Re: license of parser data files?
Anon Sricharoenchai
On 7/8/08, Christopher Manning <man...@stanford.edu> wrote:
> Hi Anon,
>
> The only one of these questions that has a clear answer is 2: Yes, these
> grammars are constructed from the Penn Treebank.
>
> For the rest, I'm not a lawyer, and we haven't done formal investigation of
> this, but I suspect the answer is a gray area. In practice, a number of
> people/groups have for many years distributed grammars ultimately derived
> from the Penn Treebank and LDC has never complained, but nor has it been
> officially sanctioned (as far as I am aware).
>
Does those in practice also include the commercial distributions?
> It's not at all clear to me that UPenn holds copyright on the grammars.
Did you also mean that no-one could hold copyright on the grammars in
the treebank corpus?
> Actually, it's not clear to me that a court would say a grammar file can be
> copyrighted.
The grammars may not be copyrighted,
but the grammars accompanied with Stanford Parser, contain some
statistics information.
LDC also reserve the right on those statistics information as follow,
https://secure.ldc.upenn.edu/access/userAgreement.jsp
``Small excerpts of text or audio data from LDC-Online materials may
be displayed to others or published in a scientific or technical
context, solely for the purpose of describing the research and related
issues. Statistics and other summaries of LDC-Online materials may
also be published in the same context. Except for such publication of
small excerpts or statistical summaries in scientific or technical
works, neither LDC-Online materials themselves, nor access to them,
may be sold or transferred to others.''
The above is just the agreement between anonymous users and LDC.
I'm not sure what is exactly the license agreement between,
1. Stanford and LDC (whether anyone can use the LDC's statistics data,
that is distributed by Stanford, for any purposes or not),
2. LDC and UPenn,
3. UPenn and the public (whether UPenn allows anyone to use those data
for any purposes or not).
> It doesn't seem to meet the usual grounds for being a creative
> work.
>
It depends on whether treebank corpus is a creative work or not.
If treebank is a creative work, then its statistics data may possibly
inherit the copyright status.
http://www.gnu.org/licenses/gpl-faq.html#GPLOutput
``More generally, when a program translates its input into some other
form, the copyright status of the output inherits that of the input it
was generated from.''
> Best,
>
> Chris.
>
>
>
>
> On Jul 8, 2008, at 6:21 AM, Anon Sricharoenchai wrote:
>
>
> >
> > Hi,
> >
> > According to the stanford parser downloaded from
> >
> http://nlp.stanford.edu/downloads/lex-parser.shtml#Download,
> it
> > includes the parser or data files such as, wsjPCFG.ser.gz,
> > englishPCFG.ser.gz.
> > My questions are,
> > 1. What is the license of these data files?
> > 2. Are these data generated from Penn Treebank?
> > 2.1 If so, then University of Pennsylvania should hold the copyright
> > of these data, and are licensed for research purpose only?
> >
> > Best regards,
> > Anon.
> > --++**==--++**==--++**==--++**==--++**==--++**==--++**==
> > parser-support mailing list
> > parser-...@lists.stanford.edu
> >
> https://mailman.stanford.edu/mailman/listinfo/parser-support
> >
> >
>
>