www-legal-discuss mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Richard Eckart de Castilho (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (LEGAL-309) Apache OpenNLP wants to release models trained on Universal Dependency under AL 2.0
Date Fri, 07 Jul 2017 15:44:00 GMT

    [ https://issues.apache.org/jira/browse/LEGAL-309?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16078272#comment-16078272
] 

Richard Eckart de Castilho commented on LEGAL-309:
--------------------------------------------------

Ok, I am not very familiar with the concept of fair use, since this is a US-specific thing.
I agree that there is little risk since the data will be separate from the code and could
be easily removed if anybody would complain. But I wonder why the fair use wouldn't apply
to any other data such as the OntoNotes corpus from LDC which OpenNLP previously OpenNLP asked
to be using as a training corpus and was denied it.

> Apache OpenNLP wants to release models trained on Universal Dependency under AL 2.0
> -----------------------------------------------------------------------------------
>
>                 Key: LEGAL-309
>                 URL: https://issues.apache.org/jira/browse/LEGAL-309
>             Project: Legal Discuss
>          Issue Type: Question
>            Reporter: Joern Kottmann
>            Assignee: Chris A. Mattmann
>
> The OpenNLP project develops statistical natural language processing software which needs
to be trained in order to produce a model that can be used to perform one of our supported
tasks such as part-of-speech tagging or lemmatization.
> We would like to know if it would be possible to train models on data included in UD
which itself is licensed under various licenses and then release the trained models under
AL 2.0.
> If you go to [1] you can see a list of data files and their license.
> Here is a list of the licenses:
> CC BY 4.0
> CC BY SA 4.0
> CC BY-NC-SA 2.5, 3.0, 4.0 and without version
> CC BY-NC-SA US 3.0
> CC BY-SA 4.0 
> GPL
> LGPLLR
> The models we would like to train on that data are:
> - Part-of-Speech models (contains bigrams and a set of individual words of the training
text)
> - Lemmatizer (contains a set of individual words of the training text)
> As far as we understand individual words or very short phrases extracted from a corpus
are not protected by its original copyright. The above licenses as far as we know don't forbid
to derive statistics from its content. 
> [1] http://universaldependencies.org/
>  



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: legal-discuss-unsubscribe@apache.org
For additional commands, e-mail: legal-discuss-help@apache.org


Mime
View raw message