nji the upper Hawa managing 2 and the lower housing.”

Even this particular machine-translated gibberish is very useful for my purposes because a quick look at the machine translation answers a number of questions a translator might have. But clearly, no amount of editing would create a real translation from these words. Human brains were not designed to “post-edit” nonsense that can be produced only by machines. How can you abuse humans by asking them to engage daily in such a futile effort as “editing” of the detritus that is left by machines when they are done with mechanical processing of human language? There ought to be a law against this kind of abuse of humans by other humans who use machines to inflict torture in this manner.

I edit human translations all the time. But when I edit translations by human translators, I simply look for typos, omissions and inconsistencies, because those are real translations produced by really good translators, not artificial, nonsensical constructs generated by a computer with an algorithm. The mental process taking place in my head during the editing process is very different from what goes on in my head when I look at machine translation. As Rudyard Kipling might have put it, had he been exposed to machine translation, “Machine is machine, and human is human, and never the twain shall meet”.

Just about everything in the so-called translation industry is so transparently fake now that the entire industry resembles politics at the highest level of government.

Here is a title of a translation seminar I saw somewhere today: “Quality metrics for measurement of translation quality with crowd-sourcing”. When you throw a huge number of words at an anonymous crowd to save money by circumventing real translators and then ask the crowd to please translate the words, either for free or for next to nothing, what kind of quality can you possibly expect? To use the word “quality” in this context is really a bad joke. Even if we were to measure “the quality of translation” of experienced and well-paid translators, and even if it were possible to design “verifiable metrics” for this purpose, how can we ignore the fact that the only true measure of quality is the satisfaction of clients who pay for the translations? By definition, it’s a highly subjective criterion.

Application of universal quality metrics to quality of translations makes as much sense as applying universal quality metrics to other intellectual processes and activities, such as writing of detective novels, or even better, romance novels. It is generally easy to tell a really bad novel: the readers will get tired of it after a few pages, never finish the book and then in won’t sell.

Similarly, it is very easy to tell a really poor translation. But universally applicable, measureable and verifiable quality metrics for identification of really good translations? There is no such thing. Despite the fact that there are no “objective metrics”, pointy heads in the “translation industry” insist that they’ve found a universal quality standard and that there really are universal quality metrics that should and must be used to measure translation quality.

And by chance it so happens that the same pointy-headed experts are selling a webinar series, and it must be our lucky day because we can buy the entire course from them right now and participate in their webinar for only a few hundred dollars.

When the old leader of North Korea died, whatever the name of the father of the current North Korean leader was (I call the son the Chubby Leader), immense, overwhelming, incessant and extremely realistic grief of a half a nation divided by a demilitarized zone and a common language was on display for TV cameras for all the world to see how much the North Koreans love their leaders.

If you as a citizen of North Korea should somehow not be completely despondent for days, weeks, or perhaps months while mourning the passing of one dear leader of a new Korean dynasty, at the very least you would be sent to a reeducation camp. On the other hand, some or perhaps most of the sadness and despondence in the faces of the people exposed to the most pernicious propaganda on this planet seemed very genuine.

When people are exposed to propaganda for a long time and nobody offers an alternative to that propaganda, they will ultimately start believing what the propaganda claims, no matter how nonsensical those claims may be.

The leaders of the so-called translation industry understand that the task facing them is to control the thought process happening in our heads, and I have to say that so far, they have been doing a very good job of it.



  1. I loved “Could it be that the madness clouding my mind is ordinary insanity rather than divine madness? Perish the thought.”
    I did find the father and daughter scene somewhat troubling but I would never question your absolute right to show it. 🙂
    BTW are you still planning to visit the CR or did we already miss you some-how? I’m planning to head to NYC to see my daughter and her husband and my granddaughter before too long and would like to take the opportunity to meet-up with you – provided, of course, that you also have the time and the interest – but I’ve spaced-out exactly where you’re located but I think – and hope – that it is in the East.
    All the Best!


  2. Hi Michal:

    Thanks for your comment.

    Yes, I will be one of the speakers at the BP16 conference in Prague between April 15 (tax day in US) ~ 16, 2016, so maybe we can get together there.


    • That sounds terrific! It would delight us if we can lure you to Kutná Hora too if you have any wiggle-room and we would happily pick-up the tab for your quiet and friendly accommodation too. As I’m sure you know it is one of the most historic and spectacular Cities in the Czech Lands and very easily accessible from Prague – just over an hour by train.

      One of these days, fairly soon, I’ll also need to make it across the Great Water. My daughter, who was born in our basement-flat in Noho – a home-birth that took place so rapidly that it pre-empted the arrival of the male midwife whom we had on-call – after living with me here (I had been granted custody of her) first in Prague and then in Kutná Hora and then again in Prague with a good Czech lad called Luděk whom she later exported to NYC and subsequently married and for a number of years they lived together on the Upper East Side. They visited us here during that time and Luděk owes them a return visit which will also give me an opportunity to catch up with all their news and their shenanigans :). Eventually Tara will also be able to tell you anything and everything you might want to know about the Czech Republic.
      Anyway to get back to the main agenda – I’ll be delighted if we do have an opportunity to meet and if it doesn’t work for you to come visiting I’ll gladly come to the, relatively, Big City to shake your hand [NOT to shake you down! :)]

      All the Very Best!

      Michal Pober

      Liked by 1 person

  3. I am currently committing the sin (maybe venial) of attending the ATA conference, where I am learning a lot of useful stuff, as well as, perhaps, having my brain washed.

    I have been to a couple of sessions on machine translation, going into a bit of detail about how it works and what it can and cannot do. (When I digest and organize it, I will put a summary of what I learned from it on my web site,, perhaps in a couple of weeks.)

    When it comes to MT involving Japanese, the expert giving one session made it clear that researchers in MT consider Asian languages about the most difficult to handle. They estimate that pairs involving English and most European languages rank the highest in general quality as compared with human translation (some around 80% or so, as a rough measure of how useful they usually are in comparison), whereas Japanese, Chinese, etc., are currently at about 10%, according to him.

    I asked him why this was true, and suggested that it was probably because English is much more closely related to the European languages than to the Asian ones. He replied: yes, to some extent, but he thought it was due more to the fact that MT researchers have worked much more on European languages and are just starting to give more than minimal attention to the Asian ones. My reaction to this reply (which I did not share with him) is that he was probably being at least a bit (!) optimistic about future progress in Asian languages, because of course investigators in any field have to have a rather rosy view of the future of their work to keep plugging away. I think the present lousy situation with Japanese has a lot to do, not only with the extreme foreignness of the language from English (given certain technical factors, that might be handled eventually), but especially with certain peculiarities of how the Japanese themselves usually use their language.

    For example, Japanese people routinely omit as much stuff from their sentences–stuff that we English speakers think is needed–as they can get away with and still be understood by the listener or reader, starting with the subjects of the sentences. The human language user (and natural languages have been invented by humans, not machines, after all) can fill in the context that was not expressed, but computers can’t “read into” the actual words to figure out how to understand the sentence. In fact, computers just don’t understand the meanings of sentences or words at all. I think that such problems might be insoluble for MT, or at least won’t be solved for a very long time, so this expert’s optimism might not be very realistic. But to give a good argument for that view requires knowing a lot of technical information about MT. One can’t just throw out a blanket statement that “MT is just awful,” as a lot of translators do; one has to understand pretty well how it works, and as the guy in the Ghostbusters movie said, “It’s technical.” In any case, the present state of Japanese-English MT is certainly awful. I never find it useful in my work.

    As for the jargon a lot of translators use, jargon is found in just about any field. Outsiders usually find it ridiculous-sounding and certainly incomprehensible. But insiders learn what it means, and decide for themselves how much of it is really useful and how much is just pompous.

    Liked by 1 person

    • Well, we picked a pretty MT-proof language, Zenner. That was some smart thinking if I say so myself!

      As far as I can tell, Japanese will be pretty MT-proof for at least another century.


      • Sure, such percentages are only “pseudo-percentages,” and a single misplaced or absent “not” can ruin a translation of a sentence. But such pseudo-percentages don’t deal with individual sentences or words; it’s just a way of saying that, on the whole, MT applied to European languages-to-English jobs is a lot more useful for most purposes than when it’s applied to Japanese-to-English. There’s no question about that being the present situation.

        Whether Japanese is “MT-proof” or not depends on many factors. If a law firm is faced with 500 Japanese documents in a discovery stage of a suit and has to figure out which of them needs accurate translation, running them through even a lousy MT system, like (free) Google Translate at least gives them a rough idea that maybe only a few of them are really relevant to what the affair is about, and those they can give to a human translator. That translator can do all of those few documents and it saves the law firm a lot of time and money. Before, the translator might not have gotten any work from that client at all.

        In other words, MT even now can certainly be applied to Japanese documents to get a result that the client can use for his purposes, and it may benefit the translator too. You can’t make a blanket statement about this subject.

        I also had a thought about this business of using terms like “LSP” instead of “agency.” Usually you see the term “LSP” in a pair with “vendor,” the latter referring to what was called a “translator,” or “independent contractor,” in the good old days. Why the heck is a translator called a “vendor.” That’s sort of a technical term that agencies use, obviously because the translator is selling her services to the agency. It’s a term from the agency’s point of view. The agency then turns around and faces its clients, and since it probably furnishes a lot of services to them besides translation in the narrow sense, it calls itself a “language service provider” in order to make things clearer to the clients. It’s providing a range of services related one way or another to language.

        I don’t see that that’s an offensive term at all. What has happened, I think, is that “LSP” has gotten to be a bit of jargon that many people in the business use for convenience, whatever position in the business they have.

        I personally wouldn’t ever call myself a “vendor”; that wouldn’t make sense at all. I’m a translator; St. Jerome was a translator; the folks who turned out the Septuagint and the King James Bibles were translators. They weren’t “vendors” (though the King James guys, at least, might have been paid for their work. Maybe they could be called “contractors.”). But I don’t see any harm in using the “LSP” term, as long as the agencies themselves don’t mind it. It’s easy enough to see how the term came about, and at this point it’s a kind of technical term, like “electron” or “gene.”

        We live in an age in which everybody gets very picky about word choices, and often gets into a big huff when they think the wrong words are being used. To me it’s usually water off a duck’s back. If someone wants to call me a “cis-hetero male” or something, I don’t mind, but I wouldn’t use a term like that myself.


      • Got it, Zenner.

        I can’t make blanket statements, but you can.


  4. Computer Science 101: Garbage In Garbage Out.

    Liked by 2 people

  5. Also: “Chubby Leader” — ROFL!!!!!

    Liked by 1 person

  6. Zenner41 said, “researchers in MT… estimate that pairs involving English and most European languages rank the highest in general quality as compared with human translation (some around 80% or so, as a rough measure of how useful they usually are in comparison).”

    80% is a number plucked out of someone’s imagination; it means exactly nothing, nada, zero. By seeming to soften it by saying it’s just “a rough measure” is a devious way of making listeners think the figure had some basis in reality, even if only approximate, as if the speaker were being honest. It’s a classic propaganda trick. Mad Patent Translator is right: repeat propaganda enough times, and people start to believe it.

    Liked by 1 person

  7. “Mad Patent Translator is right: repeat propaganda enough times, and people start to believe it.”

    Actually, Catherine, you give me too much credit. In reality, I am merely channeling Goebels and Goehring who provided the basic concepts for the propaganda that is currently used by “the translation industry”.


  8. “80% is a number plucked out of someone’s imagination; it means exactly nothing, nada, zero.”

    If you translate a sentence that says in the original language “Based on our expert opinion, the real estate property is certainly not worth 500,000 dollars”, as “Based on our expert opinion, the real estate property is certainly worth 500,000 dollars”, the translation is 93.36% correct.

    This happens all the time not only with MT, but it can also happen to careless or tired humans for instance with German or Japanese because the word “nicht” or “nai” can be separated from the verb by several long sentences.

    Assigning percentages in this manner makes about as much sense as appointing Cookie Monster to Supreme Court.

    Liked by 2 people

  9. Actually, I honestly don’t remember how the percentages in this guy’s talk were generated; I didn’t put that in my notes. It might have been something like this: a big bunch of MT-ed French, German, Spanish, etc.-into-English examples were reviewed by judges in comparison with human-translated versions by a panel of people who were asked how useful they thought the various versions would be for some purpose or other, and the same was done with Japanese-to-English translations; the panel thought 80% of the European language cases were useful enough for that purpose (the human translations presumably being 100%) but only 10% of the Japanese cases were useful enough. The point is that everyone, including the people working in the MT field themselves, thought that it was a heck of a lot more useful for European languages than Japanese. That doesn’t mean that Japanese MT is not useful at all.

    A couple of other points. One way of looking at this might be that skilled human translators only make a few mistakes most of the time, unskilled humans (your Aunt Rosie who knows some Italian) make a lot more, and MT systems make a whole lot more. But you can’t accept the results that any of these translation “systems” give you without any checking; they all make mistakes.

    One good point, I thought, that the speaker made was comparing painting with artists’ brushes to painting with a paint roller. He put pictures of the artists’ brushes and the paint roller on the screen, along with pictures of a portrait done by an artist and a very tall building. His point was that if you want a portrait done you can’t use a roller, and if you want to paint a building, artists’ brushes will be terribly frustrating. (Actually, I don’t think that buildings are even painted with the kind of roller we use in painting walls in our houses; they probably use giant sprayers or something.)

    It was also clear from that session that there are now available MT systems that you can set up and train on your own computer to do the specific kinds of jobs your own work involves (provided, of course, it makes sense in the first place, for the kind of jobs you’re doing, to put in an awful lot of work, for weeks or more, to train and tune up the system to improve its results). The problem with a thing like free Google Translate, which is what most people picture in their minds when “MT” is mentioned, is that it is put out on the web to translate *absolutely everything written* in various languages. It’s no wonder that it is sort of OK when you give it some texts and terrible with others.

    The kinds of MT systems that are really useful are ones that are only employed to work on a very limited range of texts, e.g., certain kinds of documents that a particular company produces masses of, like owner’s manuals for a certain model of car that need to be generated in a couple of dozen languages, or a pint-sized system that an individual translator sets up on her own desktop or laptop. And we still have to keep in mind that it’s artists’ brushes vs. rollers or sprays; you have to pick the tool that makes sense for a certain job. Even for the owner’s manuals, a knowledgeable person will have to correct all the mistakes in the MT system’s product with her “artist’s brushes”–the activity known by the dreaded name of “post-editing.”

    But “post-editing” (for which I think a better term is “machine-human cooperation”) is another big subject altogether. Whole books can be and are being written about these subjects.


  10. Good post Steve. I am not worried about MT at all. If it is used wrongly, it only “educates” clients about the simple truth:
    Buying translation is like buying a car. You usually get what you pay for (smile)

    Liked by 1 person

  11. Fantastic post! You made a lot of good points which I’ve been thinking of but unable to put into an elegant essay. I recently did a Chinese-English post-machine-translation test for an agency and was told I edited too much. Apparently my job was to make it barely readable in order to “train” the machine to work better. So I did the test again and was told I didn’t change enough. It was a frustrating and bizarre exercise. Next time I’ll quote them your essay.

    Liked by 2 people

    • If you want to do good work, you need to start thinking like a machine, which includes thinking of yourself as a machine!


  12. LSP was an unfortunate choice of term and acronym. Unsurprisingly, interpreters and translators pushed back by stating the obvious: it’s them that PROVIDE the interpreting or translation service not the language services companies or non-profit organizations. In the labor world and SCOTUS, Individual Service Providers (ISP or IP) are a special class of workers who are independent contractors. Freelance interpreters and translators belong to this category.

    At the last ATA conference, I was relieved to hear that they are calling themselves Language Services Companies which is what they have always been. [exhale]
    For all intents and purposes, Language Services Companies are intermediaries that can sometimes provide ancillary services as well as add value to the Service Industry, the Tertiary Sector of the economy, but not always.The good news is that they have established their own association, the Association of Language Companies (ALC) that has been up and running for some time now. I’m hopeful that one day (can’t wait) they will leave or be made to leave the ATA. This is a most necessary divorce if we want to keep the children and the spouses safe from harm and out of the Domestic Violence calendar. Should the divorce happen, I anticipate some lean years for the ATA, some acrimony in the short term but in a couple of years both ATA and ALC will see the advantages of alliances here or there for common interests but this time as equal partners. It’s time to put a stop to a business model that relies on the abuse of freelance interpreters and translators and the duping of the public at large with substandard interpreting and translation services. We are all losing under the current status quo.

    Liked by 1 person

  13. Beautifully put.

    I thought we had a profound disagreement, and I am sure that we still disagree on a lot of things

    But I consider your last comment absolutely brilliant, Milena. In just a few words, you said so much.

    By the way, do you know what your first name means?


    • No. It was the only name both my parents agreed on. My mother loved Kafka and my father was Lombard so I have an aunt named Milena as well.

      Liked by 1 person

  14. It is a Slavic name, relatively common in Czech, Bulgarian and other Slavic languages. The closest equivalent in another language that I can think of would be “Aimée” in French (I hope you speak French). “Mila” means “loved or dear” in Czech and other Slavic languages and the -a ending in Milena indicates female gender.

    I researched the etymology of the names before I agreed on what names to give to our children. I guess your parents did not bother to do that, but now you know, anyway. Better late than never, right?

    Most Slavic speakers would understand what your name means.


    • Thank you. Beloved, yes I was and still am very much loved. And by the way, we agree on a lot more things that you can imagine. 🙂


  15. […] Granting this kind of extraordinary, illegal power to a translation agency would compromise your relationships with all of your other customers, both direct clients and agencies. Should they somehow find out about it, they would naturally stop working with you because signing such an agreement with one agency would mean a breach of any other contract with any other client. Not even employees may be asked to grant such permission to their employers to be spied on in their own houses. Something like that would be clearly illegal in most jurisdictions on this planet, with the possible exception of North Korea, as I wrote in another post on this blog. […]


Leave a Reply

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

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

Google+ photo

You are commenting using your Google+ 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 )


Connecting to %s


%d bloggers like this: