Nice thread on reddit about this: http://www.reddit.com/r/MachineLearning/comments/13zvxe/ip_law_and_machine_learning_who_owns_the_model/
On Sat, Nov 24, 2012 at 11:07 AM, Rupert Westenthaler < [email protected]> wrote: > On Fri, Nov 23, 2012 at 11:39 AM, Andrea Di Menna <[email protected]> > wrote: > > Even though I am not copying models in the datafiles dir, it looks > > like those models are anyway available in the stable launcher. > > > > My questions follow: > > 1) Are the en model from lang and ner bundles licensed with a Apache > > 2.0 license? > > No. This is the reason why you get Messages like that during the build > > * > * WARNING - this build downloads some OpenNLP files that are *not* > * licensed under the Apache License, and have more restrictive usage > * terms than the Apache Stanbol code. See STANBOL-545 for more > * information: https://issues.apache.org/jira/browse/STANBOL-545 > * > > > 2) Is there any safe/preferable way to remove those models from a > > Stanbol instance without completely disrupting the Keyword Linking > > engine? > > They KeywordLinking engine only requires Tokens. Those are also > available if no models are present. However this will have an influence > on the Results and the Performance. > > > I am wondering if those models are absolutely needed for the purpose > > of Keyword Linking or if the related bundles can be safely removed > > from the Felix console. > > > > just exclude/remove all org.apache.stanbol.data.opennlp.* bundles > > Regarding Licenses: You will find a lot of relevant posts on the > OpenNLP mailing lists. > > best > Rupert > > > > > [1] https://issues.apache.org/jira/browse/STANBOL-545 > > [2] http://opennlp.sourceforge.net/models-1.5/ > > > > -- > | Rupert Westenthaler [email protected] > | Bodenlehenstraße 11 ++43-699-11108907 > | A-5500 Bischofshofen > -- Pablo N. Mendes http://pablomendes.com
