On 01/22/2013 12:48 AM, Benson Margulies wrote:
On Mon, Jan 21, 2013 at 6:05 PM, Masanz, James J.<masanz.ja...@mayo.edu> wrote:
>
>Many of the models are derived from data that is not publicly released. But
those*models* have been contributed to Apache cTAKES.
>
>This is the primary mechanism for distribution of these models.
This discussion has come up before with SpamAssassin and OpenNLP.
The exception for SpamAssassin, as I have understood it, is that there
is a place where all members of the community have access to the
source texts behind the models. At OpenNLP, again according to my
memory, the solution is to treat the models as not part of the release
at all.
Yes, thats right, OpenNLP does not release any models at Apache. There is
a page over at SourceForge where people can download them and to help
the people who need to train them self we release the corpora specific
training
code as part of the OpenNLP release.
I fear that cTakes needs to have an interaction with LEGAL to adopt
the SpamAssassin model, since, from a strict constructionist
perspective, the source of the models is precisely what you cannot
release.
As far as I understand it a part of the data is confidential and can't
be shared
with the community at all, is that right?
Jörn
