The three dimensions of patent translation

1. Accuracy

Accuracy is a measure of the extent to which a translation captures the meaning of the original. For a translation to be entirely accurate, the meaning of the translation must be identical to the meaning of the original, that is, if no meaning has been added to and no meaning taken away from the original. Perhaps this can be best appreciated in the case of a patent claim – if the scope of the claim in the Japanese original is not identical to the scope of the claim in its English translation, then the translation is not accurate.

Now ordinarily, we want our translations to be accurate, and this is pretty much non-negotiable in the case of patent translation. If there is a choice between a more accurate and a less accurate rendering, we will usually opt for the more accurate one. Of course there are exceptions to this; perhaps there is a mistake in the original that we want to fix or we want to rewrite the claims in order to obtain a broader claim. – But this goes beyond translation and enters the realm of editing and rewriting. Whether it’s OK to do so will depend on the kind of translation we are preparing (PCT? Paris priority? For filing? For reference?). Let’s keep it simple here and assume that the original is flawless (as unrealistic an assumption that might be) and that we do not want to “improve” it. In that case, there is a simple test for the accuracy of our claim translation: Can we imagine subject matter (e.g. a device or a method) that falls into the scope of the original Japanese but not into the scope of the English translation? If yes, then our translation of that claim is not entirely accurate. A similar test can also be applied to the rest of the specification: If it means something different, it’s not accurate.

Perhaps one remark regarding the term “meaning”: By this, I mean the objective meaning, i.e. the meaning that a person skilled in the art would have ascribed to the original, as opposed to the meaning that the author of the original had in mind. This may be relevant when there is a difference between the two.

What are the consequences if a translation is not accurate? Well, that depends entirely on the details of a given case – in most cases, an inaccuracy will have no consequences at all. In fact, it is probably fair to say that noone has ever created a translation of an entire patent specification that is 100% accurate; that seems pretty much impossible as even the best matches in technical terms often have subtle differences in meaning that might mean that it is impossible to arrive at exactly the same meaning and scope in the target language. – But in some cases, there may be consequences and those may be fatal. For example, a different claim scope may mean that the resulting claim is not novel/inventive or that the resulting patent does not cover a competitor’s embodiment anymore. I have talked about this problem in detail at an IJET conference in Osaka in 2018 and will summarize this in a different post.

2. Readability

By readability, I mean the ease with which a text can be processed by the reader. This is perhaps what is most difficult to some non-native translators (i.e. translators translating into their second or third language). While they may be able to translate with a high degree of accuracy, their translation may be immediately recognizable as that by a non-native speaker. Odd word and grammar choices may lead to a translation that is perfectly understandable (and thus accurate) and yet irritating or jarring to the ear. Needless to say, such problems are not limited to non-native translators and even some native English-speaking translators struggle with writing English that is as readable as the original.

I should perhaps stress that I am not talking about clarity (in a patent sense) here. For example, if the original is clear and the translation is not, e.g. because it contains an unresolvable ambiguity, then this may be an indication that the meaning of the original has been altered, which may mean that the translation is inaccurate. And if the original is unclear and the translation is similarly unclear, then this may not necessarily be a problem with readability as I use it here, but rather a problem with the original.

We can increase readability by avoiding grammatical ambiguities and by using certain “guideposts” for the reader, such as parallel constructions, careful use of relative pronouns and certain, seemlingly inconsequential words like “also”, “however”, “nevertheless” and many others. that act as the glue that ties the translation together.

Let’s have a look at some examples:

薄膜１０は、蒸着法により形成された薄膜である。

a) The thin-film 10 is a thin-film that is formed by a vapor deposition method.

b) The thin-film 10 is formed by vapor deposition.

Translation a) contains 14 words whereas translation b) gets the same meaning across with only 8 words. And while translation a) is perfectly grammatical English, I venture to say that translation b) is more natural in the sense that it is closer to what a native English speaker might have written when trying to express the same idea from scratch.

Of course, one might argue that the original contains the expression 薄膜 twice, so shouldn’t it also be present twice in the translation? The answer is clearly no, as this has no bearing at all on the accuracy of the translation. Both translations a) and b) clearly mean exactly the same thing. The fact that one is superficially more similar to the original does not make it more accurate. We will get into this issue in some more detail in the next section on checkability.

Similarly, the expression “vapor depostion method” might be superficially closer to the original 蒸着法, but that does not mean that it is better. In fact, a reader might wonder what the word “method” is doing there, when it is clearly not necessary. Is there a difference between “a vapor deposition method” and simple “vapor deposition”? “a vapor deposition method” implies that there are many different such methods, whereas “vapor deposition” focuses on the commonalities of all these methods. Also, the overall structure of the sentence is a somewhat inelegant “The A is an A that…” All these are admittedly minor issues, but they will slow the reader down, and more so than just by using 14 instead of 8 words.

So what are the consequences if the readability of our translation is poor? Compared to problems with accuracy, which may be fatal, the consequences of lower readability are typically not fatal, and therefore less serious. Most importantly, the Examiner as well as the overseas prosecuting patent attorney will take longer to read and understand the specification. In the case of the (e.g. US or EP) patent attorney, this will lead to higher prosecution costs, as they will bill more for poorly readable English that takes longer to read and understand. There is also the effect that poorly readable English is more prone to mistranslations if taken as a basis for translation into other languages (e.g. German or French).

Now, although the most serious problems are those where the translation is inaccurate, the most common problems are problems of readability. In fact, when I correct patent translations done by other translators, my corrections relating to readability issues typically outnumber those relating to accuracy issues by a factor of at least 10:1.

Let’s look at the last dimension:

3. Checkability

Presumably, a translation is easiest to check if it does not require any corrections at all – the job of the checker is done and they can go home. But this is not what I mean by checkability. One part of a translation checker’s job is to confirm how the translators got from the original text to the translated text. Presumably this is easiest if there is the greatest similarity in terms of structure of the text, i.e. if verbs remain verbs, nouns remain nouns, all pieces of the original can be accounted for in the translation and vice versa. In other words, what I mean by checkability here is the “conservation of lexemes”, and indeed, I might have called it that, if checkability weren’t more succinct.

There is a school of translation out there that states more or less that the conservation of lexemes is a goal that is inherently worth striving for – this school is known as “lexeme-based translation”. In essence, this approach to translation says that, whenever possible, there should be a perfect correspondence between the lexemes in the original and those in the translation. I look forward to discussing this notion in more detail in one of my future posts. Suffice it for now to say that checkability (conservation of lexemes) is the least important of the three dimensions. Rather, the order of importance of the three dimensions is this: accuracy – readability – checkability. In other words, no compromises should be made with regard to accuracy. Some compromises may be made with regard to readability if that leads to a more accurate translation. But neither accuracy nor readability should be compromised for the sake of checkability.

Let’s recall for a moment the translation example above. Clearly, translation a) is more lexeme-conservative than translation b). Pretty much all lexemes are accounted for. And yet, as we have seen, it is not the better translation.

Or to take another fairly common case: Many Japanese specifications contain passages, often several pages long, in which every single sentence or paragraph will start with また、…　. Here we have the choice to translate this また (e.g. as “Furthermore” or “Moreover”) or to ignore it in all cases where adding a corresponding word in English does not aid in increasing the readability of the translation. If leaving out the corresponding “Furthermore” or “Moreover” will not change the meaning of the sentence (as in most cases), then there is simply no point in recreating it with the same frequency as in Japanese. In fact, doing so will likely be distracting and therefore diminish readability.

So as we have seen, there are often tradeoffs between the different dimensions. In fact, some of the most frequent questions in patent translation can be clearly answered if looked at as described above. Is it OK to split up sentences in a PCT translation? The answer is clearly yes, if doing so does not reduce the translation’s accuracy. Is it necessary to translate every など in the original? Only if leaving it out changes the meaning. And so on.

Here’s another example:

絶縁性を有する材料

a) a material having an insulating property

b) a material that is insulating

Clearly, translation a) scores higher on the checkability scale, but translation b) is not only more succinct, but also more natural.

Here’s an example of a tradeoff between checkability and accuracy:

各センサ３は、各信号線４に接続されている。

a) Each sensor 3 is connected to each signal line 4.

b) The sensors 3 are respectively connected to the signal lines 4.

c) Each of the sensors 3 is connected to one of the signal lines 4.

Taken literally, translation a) does not mean what is shown in the drawing. Rather, it corresponds to the following situation:

Thus, it’s confusing at best and arguably completely inaccurate. Translations b) and c) are vastly superior, although translation a) scores higher on checkability.

Now, many translators ask themselves how far they can veer from the original. This little demon that they may hear in their ear arguing for greater superficial similarity to the original is the checkability demon. It is sometimes hard to ignore, in particular for inexperienced translators. But there are many cases, where it should be ignored and I will give many more examples of such cases as this series progresses.

Now, you may ask, what if you have specific instructions from the client to translate something a certain way that runs counter to the above? In that case, by all means, stick to the client’s instructions. But do not readily assume that the client will not understand what you are doing just because that’s what happened with other clients. Be ready to stand your ground and don’t compromise on quality. That’s what makes you a professional.