[Patents] Eurolinux position on Community Patent

[phm@a2e.de (PILCH Hartmut)]

> >    While the burden of having to provide translation for 22 languages may
> >    be formidable, requiring all Europeans to be fluent in English or any
> >    specific set of official languages just transfers the same burden from
> >    patentees to citizens and runs counter to the EU's professed
> >    committment to cultural diversity.

> I completely agree (well, except that I have never counted the number
> of languages in the EU, I did't know they were 22, but it will change
> with time, and maybe the exact number is not very important).

When people argue that translation requirements need to be waived,
they often mention this number, which counts in a group of languages from
eastern and southern european countries that are not yet in the EC.

> >    Machine translations could be accepted, if they reach a certain level
> >    of reliability. By using intermediate languages such as UNL or Lojban
> >    (Logical Language), a patent applicant can create an unambiguous
> >    semantic representation that can be reliably machine-translated into
> >    any number of languages. Such a semantic representation could moreover
> >    prove helpful in disambiguating claims and descriptons during patent
> >    litigation.

> If that is technically feasible, it would be a great advance if it was
> possible, both for law and NLP.

There are only prototypes so far.

It is clear that the attainable level of translation is significantly
better than that of babelfish or any program that translates between
normal languages.  Especially in patent descriptions I would estimate that
most of the automatic translation mistakes are due to syntactic ambiguity
of long sentences.  This does not exist in a language like Lojban.  If
encountered, the parser would report an error and it would not be valid
Lojban.  Also, UNL is being used for productive work already according to
a similar rationale.  Of course, even texts generated from such
intermediate language will not attain the level of a good human
translator, but it can be expected that they will be understood
correctly by a careful reader.

> >    Another approach would be to stipulate that a Community Patent is
> >    potentially valid in all countries but can be enforced in each country
> >    only after a translation in that country's language has been
> >    published. Also, the claims are more important than the description.

> There's something I don't quite understand with this idea: - what does
> potentially valid mean?. Would it be similar to the priority date,
> that now allows patentees (for a while) to patent something in one
> country when it is no longer novel because they patented it in another
> country when it was novel?

Yes.  This is of course just an example of a possible compromise.  The
purpose was to show that there are many parameters to fiddle with.

> So, they would get their patent in those countries where they have a
> translation of the patent and simply a sort of unlimited "license" to
> patent in those others when they translate it to those languages?.
> That does not seem much better for the rest of us than granting
> patents in (say) English, because unless we learn English we wouldn't
> be able to know that somebody is waiting for us to compete in their
> market to translate the patent and sue us.

We would be warned by the imperfect machine translation provided at the
beginning (according to the exemplary combined proposal).

> - How does it relate to the automatic granting and patent insurance
> company market previously proposed?. If patents are granted without
> warranty of novelness, etc. and those warranties are left to a betting
> game, how does language enter in that game?.

It would be a problem if people could publish let's say in Hungarian only,
with only an imperfect Logical Language version for the rest of the
people.  Indeed the bounty (betting sum) could be somehow a function of
the number of readers reached: the more/better translations the lower the
bounty.  This would reflect the public costs.

> - How fair is it to grant any privilege over the freedom of people to
> somebody in exchange of disclosure of anything if that disclosure is
> not in a language those people understand, and therefore does not
> really teach them much?. Of course it allows people to translate it,
> to incorporate it into textbooks that are then translated, etc. and
> this could allow those that don't speak the patent language to learn
> it indirectly. So it isn't a complete disclosure but also it is better
> than nothing. It makes sense that the privilege granted over the
> freedom of those who don't speak that language is weaker than the
> privilege over freedoms of those who do.

You answered it very well already.

> - How do we define countries?. The mapping between language and
> territory is not as simple as one state one language.

Currently there is already a set of languages whose translation is
required by national patent offices.  I guess some people might like
Catalan, Basque, Sardinian, Welsh, Rhaeto-Romanic, Sorbic etc inserted,
and disputes about what is a nation could ensue.  We didn't want to get
into such disputes, and currently they are groundless.  However, the
method of weighting the number of native speakers reached against the
invalidation bounty sum as suggested above could provide a solution that
would accomodate any language, however small, even beyond the possibility
of reliably understandable machine-translations from LL.

-- 
Hartmut Pilch                                         http://phm.ffii.org/
Protecting Innovation against Patent Inflation      http://swpat.ffii.org/
95000 signatures against software patents       http://www.noepatents.org/