Posted by: patenttranslator | May 27, 2015

But it only makes perfect sense because it is not really a machine translation

Last week I gave two different cost estimates to two different clients for translating Japanese patents. One document was very short, only two pages, and my estimate for translating it into English was for 240 dollars.

My price quote was accepted, I translated the short document, printed, proofread and delivered it, and I already received payment for it.

The other cost estimate was a little bit more complicated because it was an estimate for translating two very long patent applications from Japanese, each of them about 60 pages, including about a dozen pages with figures.

My estimate for this potential project was about seven thousand dollars for a non-rush translation (at the rate of no more than two thousand words per working day), or about twelve thousand words for translating both patent applications on a rush basis.

I know that when I submit estimates for expensive projects, nothing will usually come out of it because the cost is so high. But sometime I do get the job, so I do it anyway, although preparing such a cost estimate can be very time consuming. You have to be careful because while you do not want to present an estimate that is too high (because it may be compared to other estimates), you do not want to submit an estimate that is too low either because if your offer is accepted, you will be stuck with the quoted price.

The second cost estimate (for two long documents) was not even acknowledged. Maybe the patent lawyer was mad that a mere lowly translator would dare to ask for so much money (for a simple translation!), or maybe his mother never taught him that it is polite to say thank you when people do something for you, even people who make less money than you, especially if they do it for free, although it takes them a long time to do it because it must be done right.

I think what happened with the second cost estimate was this: the patent lawyer looked at my cost and time estimate, sent it to the client, and his client said something like: “What? No way, this is much too expensive! Could you use machine translation instead to find the information that you need?”

So the patent lawyer went to the EPO Website, downloaded machine translations for both long patent applications, and since the machine translations that he saw made perfect sense, he assured the patent law firm’s client that a translation would indeed not be necessary in this case.

Well, I did the same thing, because I generally download a machine translation of patents that I am translating and then look at it while I am translating, especially in the beginning when I am still trying to establish the terms that I will be using. I normally have a separate folder for each patent with several related files in it, such as a PDF file and a text file of the document, and a machine translation of the document if one is available.

Just so you know what a typical machine translation of a paragraph from a Japanese patent looks like, here is a sample:

“The provision 3 is molded simultaneously stop sliding on the occasion of this extrusion molding. As long as it is this material that has ****** and for which it produces the skid effect although the provision 3 is molded with a rubber material stop sliding, what kind of material may be sufficient as EVA, an elastomer, etc.”

The machine pseudo-translation really only makes sense if you can read the Japanese text, it is helpful to me, especially when I start translating, although it is not very helpful to a client who cannot read the original text. Which is of course why he ordered a real translation of it in the first place.

I too downloaded machine translations of the two large files from my other cost estimate, the expensive one that was ignored. When I looked at them, I could hardly believe my eyes because the translations were …. perfect. I kept looking at the Japanese text and comparing it to the English text, but I could not see anything that would need to be changed.

And then I remembered that not so long ago, when I was translating another Japanese patent, I found a machine translation on the EPO website that was also perfect. I also remembered that the Japanese patent that had a perfect machine-translated version was in fact only the Japanese version of a patent application that was originally filed in English, just like the two long patents from my current expensive cost estimate were originally filed in English.

In that other case, several months ago, the client asked me to go ahead with the translation anyway. It was a much shorter document, so the final cost was only a few hundred dollars. I was translating that document while looking at the machine-translated version, feeling rather foolish … because what was the point of translating it when the machine-translated version was so good … until I noticed a paragraph that again made perfect sense in English, but was not included in the Japanese text. And then I also found paragraphs that were there in the Japanese text, but not in what was presented on the EPO website as a machine-translated version of the same documents.

At that point I realized that the client who ordered the translation probably knew that what one can download from the website as “a machine translation” is in fact sometime the original document that was filed in English, not a machine translation.

The reason why “the machine translation” was so perfect was the fact that instead of a document translated with software algorithm, it was the original document that was written in English. Incidentally, this is what GoogleTranslate is based on: instead of using algorithms for “translating text with a machine”, it uses algorithms to search for, find and substitute a similar text in another language for a translation.

The reason why the Japanese text in certain passages did not correspond to the English text was that the “benrishi” (patent agent) who prepared a patent application originally filed in English for filing in Japan modified the text of the application in Japanese, added different content to it and removed other content from it to make the text compatible with Japanese patent law.

After a while I stopped even looking at the “machine translation” because it was too distracting.

Sometime only the claims are changed in patent applications that were originally filed in another language, sometime portions of the text, and sometime a lot of the text in the patent application is changed in this manner.

Does the lawyer who asked me for a cost estimate for those two long patent applications know that? He probably does, and maybe he even told that to the client who decided not to go ahead with an expensive translation project.

But the fact is, neither the patent lawyer, nor his client have any idea what is really contained in the patent application that was filed in Japan, because what was stored on the EPO website and presented as “a machine translation” was not really a translation at all.

It made perfect sense in English, but it only made perfect sense because it was not really a translation, let alone machine translation.


  1. Well, the first thing I do when faced with a Japanese patent publication (or indeed one in any language other than English) is to look at the patent family in esp@cenet and see if there is an “equivalent” English-language application. But as you note, there can be differences.
    If the Japanese application is the national phase of a PCT application, then the Japanese text ought to be a verbatim translation of the original PCT text: if that’s English, you’re home free; if the PCT is in some other language but there is a corresponding Australian or US national phase application, for example, that again ought to be a verbatim translation of the PCT and hence the same as the Japanese. If your PCT original is Japanese, just look for English-language national phase applications from that PCT. But this only works with PCT national phase applications.
    The difficulty arises with non-PCT applications, where “equivalents” (applications with a common priority claim) may have been tweaked into good local form before filing in the various countries. And the US further complicates it with continuations-in-part: is your foreign publication textually similar to the parent or the CIP?
    Also, esp@cenet is not 100% reliable on patent families – though in my experience they are more often over-inclusive than under-inclusive.
    Still, it’s a good start if what you want to know is what the application is about rather than necessarily word-by-word detail.


  2. To supplement what I just wrote, this applies to published applications only – the claims of patents potentially differ from country to country.


  3. Thank you for your comment.

    Thanks God for the complications – you never know what’s really in it until and unless it is translated.

    Except when the document is too long and the translation would thus be very expensive.


  4. “the claims of patents potentially differ from country to country”

    Thanks God for that too.

    I have been working on two projects involving translations of claims only for quite some time, one of them for about 7 years now.


  5. Good information here guys. We (small translation company in New Zealand) occasionally translate patents and have some understanding of these sorts of issues – more now, thanks. So basically any part of an existing reference translation/text can differ from what you’re translating, and you never actually know by how much and in which parts until you do the job and look closely at the precise wording throughout. Many traps for the unwary there!

    We find quoting for these jobs tricky. You can ignore any possible help/time saving from existing related translations/texts and just charge on word volume, but that can be pricey and as you’ve indicated often the quote won’t be accepted. But the alternative isn’t easy – trying to assess how useful the other text is and making an allowance for that. And very time consuming as you say. I’m not sure we have much more success when we do that anyway.

    Not even acknowledging your quote is bad form, but pretty typical in our experience – across all industries. We have a quote follow up process for the jobs we want to get, and have modest success with that. Enough to make it worthwhile doing anyway.

    Enjoyed the post.


  6. Thank you for your comment, Dennis.

    I wonder, what is your follow up procedure?

    I would like to learn something from you too :).


  7. When we quote we enter a follow up time/date in our project management system, and then get an alert to follow up at that time. It can be anything from an hour to a couple of weeks or longer depending on client, urgency, stage the project is at etc. We only do it for jobs we definitely want, the others we mark as no follow up.

    Our best results generally are from phoning the client, but that’s not always possible these days and getting an answerphone is no good. Also, some PMs are pretty reluctant to phone clients. If not, we do it by e-mail:
    “We were wondering if a decision has been made regarding the project for which we quoted below? Please let us know if there is anything else we can help you with for this project.

    If our quotation was unsuccessful, we would value any feedback you can give us on the factors which influenced your decision – price, turnaround time, etc.

    Thank you for your time.”

    Most clients respond to this and tell us where they’re at or why we missed out. Looking at it now I might tweak the wording a bit … I think the key is not to be pushy and to ask for information. And to make it look personalised rather than a template.

    Hope this is useful.

    Liked by 1 person

    • Thank you (both) for this description of processes.

      We do something similar (although we haven’t automated our reminders), and have noticed a slow move away from voice communication over the years. Clients seem phone-shy.

      We’re trying to cultivate actual, human relationships with the people we work with, and project follow-up is among the nice side effects of this.


      • “We do something similar (although we haven’t automated our reminders), and have noticed a slow move away from voice communication over the years. Clients seem phone-shy.”

        The marketing industry is slowly killing phone as a means of communication. It already killed e-mail, for the most part, as a means of communication, because 95% (probably 96.7% by now) of our e-mail is odious marketing junk. Young people no longer use e-mail. Although they still need to have an e-mail account it they go to school, etc., they use social media and various cell phone applications instead of e-mail.

        After e-mail, phone communication will be the next victim of the marketing industry.

        I have a talking phone call ID, which means that I listen to the machine announcing who is calling, and unless it is a name I recognize, I don’t answer the phone anymore as I used to just a few years ago.

        That may be why your clients too seem phone-shy.


      • Replying to our host: 90% of the phone calls I get are odious marketing junk. Be it “Rachel of cardholder services” or boiler-room “wearecallingfrommicrosoftyouhaveaproblemwithyourcomputer” scams, that’s where most of the traffic on my home line originates. Things are a bit better on the business number. But not a very *big* bit. Business communication has gone over almost entirely to written form, usually by email. Thinking about this on my work, I would guess is that it is in order to create a solid “paper trail”.


  8. Very useful.

    I think I will start using your system with the follow up e-mail for really interesting projects.

    Thank you.

    Liked by 1 person

  9. Great. Good luck!


  10. This is an incredibly valuable exposition. Consumers of translation services absolutely must see examples of both nonsensical MT products and products that [albeit accidentally, at times] reveal the misrepresentation/falsehoods that bolster up the now widely held view that accurate and precise products can reliably be produced via MT. Dropped content as a dead giveaway that a human was involved is a great investigative finding, for example. The problem remains, however, of the client whose staff includes no person competent to troubleshoot finished products for collateral evidence of errors and false representations regarding the mechanisms used for translation. . . .I think this phenomenon is rampant due to an even larger problem, namely the widening and deepening monolingual illiteracy (in both technical and nontechnical language) of business and other professionals coming up in the ranks today. . .


  11. “I think this phenomenon is rampant due to an even larger problem, namely the widening and deepening monolingual illiteracy (in both technical and nontechnical language) of business and other professionals coming up in the ranks today.”

    Monolingual illiteracy (I like the term) is a worldwide phenomenon, but it seems to be particularly pronounced in English-speaking countries.

    Instead of study of foreign languages, they have spelling bees here in US, where kids who can memorize how to spell words in their own language without having a clue about the etymology of the words or any knowledge of any other language are celebrated as super intelligent geniuses.

    In most other countries, if you don’t how to spell words your own damn language, you are considered pretty stupid.


  12. Of course, as you know, the EPO site will tell you if the application has been filed in other countries. The patent attorneys are of course aware of this too but it is always worth checking, to compare terminology etc. I have heard it is also possible that if another translator has prepared a translation of the same document using Google translate, then that translation or some of the terms selected may be available too. I would be interested if anyone can confirm this.


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: