When the translation is the prior art: or is it?

Translations have long engaged IP law and practice. In the 19th century, the fundamental issue focused on whether copyright extended protection to a translation. After an initial period of disagreement (U.S. law said “no”, English law said “yes”), consensus was reached in favour of protection. Later on, the right to translate reached the international arena, as dispute arose at the Berne convention level about whether to allow translations to take place in developing even without explicit authorization in name of encouraging the flow of information. The issue of translation has taken on a more meat-and-potatoes aspect in the patent system, namely the requirement for translation of the patent application and the attendant costs as an aspiring applicant seeks to achieve broad scope of protection. But even in the patent system, translations seem to raise legal questions, as this Kat was reminded this week as he exchanged coffee-klatch thoughts on the subject with a colleague.

In particular, this Kat got to wondering about how translation interacts with a determination of whether a given document constitutes prior art. Let’s assume that the document at issue is written in a language other than that of the jurisdiction in which the patent is being adjudicated, such that neither the parties involved nor the adjudicator have any direct understanding of it. As such, each side commissions a translation. Not surprisingly, the two translations are not identical. More than that, however, they differ with respect to the translation of several key sections. If one relies on the translation of Party A, the reasonable conclusion is that it does not constitute prior art. However, the translation of those sections as contained in the translation submitted by Party B lead to the opposite conclusion. So is, or is not, the document prior art?

If no issue of the language of the document itself is at issue, then all sides are working from the same text. A difference of opinion may arise as to how to understand the text, but there is no disagreement over what the text says (as contrasted with what it means). That is why we have an adjudicator. But in the scenario described above, the parties do not have any ability even to know what the document “says”. Instead, all they can do is refer to two different texts in translation that each purports to convey the language of the original text into the lingua franca of the parties. This in effect means that there is no original text (indeed, it can be assumed that if additional translations were commissioned, each of them would also differ, one from the other). Not only is there a difference of opinion on whether the document constitutes prior art, but there is also a disagreement over what the document actually is.

There also seems to be a bit of a legal fiction at work here. This Kat is reminded of the well-known query about what do we mean by sound: if a tree rustles in the wind, but no one hears it, has a sound been made? In our situation, from the point of view of the consciousness of the parties involved, the putative prior art document is only embodied in the translation commissioned by Party B, which post-dates the filing date of the application. A translation may be, in modern copyright parlance, a derivative work, but it still stands as a work separate from the underlying work. But in order for the original document to be treated as prior art, we need to deem the translation as being identical in both time and substance with the original document. The tendency to treat two distinct acts as a unified whole reminds this Kat of trade mark licensing law, where the quality control function was developed to support the conclusion that use of the mark by the licensee is deemed use by the licensor. Something similar seems to be work in our situation as well.
When the translation is the prior art: or is it? When the translation is the prior art: or is it? Reviewed by Neil Wilkof on Friday, March 27, 2015 Rating: 5


  1. I've had exactly such a case in opposition before EPO. There were genuinely two possible translations of a swedish word, upon which novelty hinged.

    I was pretty convinced I was right, and implied to the OD that they should find a Swedish speaking colleague to settle the issue.

    I never found out whether they did, but I won.

    This episode had led me to believe that prior art should be considered through the eyes of the skilled person who is a native speaker of the language of the prior art document.

    A more interesting, albeit esoteric, question arises: Is a suboptimal translation not prior art in its own right?

    Is it not fair to assume that, at the time of publication of the prior art, a skilled person not understanding the prior art would have it translated? And that therefore all easy to make translation errors, which nonetheless change the meaning of the disclosure, in all languages, are also part of the prior art?

    A case can be made for that.

    I am waiting for an opportunity to test this argument.

  2. A similar case was decided by the EPO BoA in T 77/87 where the abstract of a tekst was found to be novelty destroying, but the Original tekst was not. In that case the Board decided that the Original tekst prevailed (even if the abstract itself was published before the critical date).

  3. An even trickier question for practitioners is the risk of a translation adding new matter to a patent application when the patent is granted in a different language to that of the filing. This is certainly an issue for many such PCT and European applications.

    For this, the translation does not even need to be erroneous: if the closest translation of a term from language A into language B has an additional meaning or ambiguity in language B that it lacks in language A, this could be enough to introduce new matter.

    As the old Italian adage goes "traduttore, tradittore". One should be extremely mindful of the dangers of translation during patent prosecution.

  4. "if a tree rustles in the wind, but no one hears it, has a sound been made? "

    Yes. Not a difficult question.

    Neither is the translation question. As the first commentator noted, a native speaker would solve the problem. If 2 native speakers disagree then that is the same problem for all documents, whatever their language.

  5. Coming from a land that rarely thinks in more than one language, I am amused (in a pleasant manner) at the question of translations.

    My take is that translations are attempts at a factual rendition, and should be strictly treated - both as to the issue of new matter (no, there is no new matter introduction) and as to the issue of veracity (no, incorrect translations are like a mistake of fact and are a separate animal from the original).

    Interestingly then, the comment about a mistaken translation that was still published before the critical date: both works should be defeating.

    The case in which multiple translations, each of which might be deemed "correct" is a little bit more nuanced. I would posit that since translation rules proceeded the choice of words in the original, that ALL possibly correct translations fall to that original. One's invention then may have two versions in Sweden. But this last one I would leave to the discretion of each individual sovereign for a final determination.

  6. As a legal translator and editor of translated documents, my take is that if the original constitutes prior art, so must the translated version. The wording can differ slightly between any correct versions, but a version that misses a key factor or adds anything not in the original is a wrong one.

    Party A should get a new translator ; )

  7. A prior-published translation will always represent prior art. No argument there.

    How a skilled person interprets a document from the point of view of assessing patentability is where the debate lies and it will be fact specific. Questions that could be asked include:
    Will the skilled person be able to repeat the disclosure? Is the mistake obvious? Will the skilled person rely solely on the disclosure or will they combine the teachings of several? IS the disclosure contradictory to common general knowledge or current thinking? What is the context of the translation? If it is sitting in a patent office file, would the skilled person be aware of it or would they rely on the original and translate that?

  8. Anonymous @ 11:12,

    Some interesting questions to be sure.

    The answers, though, are less interesting and are already present in the definitions (per sovereign) of what meets the legal definitions of prior art (for my sovereign, this would be 35 USC 102 and 35 USC 103).

  9. For the UK (for example and certainly not exclusive) the answers aren't written in such a tidy all-encompassing statute, so we can ask questions and debate. Of those that know how even on bold or italics or a combination thereof.

    The state of the art in the case of an invention
    shall be taken to comprise all matter (whether a
    product, a process, information about either, or
    anything else) which has at any time before the
    priority date of that invention been made available
    to the public (whether in the United Kingdom or
    elsewhere) by written or oral description, by use or
    in any other way.

    This leads to the interesting question of what exactly has been made available, such as by a dodgy translation.

    If it wasn't a weekend I might have trawled through that US statute.

  10. A few thoughts:

    "made available" is a Y/N question on whether any given doc is or is not within the "prior art".

    Once a doc is within the prior art, focus turns to what the notional addressee derives from it. As we know, that notional addressee has some attributes not shared by any real person (knows all the art, reads and understands all languages equally, but lacks any inventive faculty). So exactly what is the totality of technical teaching that the doc makes available is a question not easy to assess in litigation, where the opposing parties urge different technical meanings on the same prior publication. Perhaps the notional skilled person holds BOTH of the competing meanings in mind (and so BOTH meanings have been "made available" to that reader). The notional reader then assesses, on the balance of probability, which meaning is the more likely and that more likely meaning is the one that counts, for obviousness.

    One thing we know though. If the doc is not enabling it is not prejudicial to novelty. That might be enough to filter out bad translations as novelty destroyers.

  11. This in effect means that there is no original text (indeed, it can be assumed that if additional translations were commissioned, each of them would also differ, one from the other). Not only is there a difference of opinion on whether the document constitutes prior art, but there is also a disagreement over what the document actually is.

    It seems to me that there can be no doubt that it is only the original document that counts. That document certainly exists.

    Disagreement on what that original document discloses is certainly possible, but that is not different from disagreement on what a document discloses that was written in the native language of all of the parties (and examiners and/or board members and/or judges) involved.

    Translations are merely a means of providing evidence on what that original document discloses. Just like witness statements on a prior use.

    Of course a translation that is published before the relevant date thereby becomes prior art in its own right, provided it forms an enabling disclosure. (I don't know if this might have been the problem of the abstract in T 77/87 that was mentioned by a commenter.)

    prior art should be considered through the eyes of the skilled person who is a native speaker of the language of the prior art document

    I think so too. The "skilled person" requirement will usually be far more important than the "native speaker" requirement, though. I'm guessing that most prior art documents originally written in English were not written by native speakers. Ambiguities caused by language mistakes can usually be resolved by understanding what is going on at a technical level. Of course there will be exceptions.

  12. Well stated MaxDrei.

    The devil - as always - remains in the details.

  13. Derrick Grover , in a recent article of Philosophy Now , https://philosophynow.org/issues/107/The_Problem_of_Communication, reminds the philosophy community that :" Inventors may have great difficulty in recognising their own inventions by reading the words describing them. In this case, however, it is usual to include diagrams of the invention to aid comprehension. Indeed patent agents usually turn to the diagrams first before reading the words. An electronic circuit diagram would be nearly impossible to understand if it were just described in words". ……, I would suggest that the lost in translation problem that Neil refers to points out rather forcefully the importance of providing very good diagrams and drawings in patent applications, so as to at least partially compensate for translation discrepancies and poverty of language.


  14. Joe,

    the article itself is behind a paywall, which blocks a read and my comment may or not be answered...

    While a picture indeed may be worth a thousand words, a picture itself is limited to a snapshot in time, and only then to what can be captured visually. The mind's eye, as it were, is more than visual.

  15. Try telling the EPO that a picture may be worth a thousand words. They just quote Article 123(2) EPC in response.

  16. Not quite.

    T 2537/10, point 2.9 of the reasons.

    Please be careful when generalising some personal experience to statements about "the EPO" as a whole.

  17. What about machine translations? We often find that a ‘real’ translation is different from the machine translation provided by the EPO. Sometimes the machine translation can be technically plausible (and thus arguably enabled), but nevertheless not an accurate representation of the prior art document itself. Once that (incorrect) machine translation has been made available to the public should it be counted as prior art?

    What about ‘potential’ machine translations? For example, if a Japanese patent exists and the machine translation software exists and all one needs to do is click a button on the JPO browser to obtain a machine translation, then is that potential machine translation already available to the public and therefore prior art?


  18. Once that (incorrect) machine translation has been made available to the public should it be counted as prior art?

    In my view, yes.

    if a Japanese patent exists and the machine translation software exists and all one needs to do is click a button on the JPO browser to obtain a machine translation, then is that potential machine translation already available to the public and therefore prior art?

    In my view, yes. The Japanese-language version of the JPO website makes the Japanese text available to the public. The English-language version of the JPO website makes the English machine translation available to the public in exactly the same way. What exactly goes on behind the scene does not seem to be relevant for Art. 54(2).

    Since the machine translation software of the JPO most likely is updated regularly, it will not be easy to prove that a particular mistranslation was made available on or before a particular day in the past. But that is a different issue.

    If you have to use Google translate to get to the mistranslation you need, then I have my doubts. Using Google translate involves a choice thay you make now, and not one that the JPO made then. (On the other hand, if Google translate was at the relevant date the standard tool used by the skilled person, then who knows...)

  19. I was concerned at the changeable nature of JPO machine translations some years ago, having found by experiment that translations downloaded a few months apart could have significant differences. To ensure that we were both working from the same text, with the examiner's agreement, I sent him a copy of the machine translation of a Japanese citation that I was working from so there would be no misunderstanding when quoting passages of the translation. In that particular case I had no issues with the accuracy of the translation. The examiner appeared unaware that JPO machine translations are generated on an ad hoc basis and that there is no archive of translations for downloading on request. What is to happen when it is not possible to make a translation that accurately conveys the sense of the original language? An example of this is found in the EPO Guidelines part F, IV 13, 4.6, relating to the allowability of relative terms. The English text reads:

    It is preferable not to use a relative or similar term such as "thin", "wide" or "strong" in a claim unless the term has a well-recognised meaning in the particular art, e.g. "high-frequency" in relation to an amplifier, and this is the meaning intended.

    Speaking with my electronics engineer hat on, this makes no sense, as, in English "High Frequency" does not have a well-recognised meaning in the amplifier art.

    This passage almost certainly derives from the German text,

    Es ist nicht empfehlenswert, in einem Patentanspruch relative Begriffe wie "dünn", "weit" oder "stark" bzw. ähnliche Ausdrücke zu verwenden, es sei denn, der betreffende Ausdruck hat auf dem betreffenden Fachgebiet eine allgemein anerkannte Bedeutung, z. B. "Hochfrequenz" in Bezug auf einen Verstärker, und dies ist die beabsichtigte Bedeutung.

    Hochfrequenzverstärker does have a well-understood meaning on German, but its correct English translation is radio frequency amplifier, which is not a relative term. The only way to preserve the sense of the German original would be by using a different English relative expression altogether, such as "High Speed Steel" (= Schnellstahl, literally, "fast steel" in German).

  20. Ron,
    I've noticed that (relatively) recently the JPO machine translation has come with two options for some apparently ambiguous words, one of which is usually more context appropriate than the other. Helps comprehension (once your mind gets used to concentrating on the preferred term and ignoring the other) but could be a minefield for getting a correct and acceptable translation e.g. as per your high frequency example.

  21. I find the discussion here fascinating, but am deeply troubled by those that would seem to hold that something that has not been run can be deemed to be "prior art."

    If, as people are commenting, the software that runs the translations is being perpetually updated, then I cannot give credit to any translation not yet run. If simply awaiting "the push of the button" is also accompanied with the very real possibility of intervening changes (and those changes result in different translations) then such a result cannot be "prior" in any sense of the word. Even the interactive contextual choices indicate that the translation only comes into being at the time of that particular choice.

    I liken this to a simple physical analogy: I have a set of resistors (words). I can connect these resistors in parallel or in series. My instant choice of how I connect them creates vastly different results, even as the elements I have do not change.

    Do I have all possible results? It is arguable that I do. Should I have credit for all possible results? I know of no one that would say that I should.

    In another analogy, let's say I have a big box, full of electrons, protons, and neutrons (letters). Does my big box "preempt" all possible configurations? Of course, as all possible configurations can be reduced and readily recognized as being composed of electrons, protons, and neutrons. Should I receive credit for all possible configurations? Of course not. So too then, with this "big box" of letters and the push of a button that makes a translation (with intervening re-configurable translation rules).

  22. The "High frequency amplifier" example of permissible relative expression found its way into the UKPO's "Manual Of Patent Practice" (MPP) when the 1977 Patents Act was introduced. I was an examiner then, working in an examination group whose subject matter included amplifiers. We did complain to those responsible for the MPP that we always objected to "high frequency" in electronics as it had no well-understood meaning, and were concerned that the MPP approved the use of an expression that would be objected to during examination. Our complaints fell on deaf ears.

    I understand from a conversation with a professional translator some years ago that his code of practice required him to always use his best estimate of a word where there were multiple meanings. This may account for non-linguists having the perception that exact translations are always possible. Not being a professional translator, I have had no qualms about providing possible alternatives in brackets when I had occasion to provide the occasional urgent ad hoc translation, or using footnotes to explain nuances in the original language which would not necessarily be apparent from a literal rendition into English.

    A number of articles about the difficulties that can arise with translations have appeared in the pages of CIPA over the years: the problem is hardly new, but the powers that be do still seem to believe that computerised Rontranslation will solve everything.

  23. @US anon:

    The test in Europe is whether the information has been made available to the public. It is not necessary that any member of the public has actually accessed it. It is also not necessary that the information continues to be made available to the public, once is enough. So the scrolls in the library of Alexandria are prior art under Art. 54(2) EPC. That we are not able to prove much about what those scrolls disclosed is a different issue.

    So, if a patent office announces that today certain patents are published that can be downloaded by typing their patent number into a web page and hitting a button, those patents become prior art from today by that announcement even if nobody takes the trouble to actually type in those numbers. If there is an additional button to obtain the English machine-translation, then the presence of that button in my view also makes the English machine-translation available to the public even if nobody actually clicks the button.

    Does it matter whether the English machine-translation is actually generated? I don't think it does. Is there much difference between an electrical signal stored on a hard disk that, in combination with certain software present on the web server, encodes the Japanese original, or that same electrical signal stored on that same hard disk that, in combination with the machine-translation software present on that same web server, encodes the English machine-translation?

    Having said all of that, I think Ron's "Hochfrequenz"-example is very interesting. It seems reasonable that the skilled person reading a translation will take into account that the text he is reading is a translation and all the more so if the text is a machine-translation. If there is anything suspicious about it, the skilled person will likely turn to the original (whose language he can read anyway) or, if the original is not available (*), conclude that nothing can be deduced from the suspicious part of the machine-translation.

    So I think that it will be very rare for a machine-translation to disclose something that the original did not.

    (*) Is it possible in the world of the hypothetical skilled person that the original is not available when the skilled person is reading the machine-translation (assuming that the original was published at some point in time somewhere in the world)? Somewhat related question: if two documents reference each other in a way that would make the skilled person read the two documents together as a single disclosure, is it necessary that the skilled person when reading one of the two documents somehow has "practical" access to the other? Suppose both documents were only made available to the public on separate days behind a button that nobody actually clicked.

    Art. 54(2) should get a warning label: known to cause dizziness.

  24. @US anon:
    I liken this to a simple physical analogy: I have a set of resistors (words). I can connect these resistors in parallel or in series. My instant choice of how I connect them creates vastly different results, even as the elements I have do not change.

    Do I have all possible results? It is arguable that I do. Should I have credit for all possible results? I know of no one that would say that I should.

    The difference with the "one-click-translation" is that you still need to make many choices to arrive at a particular circuit. The button on the JPO website is something that in a way comes with the patent number. Selecting that button does not create novelty, but making many choices while arranging resistors does. I would say selecting (today) your own translator software instead of using the software provided by the JPO also creates novelty.

  25. Anonymous @ 21:56,

    Except not.

    Please read again the conditions present: that one click is NOT a static condition fixed in time to when the original item itself is made available. There are intervening changes to the translation software, and there are interactive options available AT THE TIME of the translation.

    If the conditions were as you suppose, then my stand may be different.

    They are not.
    My stand is not.

  26. @anonymous 28.03 08:32:
    I do not understand your remark regarding granting in a different language than filing. How can that be the case? Do you mean filing in a different language than the priority document? Surely, that only may pose a problem in case of priority questions. Any translation of the original text filed in e.g. a different country will form the basis of the procedure, regardless of translation errors or not. The translation is then the original document, because that is the document filed.

  27. I have certainly come across human-written abstracts that disclose matter not present in the original. About 20 years ago I received an EPO search report on an invention for an optical beam splitter where a groove was made in a substrate into which a thin piece of a dichroic material could be inserted. One of the citations was a German patent. Its English "Derwent" abstract referred to a "groove". This was a translation of "Riss", and it was abundantly clear from the embodiment and its mode of operation that the structure in question was in fact a "crack" of essentially zero width, made by cracking its glassy substrate, the crack itself providing the beam-splitting behaviour.

    The abridgements of Pre-1949 Act UK patents sometimes disclose matter that is not found in the printed patent. The patents in question were those filed under international convention, the files of which were laid open to public inspection 18 months after the filing date. The abridgements of such patents were based on the state of the file as laid open, and sometimes include drawings and description of embodiments that were deleted subsequently. Such abridgements do explicitly say something like "This matter does not appear in the application as accepted". The files themselves are long gone.

    I appreciate that this is not directly relevant to translations, but felt it worth a mention, as an information scientist I used to work with came across such an abridgement recently. She was disappointed to find that the actual patent did not disclose what the abridgement had promised to be highly relevant prior art. The abridgement per se did not provide enough detail to be useable.

    Unless family members of the original foreign application can be traced, the abridgement will now be the only extant record of the original disclosure. Not that this always works: as a trainee patent examiner I once found a UK patent, based on a French application, whose UK filing date was too late for it to be citable. I sent for a copy of the corresponding French patent, only to find that the embodiment I was relying on had been added when the UK application had been filed. A salutary lesson that it is not safe to rely on patent family members as an alternative to having the actual cited document translated.

  28. Anonymous at 21:33,

    My apologies for overlooking your comment.

    That being said, I think that you have taken a tangent to my discussion, and you too are missing the point of the possibility of intervening changes post-original item brought about by the software changes in the translation mechanisms.

    It is a bit of a non sequitur to that discussion whether or not the public has accessed that which can be accessed (in a truly unchanged state of being). The unpushed button simply is not something actually in being. The ability to access by pushing a button is quite a different thing than the mechanism of employing the translation mechanism.

    You too, need to pay attention to ALL of what I posted.


  29. I have a view that seems to differ from almost all of the commenters so far.

    Firstly, as some commenters have already alluded to, there is no such thing as a “correct” translation. The vocabulary and grammar of different languages in general always create different nuances and meanings from each other. A clear example is that a language such as Japanese which has no singular and plural will usually convey a wider range of meanings than a language such as English which has clear singular and plural.

    But on the main issue, I do not believe that a translation can ever disclose (for prior art purposes) something that was not present in the original text. Even if the translation is prior-published, I consider that the notional skilled person must be taken to be aware of its status as a translation, and to disregard any teaching from the translation that is not present in the original. I have used this successfully at various patent offices in relation to post-published translations; but even in relation to prior-published translations I believe that it holds good and I would be somewhat optimistic about the prospects of success. Thus, if the English text used a singular word, but a Japanese translation appeared to also teach a plurality, I do not believe that the teaching of the plurality is genuinely present in the prior art.

    Some support for this view can be found in T 77/87, referred to by Millipede above, where a (prior published) abstract that was not consistent with the original was held not to be novelty destroying even though on the face of it, it was. As discussed in the CaseLaw of the EPO book:

    In T 77/87 (OJ 1990, 280) the abstract published in the journal "Chemical Abstracts" did not correctly reproduce the original paper. The board stated that the original document was the primary source of what had been made available as a technical teaching. Where there was a substantial inconsistency between the original document and its abstract, it was clearly the disclosure of the original document that had to prevail. The disclosure in the original document provided the strongest evidence as to what had been made available to the skilled person. When it was clear from related, contemporaneously available evidence that the literal disclosure of a document was erroneous and did not represent the intended technical reality, such an erroneous disclosure should not be considered part of the state of the art.

    Best wishes



All comments must be moderated by a member of the IPKat team before they appear on the blog. Comments will not be allowed if the contravene the IPKat policy that readers' comments should not be obscene or defamatory; they should not consist of ad hominem attacks on members of the blog team or other comment-posters and they should make a constructive contribution to the discussion of the post on which they purport to comment.

It is also the IPKat policy that comments should not be made completely anonymously, and users should use a consistent name or pseudonym (which should not itself be defamatory or obscene, or that of another real person), either in the "identity" field, or at the beginning of the comment. Current practice is to, however, allow a limited number of comments that contravene this policy, provided that the comment has a high degree of relevance and the comment chain does not become too difficult to follow.

Learn more here: http://ipkitten.blogspot.com/p/want-to-complain.html

Powered by Blogger.