[Gauss-parl] adobe equivalents

[Carsten Svaneborg]

On Saturday 02 April 2005 13:44, Roland Orre wrote:
> I would like to see all patents downloaded,

Everything is under ~zqex/PatDat/pat

the subdirectory structure shows which searches I have made.
The most recent ones are topic/word/* and last week
topic/company/adobe

However, since each patent only occurs once in directory
hierachy, the adobe subdirectory only contains those patents
that were not already downloaded somewhere else in the three
previously.

If you want those, then I suggest you write a search script
that takes a patent list (as topic/company/adobe/_adobe) and
transforms that into a list of filenames.


> the issue of classifying which are software or not is something
> we need both types for.

True. However, I think the main problem for the moment is to
remove patents from the DB that are not swpat, and I suspect
most of these will be in the crossover between hardware/software,
which is why I don't think that non-swpat such as chemicals,
textiles or milking-machines are really relevant for the classification.

I think the outliers are easy to find and delete manually by looking
at http://gauss.ffii.org/Browse/Ecla/


> Further on, if we see the gauss as a service as well, not only
> a political instrument, then it is a clear advantage to be able
> to search for any patent, and get its classification as
> software vs real.

I don't really care about non-swpat patents, and I think swpat is
more than enough to handle. So my only interest is to get rid of them
(both swpat and non-swpat ;*)

What is relevant is to create the best possible statistics about
swpat, and document which software patents are being applied
for, and granted.  This means defining some fuzzy threshold,
and deleting whatever remains.


The relevance of the classification is that the EPO makes weekly
publications of new applications with abstract, it would be extremely
handly to have a script that sorts these after swpat probably based
on abstract (title,ecla,applicant) only, and then automatically adding
highlikelyhood swpat applications to the DB.

I have a backlog of these files for quite a number of weeks, but so
far only a pretty printing script.

N.b. On my timescale implementing this is probably somewhere
between half year and infinity into the future, so if somebody are
interested in doing it, please give a shout.

-- 
  Mvh. Carsten