Posted by: patenttranslator | July 19, 2016

Too Many Tools Spoil the Translation

I made an interesting discovery recently when I was translating a fairly long Japanese utility model into English. Usually, when I am translating the text of Japanese patent applications, I automatically download a version of the machine translation of the Japanese text either from the Japan Patent Office, or from the European Patent Office website to use as an expanded dictionary, although with many caveats, of course. Machine translations into many languages, including into English, are available for all relatively recent Japanese patent applications going back more than 20 years. But they are not available for older patents and utility models, which belong to a lower category, inventively speaking.

Even in a case like that, I can simply convert a PDF file to MS Word and create machine translation manually, for example with GoogleTranslate or with Microsoft Translator. If there are just a few very slight imperfections in the printed copy, the conversion to a digital file will result in a text that is often completely nonsensical. But even if I have a perfectly legible PDF copy, as was the case with this utility model, the conversion from PDF to a digital file creates too many obstacles for a machine translation program because the spacing between Japanese kanji, hiragana and katakana characters, even though they might have perfectly regular spacing, throws off the algorithms of the machine translation program because there are no spaces between Japanese words.

It is not even clear what a word is in Japanese, so the Japanese text may look like this: itisnotevenclearwhatawordisinJapanese, or like this: i t i s n o t e v e n c l e a r w h a t a w o r d i s i n J a p a n e s e. When the wrong characters are connected together, the result is obviously completely wrong. For some reason, machine translation programs can handle the first alternative for Japanese texts relatively well, but not the second alternative.

Here is an example of what happens when a perfectly legible PDF file in Japanese is converted to digital form and then put through GoogleTranslate:

“In Tsu by the and the child that same map di-work for the person E students that Ki out and this you promote the blood circulation of the head part. Placed by the comb-shaped heat scratch preventing portion to the peripheral edge evenly of the head portion, order to carry out the heat scratch prevention treatment of all rectangular position to the head part, be sampled rate de la Lee ya when you for the use and to or cormorant child to the skin to the heat scratch you can in prevention to. Because the thermal scratch-resistant full-time (Note1) are placed at the site had close to ground clip on the head part, it is sampled Les To de la Lee You can ya di one hand in the prevention and this bovine or by the heat scratch.”

Even this kind of “machine translation” may still be somewhat useful to me because some of the words are translated correctly, such as “rectangular position”, (actually, now I remember that I did not use these words in my translation, so this must have been wrong too). But then again, because not too many words are useful, this kind of machine translation is completely useless to a non-translator.Based on the machine translation above, do you have any idea what this new utility model about a minor invention is about? Hint: it is not about “a cormorant child”. Sadly, there were no “cormorant children” in the Japanese text at all.

Incidentally, this is also one reason why I don’t bother using any CAT tools, although the main one is that I simply don’t need them, don’t like them, and despise the way predatory translation agencies are trying to use them to extort illegal discounts for “full and fuzzy matches” from hapless translators.

The discovery I made while translating without having access to a machine translation was the realization that when I don’t have to look at the words in a machine translation because there is none, I experience an exhilarating, almost forgotten feeling of freedom: freedom for me to translate the text the way I understand it, without having to pay attention to what an algorithm thinks the text means.

In the prehistoric times before machine translation was available to me and my clients (20 or 30 years ago), I would usually have three dictionaries on my desk – one to look up characters that I did not know, or used to know but somehow forgot, one standard Japanese dictionary, and one technical dictionary (often more than one).

But then machine translation came along, and instead of looking up characters, words and technical terms in paper dictionaries, I do all of that faster online now.

It speeds up the translating process if I translate a very complicated text in a field that I don’t know all that well, such as biotechnology. But if I have to look at a machine translation when I translate a reasonably simple text in a field that I know quite well, for example a patent about electric design, it slows me down when I have to look at a machine translation as an additional source of information.

And it is definitely advisable to at least take a look at an official machine translation obtained from the Japan Patent Office or the European Patent Office website because I have to assume that my client has this translation and possibly expects me to use the technical terms contained in it. If I don’t used them, I need to have a good reason for not using them, namely because I know that they are wrong.

I really like the feeling of freedom that I experience when I can simply ignore stupid algorithms.

It’s like the old times. I put on some weird music, increase the volume during an easy passage when everything is pretty clear, turn it off for a while if I come across something that for the moment I don’t understand because I have to figure it out first and the music is distracting, and then turn the music back on, but usually at a lower volume.

This, to me, is the natural rhythm of translation, or the natural rhythm of the way I translate. I think that too many tools – CAT tools, machine translation, dictation and transcription software, and God knows what other tools highly entrepreneurial tool merchants with many bills to pay and a limited number of potential customers may come up with – somehow disrupt the natural order of things in the Universe, which tends to send celestial bodies a few degrees off their usual trajectories, thus creating dissonance in the soothing and restful music of spheres that I need to hear in my head when I am on a roll as the mystery of the meaning hidden in a foreign language suddenly becomes clear to me.

Too much of everything spoils the fun for me. J. R. Tolkien wrote the entire Lord of the Rings trilogy in his spare time, usually at night while sitting on his  bed in an attic by pecking on a typewriter with only two fingers. No other tools were needed.

Tools are a good thing only if we can use them if we want to use them because we find them useful. Too many cooks spoil the broth, and too many tools spoil the translation, that’s what I think.

And when we are asked to use certain tools for our work, we have to make sure that the tools that we are told to use (or else!) will not turn us into …. tools.



  1. I never use CAT tools and never will. The translations they produce are awful and mechanical and 90% of translators are not professional editors anyway and cannot spot mistakes. As for machine translation, sometimes it is very good sometimes it is awful (Linguee reproduces some appalling translations from the web) and sometimes Google Translate for instance leaves out a negative making the translation say the exact opposite of the original meaning!

    Liked by 2 people

    • Considering that CAT tools are a translation *aid*, and don’t produce translations on their own, I’d say you need to put the blame on the translator rather than the tool 🙂 MT, of course, is a completely different matter …


      • CAT tools may be an aid for bad and inexperienced translators but for experienced translators they are just a trick for translation agencies to pay as little as possible to translators.

        Liked by 2 people

      • CAT tools’ sole and only purpose is to serve as a pretext to extort a lot of money, even for partial segment analogies (up to 50% analogies!), from a profession of freelance translators that has always been known for being poor.

        So it is all the more immoral.

        And stupid because good translators are leaving the profession, leaving end customers with juniors and amateurs, which is certainly not what they want.

        n fact, the parasite SDL has just discovered, after 30 years of so-called “experience” on the translation market, that customers want quality above speed and low rates.

        So they have totally spoiled the translation market by making intermediaries and end customers think they are morally entitled to rebates for segment repetitions.

        Even when we work for large direct customers, the extortion grid is presented to us as a due.

        CAT tool producers should be kicked out of the translation market.

        These are not computer AIDED translation tools.

        They slow us down and kill our creativity, with everything split into segments and ill-conceived terminology databases, designed by people WHO DO NOT HAVE THE FAINTESTED IDEA OF WHAT TRANSLATION IS ALL ABOUT, just like the mechanical engineer who created after only 2 years experience as an amateur technical “translator”.

        The translation market is full of amateurs, crooks and parasites.

        There are not enough barriers to this profession.

        It is a TOTAL MESS.

        I think companies like SDL should be sued for extortion, forced dumping and the like.

        Liked by 1 person

      • I am neither – I trust – a “bad” nor an “inexperienced” translator, yet I use translation memory. I have pointed out previously – several times by now, I’d guess – how TM enabled me to get through a dozen related patents in an extremely short space of time, in a way that would have been totally impossible if I hadn’t been using it.

        I’ve been using TM since late last century, although admittedly I’ve only been freelancing for 7 years. In that time, I have only come across one client who asked me to use TM and apply any sort of discount, and that was because I was doing something for them which involved a lot of boilerplate text and they obviously didn’t want me reinventing the wheel on that. I don’t remember how I charged them, but I don’t think it was on the basis of x% for exact matches and y% for fuzzies.


  2. I don’t any put blame on anyone. Why should I?


  3. “And it is definitely advisable to at least take a look at an official machine translation obtained from the Japan Patent Office or the European Patent Office website because I have to assume that my client has this translation and possibly expects me to use the technical terms contained in it. If I don’t used them, I need to have a good reason for not using them, namely because I know that they are wrong.”

    As is frequently the case, even when using human translators. I have a rule never to take the translated title of an EP, or the translation of an abstract, as gospel. They *may* be right, but they may equally be wrong: as I understand it, the translators are working completely out of context, and probably don’t have access to the rest of the text/the drawings to give them any help. I don’t use machine translations at all, but I do sometimes refer to the US version, if one exists, to see if the translator there has understood something the same way I do.


  4. The people who translate patent summaries usually work for an agency that has a contract with WIPO or EPO. Some agencies pay relatively well, most pay very little, which is why they have to work with subprime translators and even the better translators have to try to work quickly because they are being paid by the summary, not by the time they spent on it. As you said, a lot of these summaries are not very good, and sometime even the title is mistranslated – when we talk about European languages.

    With Japanese patent summaries, the problem is mostly that the as far as I can tell, all English summaries/translations on the JPO website are produced by native Japanese speakers whose English is often not up to the task.


    • It is bad enough summarising patents, not too difficult I imagine but yesterday I was approached by an agency in the Phillippines (I never work for third-world agencies, too hard, if not impossible, to collect your payment) who wanted me to summarise a lease! They had a big glossary with all the words you might need to know for a lease. If the summarizer needed definitions for “landlord” and “tenant” how could they summarise a lease, without knowing what was and wasn’t important! This was probably for some lawyer, whoever he
      /she is, do not consult them!


      • Summarize the Bible, both the New and Old Testament, in no more than 250 words – without leaving out anything important!

        Liked by 1 person

  5. “It speeds up the translating process if I translate a very complicated text in a field that I don’t know all that well, such as biotechnology. But if I have to look at a machine translation when I translate a reasonably simple text in a field that I know quite well, for example a patent about electric design, it slows me down when I have to look at a machine translation as an additional source of information.”

    Looking at additional sources of information is always going to slow you down, whatever they are 🙂 That’s why I don’t automatically look at other translations, if they exist.

    I use translation memory basically for one reason when it comes to patents: consistency of terminology and phrasing. I want to be sure that I am calling something a “connecting element” on page 1 and also on page 30, without accidentally slipping into calling it a “connection element”, for example, somewhere en route. If the terminology database and lexicon are sufficiently full, I can actually save myself a great deal of typing this way, which in turn reduces the risk of my developing RSI again, so that’s another benefit. (It’s also particularly useful when, 3 months after you’ve translated the patent claims, the client suddenly decides to get the description translated as well, because you don’t have to go through your previous translation and see how you translated everything again). Admittedly, there are (very odd) occasions when I do get an exact match or a high fuzzy match, but they are sufficiently rare that I would discount them.


  6. “I use translation memory basically for one reason when it comes to patents: consistency of terminology and phrasing. I want to be sure that I am calling something a “connecting element” on page 1 and also on page 30, without accidentally slipping into calling it a “connection element”.

    After almost 30 years of translating patents for a living, I feel that I don’t need a CAT tool for that. I just write down on a piece of paper “connecting element”, sometime only in English, sometime in both languages, and refer to it while translating. That’s all I need.

    Liked by 2 people

    • There *is* a lot more than that to it, of course. Plus writing things down on a piece of paper doesn’t help with the RSI, in my case.


  7. RSI? Repetitive Strain Injury? If that is what you mean, with all due respect, that is a weird argument.

    I understand why people who have RSI use dictation. In fact, they have no other choice and this must be a liberating tool for them.

    But writing (not typing) characters and words on a piece of paper is a different motion which is not as harmful as typing.

    I think that the best medicine against RSI is when the translator can charge decent rates because then he or she does not need to type as much to make enough money. In other words, the best medicine against RSI is to dump agencies paying low rates.


  8. CAT tools have been designed by non-translators who thought that translating meant replacing one word by another, more like technical translation.

    Yet CAT tools are imposed upon the whole translation profession, in all specialisation areas, because they provide an excuse for extorting huge rebates.

    Those rebates are based on word counts and not on TIME count.

    Just encoding one’s terminology in SDL MultiTerm takes forever.

    CAT tool segmentation kills our creativity. Even more with MT.

    99% intermediaries and 100% CAT tool producers DO NOT HAVE THE FAINTEST IDEA OF WHAT TRANSLATION IS ALL ABOUT, yet they make fortunes and translators remain poor, whereas computer technology should have taken them out of poverty – I am talking about wordprocessing systems of course.

    Who are the 65% idiots who answered in an SDL poll that they were supposedly working faster with SDL Trados?

    Measured how?

    Compared to using what? MS Word & AutoCorrect? Impossible!

    Plus 65% is far from a consensus. If it were at least 80-90% of translators answering the poll question, maybe, yes, but just over half of them is a sure sign that they did not measure anything and are just guessing.

    Who were these “translators” since the profession is open to just anyone? Amateurs? Junior translators who don’t know better tools?

    This is the poll result that SDL crooks are using to pretend that their tool makes us gain time – which is a total lie, except maybe in a few areas – and, still, I am sure ALL translators would work at least as fast with MS Word coupled with AutoCorrect macros!

    If ALL translators had used MS Word coupled with AutoCorrect macros, NOBODY WOULD HAVE EVER BOUGHT THOSE STUPID EXTORTION TOOLS which decrease quality and slow us down considerably.

    Translation schools should also play a better role at preventing that junior translator fall in the hands of such crooks.

    And that they fall prey to websites like who assure newcomers that they might not make enough money if they don’t buy 1) Trados Studio, 2) Wordfast, and if possible a third CAT tool, in order to please as many intermediaries (Proz’s actual clients, not translators) as possible.

    Proz is NOT a translators’ network (contrary to what they misleadingly advertise) and is DISINFORMING translators, to their own profit (they get commissions on every sale) and that of their real customers: intermediaries/brokers/resellers.

    Liked by 1 person

  9. At the beginning of this century, I really thought that internet would make translators less dependent on agencies as they would be able to connect with direct clients quite easily through their websites, which is not a very expensive or complicated proposition since even I was able to do that.

    Instead, translators allowed outfits like Proz to set them up for rock bottom rates when hordes of translators must compete for one measly job, and then they meekly accepted the imposition of CAT tools on them with demands for low rates or free work for “full and fuzzy matches”, which is an extortionist and illegal act, because it is simply a transparent attempt at wage theft.

    I think that there are now two types of translation markets: the market that is serviced by “the translation industry” which produces mostly mountains of crap, and the market that requires translations that can be produced only by highly educated and highly skilled translators.

    If highly educated and highly skilled translators agree to work for translation agencies, usually but not always the big ones, which keep imposing more and more restrictive terms on them, while offering lower and lower rates, our profession will soon be history and most translation will simply be crap sold as real translation by the agency hawkers and pranksters who came up with concepts like “fuzzy matches and full matches”.

    I don’t know whether this will happen, but it certainly could. Fortunately for me, I am planning to semi-retire soon, and eventually I will probably retire completely.

    Liked by 1 person

  10. A second parameter is the much later use of social media to share experiences, opinions, requests, feedback, etc. Unfortunately, by then so-called CAT tools had already settled in. But fortunately, this was before they also tried to impose PEMT upon us, which most translators seem to reject, thanks to Internet social networks, probably.

    Also, Internet social networks have allowed translators to be pushed to use other CAT tools (e.g. memoQ) than the predominant ones. So now SDL has to gradually change its extremely arrogant attitude of a company that thought it was reigning over the translation profession worldwide, with little interest for translators’ demands since they are an intermediary and most of their good customers are intermediaries – to an attitude of service, taking into account increased competition on the CAT tool market.

    Unfortunately translators have been made to believe that CAT tools had been designed in their interest (which is totally untrue), that they are working faster with them (which is totally untrue most of the time – officially at least for 35% of translators, which is a huge amount of unsatisfied users) and that fuzzy match extortion grids were legitimate. It is so anchored into their minds – and that of their customers of course, be it intermediaries or direct customers – that we can only work on the minds of new translators and even so, they will face persuaded intermediaries sooner or later.

    But today, for the first time, I am receiving an offer from an intermediary that does not request fuzzy match rebates for less than 85% analogies! Is that the beginning of a change for the best? Let’s hope so! It is a Dutch agency. Dutch agencies are good people, usually…

    Also, now is the right time for translators to ask what they want from SDL! 😉


  11. One of the most popular posts on my silly blog is this one:

    I wrote it exactly six years ago, people still read it – 4,389 people read it just last year, 5 years after I wrote it and they keep commenting every now and then. It probably has the highest number of comments among all of my posts, and I have written 618 posts ever since I have been stricken by a horrible case of the blogging disease six years ago.

    This post probably cost SDL hundreds of sales of Trados software, and I am happy about that.

    I think that social media is a powerful weapon that we translators can use to inform ourselves and to fight against things that we don’t like in our “industry”, while we still have the freedom to say on our blogs pretty much anything we want in our respective countries.

    This freedom may not last forever, bloggers in countries like China, where even the access to a blog like mine is blocked, are certainly not free to disagree with powers that be.

    So we’d better use our freedom while we still can do that.

    Liked by 2 people

  12. To answer your question:
    Well if it isn’t about “a cormorant child”, which I must admit was my first thought and something definitely worth patenting, I’m pretty sure it must be about a “heat scratch”.

    Great invention! A mechanical device to gently ease unwanted itches by gently scratching the offending area whilst applying soothing heat.

    Obvious really.


  13. I read your original article and liked it and enjoyed some of the stupid comments even more. I don’t think anyone pointed out that the main danger in using CAT tools is that you can no longer translate without them, you get into a dangerous mindset. I have always refused to use them and still get plenty of work, even in repetitive texts. I use search and replace a lot and search and replace is reliable when very similar but not identical terms need translating, in fact it is more reliable than human editing and I know because I am also an editor of English language non-fiction (and an americanizer/angliciser). “This post probably cost SDL hundreds of sales of Trados software, and I am happy about that”. Unfortunately, it probably didn’t because most new translators who studied translation are taught to use CAT tools and because most CAT (except Wordfast) tools don’t work for Macintosh anyway.


  14. 1. OK, so maybe I prevented only 10 translators from wasting their money on Trados.

    That still must count as a good deed in my karma record of good and bad deeds, which should mean that in my next reincarnation I hopefully will not be born as a venomous snake, or become a Wall Street banker.

    2. When I was young, our teachers at the university were teaching us how to learn foreign languages and how to think.

    Now instead of teaching young, budding translators foreign languages and how to think, they teach them how to use tools that will do the thinking for them so that they themselves will become tools.

    What a wonderful world!


  15. Hello. As with many here, I agree that CAT tools exist for the sole benefit of the people who sell them and that of the middle men, a.k.a., the people in the ‘translation industry.’ I have used them a few times, on online clients’ platforms, and found them somehow useful, specially in the case of documents that have too many repetitions of terms, sentences, headers, etc. They save time, I can’t deny it, but not all the time. And they are extremely overpriced for what they deliver. Why am I going to do such an investment if in my field – media, marketing, advertising – it doesn’t pay off?
    There are certain repetitions that, for sake of style, may well be translated slightly differently, basically using synonyms, which makes for ease of sentence flow and reading, it makes the text more attractive. Something that someone like me, who is also educated on writing and editing, knows well. Repetitions don’t sound good in Brazilian Portuguese, in fact, they sound awful. Sometimes the source is not that well written and, when translated with the aid of MT or CAT tools, can turn into a very clunky text in the target language. So, after you deliver the translation and the PMs, who normally have no idea of what is written in the translation as they only know the source language or not even that, do their QA, they come to you asking, ‘why is X not translated as Y as in that segment’, ‘why is this word missing in this segment’, etc, etc. A futile exercise because the time possibly saved is lost with the stress of the queries follow-up.
    I had an editing job last week that, at first glance, looked in a reasonable shape. The text came to me because the client, upon receipt of the translated and, believe me, PROOFREAD document, could only comment that ‘it read terribly, as if it had been done by MT’. My job was to fix it. As I got deeper into the text I then could see that syntax was not only poor in grammar terms, it just didn’t make sense, there were countless mistranslations of sentences and terms. The translator and the proofreader were still in defence of their work, what made me think how that the problem sometimes is not only the middle men trying to save time and make money out of the translators but also some translators and proof-readers who blindly think that these tools are great and they don’t have to do a lot of work. It is all very disheartening.
    I had a very tight deadline (the client, rightly, wanted their document revised and amended asap), and missed a few – very few – mistakes because such was the state of the document that I needed a couple of days rest after 6 days of very hard work. And the agency is now saying that they might not pay me fully because I didn’t immediately spot and fix these 10 mistakes out of 20,000 words (I did fix them once they were pointed to me).
    I could not say ‘no’ to the work but I will think twice before accepting anything of the kind again. I will read it more carefully, check the deadline and negotiate on my own terms. This if they ever offer me this kind of work again because they all seem so blind to the reality of ‘time and money saving’ of those tools in certain contexts that they don’t want to know, they just want you to say that, yes, work was fine, it just needed some adjustments that could be made in no time. They want reassurance that their investment paid off.

    Liked by 2 people

  16. Exact repetitions are considered very good style in long paragraphs in Japanese technical texts and in patents in particular, but in English they look really silly.

    So I try to vary it a little bit by using synonyms once in a while to avoid a robotic effect.

    Of course, I use the same words for the same technical terms, but that does not mean that I have to write like a robot.

    But if you rely on CATs, I think you have no choice but to be a human robot.

    Liked by 1 person

  17. Every language has its quirks that do not translate from one language to another. Hebrew is miles shorter than English, especially in some types of text (military and police, mainly) because so many abbreviations are used. What looks like a single word to someone counting words in a translation agency who cannot read Hebrew may be three or four words in actual fact.


  18. […] es estar explicando una y otra vez que… no se puede confiar ciegamente en la tecnología. Demasiadas herramientas estropean la traducción. No se le puede confiar un texto a una máquina y que haga lo que quiera. No. Es imprescindible que […]


  19. Shared this great article as a link in one of my short posts on translation technology (in Spanish):

    Liked by 1 person

  20. Reblogged this on Translator Power.

    Liked by 1 person

  21. I have just been in contact with an agency in Paris that claims to translate fashion BUT they insist that all their French to English translators use a CAT tool! How on earth can you translate fashion texts (which are not translations but transcreations, if translated literally they would sound ridiculous!) using a CAT tool? The reason is that they don’t want to pay what the job is worth.


  22. Of course. That is why I don’t use CATs, period.


  23. Great post! I haven’t used CAT tools much but looking for ways to speed up word lookup during Japanese->English translations. For example, now I have to cut and paste words into my dictionary of choice, and even the time for this accumulates after pages. So I am looking for a way to reduce this effort. Any ideas?


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: