Posted by: patenttranslator | September 22, 2016

A Few Thoughts on the Inevitability of Exponential Progress of Machine Translation

MT propagandists always stress how futile it is to try to resist the progress of technology. Machine translation keeps getting better and it’s here to stay, they tell us. Change is inevitable and translators better get with the program if they know what’s good for them. Within a few years machine translation will be as good (or almost as good) as human translation and we’d better get used to the idea that most of us will become MT post-processors. Post-processing of the machine translation detritus is a tool that translators need to add to their arsenal of useful skills.

Technological progress is constant, not linear but exponential, it’s coming at us, poor little translators no matter what we do, etc., and so on. The only other option, for those stubborn enough not to join their more obedient colleagues, is to quit the profession and start doing something else.

Then they sometime show PowerPoint slides for better impact, complete with charts and curves illustrating the progress that machine translation has made over time. It all looks very scientific to most people – with the exception of translators who, unlike most people, actually understand what machine translation is about and how it works.

Technology keeps getting better, that is true. But that does not mean that a tool can replace the humans who are using this tool. If MT propagandists, who are usually sales people, were able to look at things as they are, they would have to recognize that the progress of machine translation has been only incremental, and very slight at that, in the last few years, in fact so slow that for most people, it is unrecognizable.

Especially when we are talking about translations between disparate languages such as English and Japanese or Chinese, or English and Slavic languages, where there has been very little progress.

One reason why machine translations of claims in Japanese patents on the websites of Japan Patent Office (JPO), World Intellectual Property Organization (WIPO), and European Patent Office (EPO) are nonsensical is that the machine translation software does not understand that the verb that belongs to the topic of the claim, called wadai (話題) in Japanese, which is similar to subject in European languages, but not quite the same thing, is hidden at the end of a long sentence, even if the sentence has several hundred words.

It would seem like a bug that would be relatively simple to fix – all a programmer would have to do is tell the software how to recognize a claim and that if a sentence is a claim, the software needs to look first for the verb at the end of the sentence and only then come back to the verbs preceding the last one. But so far, the MT geniuses working on the software package have not been able to figure out this simple fact and program the software correctly. How is that possible after at least 15 years of “exponential progress”? Don’t they understand how the Japanese language works at all?

In machine translations from English to Slavic languages, for example, the endings of nouns are often wrong in just about every sentence “translated” with machine translation.

Unlike in English, the system of declensions of nouns is quite complicated in Slavic languages. For example Czech has seven cases in singular and seven cases in plural with different endings for nouns. And because these different cases are combined with many different classes of nouns that are based on the grammatical gender and the type of the noun, a very complicated system is created in this manner. But it is still a finite system that can be programmed into software if you understand how the system works in the language.

I remember when I was studying Japanese in Prague in the seventies, Izuru-san, my Japanese friend from Kyoto who was studying art history, had the whole system of Czech nouns taped on the wall above his desk so that he could learn the damn system by looking at it every time when he was not sure about the correct ending. (Which inspired me to try and do the same thing with Japanese characters).

It took Izuru-san’s remarkable but relatively slow human brain about 2 years to learn the system of nouns in Czech by constantly looking at it. But for some reason, machine translation programmers have not been able to program a relatively simple finite system of noun endings in Slavic languages into their software in the last 20 years. And this even though it is a finite system of possibilities that can be clearly defined in mathematical terms that could be easily handled by software and hardware.

The fallacy of MT propagandists is that they seem to think that a tool can replace the humans who are using this tool. They don’t really seem to understand the difference between these two concepts – a tool is not the same thing as the end result of the work that can be done with this tool.

A powerful vacuum cleaner does not replace a janitor. Powerful weapons and other military equipment costing a million dollars do not replace the intelligence of a soldier on the front line (or the lack thereof). They’re just tools, tools that don’t even understand what it is they are used for.

If MT propagandists really were knowledgeable language and translation experts, and not mere propagandists and salesmen who are trying to sell us something, or ingratiate themselves to “the translation industry”, they would have to start every lecture by saying the following words: Machine translation is not translation and probably never really will be because it is just a tool, not the actual product. Real translation is based on the understanding of the meaning of the words, which is something that a software program will never be able to do. Machine translation is a very powerful tool that can be used both by non-translators and translators, within certain, very definite limits.

And that is all it is and ever will be.

But this is not what they say. At least I have never heard “an expert on machine translation” admit that the Achilles heel of machine translation is that it is not really translation and never will be because there is this thing called “meaning” in human languages that cannot be programmed into a software package.

Instead, they are trying to sell us their version of reality – namely what machine translation is supposed to be, a version that you know has nothing to do with reality if you know anything about translation.

To say that we will eliminate the boundaries between human languages with machine translation software is like saying that people will bridge over distances between us and the things we want to do by learning how to fly. But there is this thing called gravity, combined with the fact that unlike birds, we are too heavy to fly, even if we could learn how to grow wings. So we have no choice but to buy an expensive ticket, get on a plane and entrust our lives to a pilot who knows how to use a tool that can be used by humans for flying called airplane.

I find all this talk about the inevitability of constant and exponential progress kind of infantile because it ignores the very real possibility (probability, or intermittent certainty?) of something called regress.

If 150 years ago most people lived only to the age of about 60 and now they live well into their eighties (women to the age of 84 because they finally figured out how to beat patriarchy, and men only to the age of 77 because they are not as smart as women), does that mean that as a result of unrelenting, exponential progress, 150 years from now most people will live on average to the age of 200, and 300 years from now most people will be immortal?

Actually, the way things are going, it seems much more likely the result of exponential progress in everything will be that no people will be living on planet Earth within a few decades.

Let’s hope that this will not be the case. But even if people somehow do figure out how to survive the next three centuries on a planet full of stockpiles of thousands of nuclear weapons, a planet that is heating up more and more every year, it is clear to me that even 300 years from now, machine translation will still be nothing more than what it is today, namely a tool that does not replace the end product of the work that can be done with the tool, just like a vacuum cleaner does not replace a janitor, and a machine gun does not replace a soldier.

However, considering the steady, albeit incremental progress of machine translation in the last few years, it is possible that in about three centuries, the machine translation packages on the websites of the Japan Patent Office (JPO), the World Intellectual Property Organization (WIPO), and the European Patent Office (EPO) will be able to identify the verb that goes with the noun in the topic (話題) of the claims in Japanese patents and that some machine translation packages will be even able to match the right endings with the right case of nouns in Slavic languages.

After more than half a century of incremental rather than exponential progress, maybe machine translation programmers will figure out how to teach their software basic rules of Japanese and Czech grammar within the next three centuries or so. Although it does seem to be a tall order at this point, especially since even the assumption that there will be any human life left on this planet 300 years from is at this point already a somewhat unlikely proposition.


  1. Hmm. I think this is a little bit unfair. OK, English to Czech say is going to have all the wrong inflectional endings, but Czech has just about the most fully-developed inflectional system of any Indo-European language, and English, classically called an isolating language, has almost no inflection of nouns or adjectives, and a very much simplified system for verbs. So it’s no surprise that when there is no data in the surface structure that a machine can’t guess which case to use, to take your example. But what about Czech to English? The machine will do much better, because it’s going from a data set with more structure to one with less. I use MT all the time for Romance languages to English and while it has obvious limitations it speeds up my translating work enormously. In fact I vastly prefer Google Translate as a professional tool over the ghastly Trados conspiracy, Wordslow and all the other anti-language products out there.


  2. I agree, machine translation is an excellent tool, just like a vacuum cleaner, or a machine gun.

    I use it all the time too (MT and vacuum cleaner, not machine gun).

    What I am saying is that MT is not translation and never will be, because it is only a tool, and every MT propagandist should start her lecture by saying precisely that if she wants us to take her seriously.

    For some reason they never do that. They always start by saying “It’s not perfect yet, but …..” so that most people who don’t understand what translation is are led to believe that one day it will be perfect, sooner or later.

    Liked by 2 people

  3. MT propagandists, who are usually *sales people*, ARE quite able to look at things as they are: get translators to accept to lower their rates—and therefore increase margins further—by promising implicitly that MT will do 99.99% the job tomorrow.

    MT is translating words. Translators are conveying concepts and ideas. Making them sound natural in the target language. And no computer will be able to emulate this type of brain work in a near future.

    Why would a translator accept to leverage (i.e. clean up) MT junk at a reduced price when you can do the same at your full rate?

    Liked by 2 people

  4. Because he may be convinced by MT propagandists that he has no other choice.

    Liked by 2 people

  5. “Machine translation is a very powerful tool that can be used both by non-translators and translators, within certain, very definite limits”.
    The words ‘effectively’ or ‘expertly’ come to mind in this context. I can use a golf club, but it would be a serious exaggeration to say that I can use it either effectively or expertly :-). I suspect it requires a modicum of talent, a considerable amount of coaching/training, and years of practice.

    Liked by 3 people

  6. This just in: we’ve got a talented performer in our corner, making the point about nuance in harmony.

    This may inspire a whole new blog post. Enjoy!

    Liked by 1 person

  7. Well it’s the same old story people trying to flog their ware and to and to entrap you. However do consumers always get what they are being promised do we get from TRADOS or the like, what their marketing drivel promises? Have Microsoft ever produced anything that works reliably? How many people have had their computers crash after upgrading to Windows 10?

    Quite a few years ago in the course of a business simulation (English training in Germany) I learnt that if a company secretary gets the saying order company, the company is doomed, since company secretaries (or others in the company who look after the profits and the money) look at money only, they give a hoot whether a product works reliably or satisfactory – it must sell


  8. In fact I decided to do a little experiment with Google Translate this morning to demonstrate what I mean. Below is a passage from Wikipedia in Czech. Then there is a Google translation into Russian, which looks pretty good. Then there is that same Russian text translated back into Czech. You will see that a few words are different, but the effect you describe of having every single case wrong is not visible at all. It reads like grammatically correct Czech. So it is not that the system is not capable of dealing with Czech: the problem is a technical one of some language pairs being more difficult than others.

    Let me also add that translation of Japanese is another red herring. The fact that Japanese mainly uses a writing system developed for an unrelated language, plus the cultural peculiarities of Japan, makes Japanese translation (as I’m sure you know) into a highly specialised operation. A famous example of this was the attempt to repeat the Nuremberg Trials in Tokyo after the defeat of the Japanese. One reason the Nuremberg Trials had been successful was the use of IBM’s new simultaneous interpreting (why does everyone say ‘interpretation’?) system. If you think of your mental images from that, they are all wearing headphones. The working languages of the proceedings – German, English, French and Russian, I believe – all went quite smoothly together. The attempt to do the same thing in Japan failed completely, as the Japanese counsel would stop the proceedings every few minutes to protest about mistranslations. For that and other reasons, the Tokyo Trials were eventually abandoned half-way through, and most of the accused were simply released.

    Literatura uvádí výskyt sociální fobie u 3–13 % populace v rozvinutých zemích. Tito lidé jsou též často považováni za introverty a sociální fobie za nadměrnou úzkost, na kterou neexistuje profundovaný „lék“. Akademická sféra tuto nemoc dlouho opomíjela a její výzkum, včetně zkoumání léčebných technik, se začal rozvíjet až v 80. letech 20. století. Dnes se lehčí sociální fobie léčí psychoterapií, v těžších případech psychofarmakynebo kombinací obého.

    В литературе показывает частоту социальной фобии в 3-13% населения в развитых странах. Эти люди также часто рассматривают интроверты и социальную фобию для чрезмерной тревоги, для которых profundovaný “лечение”. Болезнь Academia и его долго не уделялось должного внимания научных исследований, в том числе изучение целебных техники начал развиваться в 80-х годах 20-го века. Сегодня легче социальной фобии лечение с психотерапией, в тяжелых случаях psychofarmakynebo сочетание обоих.

    V literatuře se uvádí výskyt sociální fobie v 3-13% populace v rozvinutých zemích. Tito lidé jsou také často považován za introverti a sociální fóbii pro nadměrnou úzkostí, u nichž profundovaný “léčbu”. Academia onemocnění a jeho dlouho zanedbávané výzkum, včetně studia léčivé umění se začala rozvíjet v 80. letech 20. století. Dnes je to jednodušší léčba sociální fobie s psychoterapií, v závažných případech psychofarmakynebo kombinaci obojího.


  9. Whenever you see that a text translated by GoogleTranslate from English to Czech makes perfect sense an is perfectly grammatically correct, it means that GT found an existing Czech translation that was done by a human.

    Your comment has not been translated yet by a human, and although you use simple, short sentences, there are many more grammatical, stylistic and factual mistakes in the GoogleTranslate version of your comment than in the translation you compared after you ran it through GT to translate it into Czech and Russian, (although there are many mistakes in your example too, you see them, right?):

    Dovolte mi také dodat, že překlad z japonštiny je další falešná stopa. Skutečnost, že japonská využívá především psací systém vyvinutý pro nepříbuzného jazyka, plus kulturní zvláštnosti Japonska, dělá japonský překlad (jak jsem si jist, že víte) do vysoce specializované operace. Slavný příklad tohoto byl pokus opakovat Norimberském procesu v Tokiu po porážce Japonci. Jedním z důvodů Norimberském procesu byl úspěšný byl použití nových simultánním tlumočením IBM (proč všichni říkají, “výklad”?) Systému. Pokud si myslíte o svých mentálních obrazů z toho, že jsou všichni s nasazenými sluchátky. Pracovními jazyky řízení – německy, anglicky, francouzsky a rusky, domnívám se – to vše proběhl zcela hladce dohromady. Pokus udělat totéž v Japonsku zcela selhaly, protože japonská rada by zastavení řízení každých pár minut na protest proti nesprávnému pochopení. V případě, že i z jiných důvodů, Tokio soudy byly nakonec opuštěný v polovině cesty přes a většina z obviněných byli prostě propuštěn.

    Of course you understand the machine translation into Czech – you’re the one who wrote the original text in English. But most people would understand probably only something like 70% of what you said if all they had was the machine translation, and 30 percent would be not understood or misunderstood.

    My brother in Czech Republic, who does not know English and sometime translates my posts with GT into Czech, told to me many times that he does not understand the machine translation much, although I would of course understand the machine translations because I am the author of the English original.

    It happened to me a few of times that I printed out a machine translation of a Japanese patent and to my consternation, I saw that it was perfect, just like a human translation. But as I continued translating the Japanese text while looking at the machine translation, I realized that GT simply substituted an English version of the same Japanese patent that was filed with the US Patent Office when I saw that there were passages in the English machine translation that simply did not exist at all in the Japanese version because there were some differences between the version that was submitted in Japanese to JPO and the version that was submitted in English to USPTO.

    And the reason why a real, human translation was needed was the fact that the law firm needed to know what the differences were.

    I was translating a few days ago a complicated text dealing with Japanese epidemiology research. It was complicated mostly with respect to specialized terminology in this field, which I was not very familiar with. When I ran passages that I did not understand through GT, about half of the terms that I did not know at first were correctly translated with GT, the remaining half I had to find on Linguee and other online resources in Japanese (which sometime also contain English equivalents that Google was able to find for me).

    But the way the GT put the machine-translated sentences together in English made it completely impossible for somebody who does not read Japanese to understand what those sentences meant. It was very much like the GT translation of the song on YouTube provided above by Shunra.

    So GT was an excellent tool for me – the translator. It saved me a lot of time because I did not have to look for words in a dictionary and I don’t have such a specialized dictionary in this field anyway.

    But GT would not be much of a tool in that case for a monolingual person who needed to understand what the real content of that Japanese paper was. That was why this person had to pay a lot of money for a human translation.


    • “It happened to me a few of times that I printed out a machine translation of a Japanese patent and to my consternation, I saw that it was perfect, just like a human translation. But as I continued translating the Japanese text while looking at the machine translation, I realized that GT simply substituted an English version of the same Japanese patent that was filed with the US Patent Office when I saw that there were passages in the English machine translation that simply did not exist at all in the Japanese version because there were some differences between the version that was submitted in Japanese to JPO and the version that was submitted in English to USPTO.”

      And then we get back to the old question of: where did Google get that translation from, who originally translated it, and has Google breached their copyright and/or paid anyone for the right to use it?


  10. Yes of course. Sometimes with legal translations it’s the same thing. If you are lucky you get a perfect translation from Google because an identical or near-identical text has been included in the database.

    I have a friend who regularly puts passages from English or Dutch or French into Google Translate just so he can laugh about how useless it is. This seems strange and pointless to me. The better you understand its technical limitations the better you can use it. But you know all this.

    The task of providing monolinguals with reliable accurate machine translation is an impossible one. If you don’t understand that languages are different from each other then you won’t give the system appropriate tasks, and will be frustrated when the jokes and double meanings and odd distinctive phrases you naturally use in your own language don’t make it out on the other side.

    Liked by 1 person

  11. I think part of the problem is that the people who work on machine translation software are just programmers who are technical geeks but not translators. It is not just the problem of the inability of GT and other programs to find proper endings for nouns and verbs (a technical person would not be likely to pay much attention to something like that if his first language is English).

    I often wondered why these programmers don’t put together for example a list of equivalent or similar proverbs, idioms and sayings in English and other languages. Cultural heritage like this is almost never correctly translated by MT.

    I think that the reasons for this is again that the people who work for GT, Microsoft, etc., are mostly technical geeks who are not that much interested in the cultural aspects of different languages, that may be also one reason why GT users are asked to “correct” translations so that the GT employees would not have to waste their times with little details like that.

    Liked by 1 person

  12. Having studied Czech at Lancaster University I became familiar with the fact
    that there are a good many declensions of nouns, pronouns, and adjectives
    and even the numerals are declined for case. Happily, Josef Fronek who was
    there at the same time has included grammatical tables in his excellent Czech-English dictionaries.


  13. I have his dictionary in my library of dictionaries, right next to that of Ivan Poldauf who was one of my English teachers.


  14. “A powerful vacuum cleaner does not replace a janitor. Powerful weapons and other military equipment costing a million dollars do not replace the intelligence (or the lack thereof) of a soldier on the front line. They’re just tools, tools that don’t even understand what it is they are used for.”

    Robotic vacuums and lawn mowers have gotten better and cheaper. The Air Force once used a high school drop out in the Iraq war. For one mission, the air force squadron was replaced by a high school drop out who was controlling nine drones at once from a room in D.C. He happened to be very good at video games.

    MT will just keep getting better and better. I read a Honyaku Group question today that asked about vagueness.


    GT: Compounds having one or more primary amine groups

    Another translator mentioned this: “they could have eliminated the ambiguity by choosing the phrasing 一級アミン基を有する1つ以上の化合物,
    leading to the conclusion that the latter of your interpretations is correct.”

    GT: One or more compounds having primary amine groups

    That level did not exist even three years ago.


  15. Thanks for your comment, Jimmi Hendrix’s Ghost.

    Yes, the lack of plural in Japanese can be a problem, especially when translating patents. Personally, I don’t understand how a language as elegantly put together as Japanese could have overlook such an obvious deficiency.

    I have just one other comment: A high school dropout may be much more intelligent and competent than somebody who went to and graduated from the best schools in America. I have in particular in mind Edward Snowden and a certain US president (perhaps you can guess which one).

    Liked by 1 person

  16. […] MT propagandists always stress how futile it is to try to resist the progress of technology. Machine translation keeps getting better and it’s here to stay, they tell us.  […]


  17. Plurals is an excellent example of my original technical point. Chinese has no plurals except in personal pronouns (我们,咱们, 你们,etc.), so plurals in other languages being machine translated into Chinese can just be eliminated, but you can’t make a machine that is going to be able to perform the reverse operation correctly every time. It is to do with technical features of the particular language pair in question.

    As for the question
    ‘I often wondered why these programmers don’t put together for example a list of equivalent or similar proverbs, idioms and sayings in English and other languages. Cultural heritage like this is almost never correctly translated by MT.’

    This is just like my Dutch friend who happily sits there putting Dutch proverbs into Google Translate so he can laugh at the meaningless results in English. He uses this to prove to himself that it is ‘a very primitive system’. Patent translation and a lot of other technical translation is a different matter, but doing what you could call ordinary-language translation it is an everyday experience that you come across idiosyncratic expressions that just can’t be translated. Every language has these oddities. A human translator doesn’t exactly translate them, but invents some way around them. Obviously a machine can’t do that at this level of technology.

    A strange exception to this is translating French. It has often happened to me that when I produce a translation of a French text, particularly university-related French texts, the client comes back with a ‘corrected’ version in which all of these lexical oddities have been put back in, with word for word ‘translation’ which should have some other name. I think the reason for this is that they are convinced that French is so superior that it will be degraded by being translated in the normal way. French academics nowadays are expected to publish in English, and I recently worked on a paper on the philosophy of mathematics written in English by a man who was evidently very clever but who clearly couldn’t speak English. But he was capable of doing this pseudo-translation, producing a paper ‘in English’ that could in fact only be read by French-speakers. I think this is the reason for the popularity of French theorists among English-speaking university students. These texts are impossible to understand, but have a weird, hypnotic effect, as if they would start to mean something if you stared at them long enough.


    • Thanks Christopher,

      It seems that rules in a language are lexical oddities in another.

      Maybe it’s because brain connections are different from one language to the other.

      Maybe it’s because brain connections of French academics are unlike those in the rest of the population.

      Or maybe it’s because they just want to hypnotize you 😉

      At any rate, thanks for this idea that I may use to sell my French translation services in more creative ways!


  18. “These texts are impossible to understand, but have a weird, hypnotic effect, as if they would start to mean something if you stared at them long enough.”

    Ha, ha, ha, that is precisely the effect that I am aiming for in my blog posts.

    Alas, it is probably too presumptuous of me to even try doing something like that. Not being French, I can never quite succeed.


  19. Hello Steve,

    Have you read about Lionbridge’s recent patent for their GeoFluent solution? They say that first-generation machine translation doesn’t really work, so now they have second-generation, and it has a “proprietary linguistic layer” on top of the machine. What is this??

    I sympathise with everything you say, but at the same time, I also sympathise that there are customers with very complex problems who are looking for solutions. I think freelancers shouldn’t take this too personally. Ideally, I wish there were a way that LSPs could/would integrate freelance translators more in the entire process and two, perhaps we as freelancers need to integrate ourselves more in the wider content management industry to contribute to coming up with solutions.


    • Phyllis, sources inside the company have informed me that the “proprietary linguistic layer” consists of human capital purchased on the discount slave markets for burned-out post-editors. After appropriate reconditioning and medication (patent pending for the process), said capital is ready for deployment at affordable rates, 24/7 in any brave new language market.


      • Ah I see. Thanks Kevin. GeoFluent is thus a true Mechanical Turk for the translation industry.


  20. I have been offered work a number of times by Lionbridge, but I would never work for them.

    I am not interested in becoming integrated into “a content management industry”.

    I am offering an alternative to this industry for providing real translations instead of garbage manufactured by machines and shaped into semi-credible form by slave-like humanoids subservient to these machines. I have a different business model and that is what I am interested in.

    This business model that I am interested in has existed for hundreds of years and still exists independently of “the translation industry” or “content management industry” as a profession that is not “a proprietary linguistic layer on top of a machine” (a pretty disgusting idea to me), but a real profession that is still alive despite the attempts of “the translation industry” to kill it.

    I believe that this model is very much needed as an antidote to “the translation industry” and “content management industry” (what a funny name).


  21. By “content management industry” I was referring to the relatively new problem of companies having large volumes of content which they are producing in a generally haphazard and disorganised way, especially with business becoming so international, that one division may be producing content which is not compliant, without another division even knowing about it. This obviously involves multilingual content. This is more to do with marketing though and I know this is not your field.

    I actually find that people in the “content management industry” are among the most knowledgeable and unbiased people when it comes to the translation process, including the role of freelance translators. It was very refreshing to me to discover that people who are higher up the strategy chain than translation agencies are actually aware of the importance of translation quality and struggle to help companies understand it.

    Content management is not about translation, it is about organisations applying a consistent strategy regarding the content customers receive from them, of course this includes translation and localization. People in content management actually suffer because of translation agencies and are very aware of what is involved in translation quality issues. They do complain that all LSPs look alike, as GALA found in a survey earlier this year.

    I know marketing is not your field, but I would recommend looking at Val Swisher’s “The Five Phases of the Translation Workflow” at She is not a translator by the way, I think she used to be a writer before the content strategy consultant she is today. She even has a very short video on “The Content Wrangler” where she says don’t use machine translation.

    This is what I meant when I said that I think freelancers need to educate themselves about, and integrate themselves into, the wider world of what is going on. Freelancers, being at the production end, are at the very bottom of the chain, and we will of course be left behind if we don’t demonstrate an interest in the needs of other stakeholders in the general economy and somehow also contribute. This is maybe even one of the reasons why LSPs are trying to replace us with machines because it is not as if freelancers are collectively very innovative or aware of other people’s problems, being so isolated as we are.


  22. Marketing is indeed not my field so I will not comment much on what you said, Phyllis, except to say that I went to your blog and I think you have the right idea to call your business “language translation consultancy” (if I remember it correctly), with the emphasis on “consultancy” as opposed to “freelancer”.

    Freelancer has become a much misused, abused and overused term, which to most translation agencies a pliant, obedient and cheap person to be exploited as much possible.

    This is a relatively new phenomenon and I will try to say a few more things on this subject in new posts.


  23. Yes, I do refer to it as a “consultancy” to emphasise that by now I think I have enough experience beyond just actual translation to collaborate as a business with other businesses to help them achieve their goals, which may be that they in fact need an LSP and not me.

    I think there is nothing wrong with being a freelancer if that is the stage where you are. Everybody starts somewhere. I mean, “consultant” shouldn’t be abused either. I see it like graphic designers. I have actually been talking to graphic designers a lot lately because I find they have a very similar problem to translators, that they are trained to “produce” and “make”, and to a high a quality too, but they have very little guidance about the business of design, and the world is changing and customers’ needs are changing in a way that is decimating them if they continue to remain excessively focused on producing.

    I think translators have a similar problem that of course, we need people to produce to a high level, but if we are doing it in a way that is not meeting the customer’s overall needs, then they don’t need that from us, and will take a machine instead.


  24. 1. I think that most customers who swap human translation for MT, even post-processed MT, will quickly come back to us, humans, again, despite all our faults, with the exception of cases when the quality is unimportant, such as when translations need to be done because somebody high enough ordered it but they are not really needed and nobody really reeds them. (There are probably many cases like that – excellent business for some outfits)!

    2. I wonder whether graphic artists are also exploited to the same incredibly shameless extent by graphic and advertising agencies as translators are by translation agencies (also known as LSPs or Lame Service Providers).


  25. Re: 1. Common Sense Advisory has done a study on this. I haven’t read it because you need to buy it, but apparently customers become “localization mature”, or something like that. Meaning, it takes them time to mature as translation buyers, and the more mature they are, the more they understand the difference between the many different providers and value talented freelancers more.

    I think it would be great if LSPs could somehow get over their fears of freelancers and integrate them in their processes, because from the customer’s point of view, it would be ideal to have the processes and management of an LSP as well as production expertise. I really don’t know how LSPs developed in this way. Every other industry invests in their talent and accepts that there is always the risk that if you don’t retain loyalty, the talent you trained will leave for a competitor or set up their own shop.

    Re: 2. I don’t have the impression they are highly exploited by agencies. They don’t talk about their agencies the way translators do. They of course work as freelancers but they are integrated in agencies too. They have contact with clients and higher-ups and learn a lot. Graphic agencies themselves seem to be suffering from the changes in the world though and look back on their glory days. One problem we don’t have that they do is that there are excellent graphic designers living in Asia for example who are very cheap, whereas with translation, you can’t direct a translator to push a pixel this way or that or produce a certain concept. You would need the actual expert translator to carry out the work.


  26. Translators have the same problem with competition from countries where the cost of living is much lower than in Western Europe or US.

    There are many smart people in Africa who want to translate French and English, and in India who want to translate for example Chinese, Japanese or German to English. Their knowledge of languages (including English) may not be that good. And the cost of living in China is also much lower for the most part than here.but they are willing to work for a fraction of the price that people living in more affluent countries would need to charge, and what they produce is probably still much, much better than MT.

    I know this because they want to work for me and I receive their resumes all the time. Sometime they even change their names so that they would sound more Japanese or German. They mostly compete directly with translators and I am sure agencies dearly love them.


  27. Hi Steve, just wanted to let you know that our little exchange inspired me to write a post ( and I have linked back to your blog. I would have preferred to write to you directly rather than here but couldn’t find an address anywhere.


    • I have not lived my life in vain!

      But I have to say, machine translation is NOT SUITABLE FOR TECHNICAL TRANSLATIONS! At all! That is just something that people who know nothing about technical translations think.

      As a technical translators of some 30 years, I think that machine translation is about as suitable for technical translation as it is for biblical translation.


  28. Regarding MT offerings from the LSP honchos, apparently they tried to leverage their existing translations to train MT engines. This occurred more or less at the same time as “big data” became fashionable, so presumably management decided to leverage their databases to try to get rid of at least some of those pesky freelancers.

    The theory behind big data is that AI performs better when trained with more data. The quantity of data is often far more important than the algorithms employed. As more and more huge datasets became available, practitioners saw the possibility of obtaining substantially better results. It’s a safe bet that some big LSP’s tried to join this bandwagon.

    However, I found the LSP’s MT offerings were far inferior to Google Translate, at least into/from Portuguese. A few years ago I participated in a LinkedIn discussion about this, and other translators said the same about German and (I think) Dutch.

    The uncircumventable, painfully obvious fact is that the LSPs and their “precisely tailored language engines” are far outclassed by Google in terms of money, IT talent, and computing horsepower. But this doesn’t stop their salespeople from trying to ram their dreck down everyone else’s throat.


    • Swamping powerful computers with “Big Data” aided by smart algorithms was supposed to result in much improved results many years ago. But it never happened because there are just too many variables to deal with. When you deal with a translation, the only arbiter of reliability that is any good is human brain. Even a mediocre, substantially underperforming human brain is likely to do a much better job than the best technology imaginable. I was again translating claims today while surreptitiously glancing at a machine translation. After a while I had to put Google Translate away because it was again supplying a set of slightly different patent claims, and it was messing with my own brain too much because it was so similar …. but wrong!


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 )

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: