Saturday, 31 January 2009

Dyson again: prior art lost in translation

The Dyson decision (noted by PatLit here) is particularly interesting from the viewpoint of an ex-patent translation proofreader, writes Luke Ueda-Sarson, who says:
"I'm convinced that translation errors are rife in patent litigation, but they just happen to go undetected. A recent American paper on the topic indicated that translation errors may be important, but the authors basically couldn't find any actual examples from court cases. Dyson looks like it is such a case. True, Samsung only came away with claim 9 intact, but there is every chance they shouldn't have got that either. Perhaps more interestingly, Dyson are really leaving themselves to further attack open with that Sanyo reference, since it's not as poor as it was made out to be.

I'm currently discussing this case on the patent translation e-mailing list on yahoo groups, and trying to get a variety of views from experienced translators on their take on the critical passage. Hopefully I can synthesize the various threads at some point.

Anyway, the critical 1977 Sanyo reference was "in the agreed translation" (sic, whatever that means) held to disclose:

"Even if the air flows within the intermediate cover 11 through the turn guide pipe 17, a predetermined amount of dust still remaining in the air is caught by means of the fine filter 18."

The reference was essentially discounted because the skilled reader wouldn't know what to make of the words "fine" and "predetermined"; they render the passage non-sensical, and the general sense is incapable of clear interpretation. Hence being unavailable for an obviousness attack.

But neither of the critical words appear in the Japanese.

A first-stab translation is as follows:

"The air travels through the inner lid 11 via the swirl-guidance barrel 17, and due to the filter 18 disposed part-way along, some of the dust that is still contained in the air is captured."

This of course is entirely different from what was proferred in court. There is no conditional "if".  The filter is just a filter, not a "fine" filter, so there is no reason to ask "why doesn't it clog?". Perhaps more importantly, there is no later contradiction that was held to exist with it requiring to be coarse. In fact, on page 3 it describes filter 18 specifically as being a *coarse* filter.  And "predetermined" is simply a mistake for "some"; "some" making perfect sense in his invention, unlike "predetermined".

Thus the facts that the judge took into account are completely wrong - not one of the problems which he cited with the translation exists in the original.

As an aside, I must say the judge has to my mind a highly improper way of referring to the translation as if it is the original document. Does not the person skilled in the art know all languages, needing no translation service? Judges should be alive to the fact that any translation may contain errors. If a translated document looks to have some weird language usage, surely (surely!), the first thought should be to ask "is this translation up to scratch?" and not say, as happened in this case, that the skilled reader could reasonably "conclude that [the people who actually invented the technology] had not fully understood the technology"!

Bad facts lead to bad law, 'tis said..."


Anonymous said...

The frequency of errors in translation is something I have looked into, statistically, and found, for example, that in WIPO abstracts of PCTs filed in Japanese, 10% of the 'English' abstracts contained material errors and 48% contained 'poor expression'. Now abstracts are not claims, but if after translation by freelance translators, checking by agencies and editing by staff translators at WIPO this level of errors still occurred, what chance has a law firm who outsourced translation to the cheapest bidder (and worst translator) ?
I have the data above in a Power Point (MS) presentation if anyone wants to see it. Interestingly, though while I do not translatate Chinese or Korean, the CN and KR English abstracts on PCT contaned many of the same common errors but the expression in English was, if anything worse,sometimes bordering on unintelligible.

Anonymous said...

Great comment from Luke. Makes one wonder how many other times an infelicitous translation has affected the outcome. German patent attorneys used to bolster their meagre remuneration, early in their careers, by translating the allowed EPO application into German. They soon realise ghow impossible it is to render in the translation exactly the meaning of the original English. Most patent attorneys (and litigators) with English as mother tongue go through their entire careers blithely unaware of this problem (and its inverse). Mind you, they do know that American English and UK English are different, the first time they see the brass plate NO SOLICITORS on the front door of some swanky US law firm offices. Luke, the "agreed" translation is what the opposed English law firms "agreed" as accurate, isn't it?

Anonymous said...

It hardly seems fair to blame the Judge. The UK system is an adversarial one. It is up to a party relying on a document to put forward the most advantageous translation.
This doesn't of course detract from the main point, which is - if as a party to litigation there are translations involved, and especially if large amounts are at stake, for goodness' sake don't take the first translation that is proferred at face value. And the point works both ways...

Anonymous said...

Max, not being an attorney, let alone a litigation attorney, I have no real idea what "agreed" means here.

(So I will be unlikely to visiting here much; I have just been pointed this way by our host)

If a translation was "agreed" to by the opposing parties, then, well, I would be a surprised. Presumably such a translation must have been commissioned by one of the two (as opposed to a neutral 3rd party, like the court...); but it that case, why would you, as a litigant, put your trust in the other party for something so critical? It can only be because it was not regraded as being important, let alone critical.

Regarding anonymous' point; I can't (entirely) agree. A judge can only use what is in front of them, it is true - but judges shouldn't close their eyes just because they believe there is nothing to see. Here the judge's own language clearly demonstrates he is regarding the translation, improperly, as if it is the original document; indeed, the fact it is a translated document seems to have been forgotten, otherwise I don't see how he could have said what he did in terms of its author's [sic] knowledge.

If a trial judge in a criminal case, where the defendant was, e.g. a Swahili speaker, made the crucial part of the judgment based on exact words used by the defendant, ( the equivalent, say, of 'Let him have it') despite the language actually quoted being that of the interpreter, and at no point mentioned the fact that the words quoted are in fact those of the interpreter, and not the defendant, wouldn't that be a clear case for potential review?

Cheers, Luke

Anonymous said...

Thanks Luke. Here's a further thought for reraders. Suppose there's fifty times more patent trials in Germany than in England. Now think how many of the EP-B publications, the basis of the litigation, are in English. Most of them, right? Does the trial proceed in English? It does not. So, where does the German translation of the EP-B come from? Is it in some sense "agreed"? How?