Skip to content

Latest commit

 

History

History
68 lines (50 loc) · 2.75 KB

LICENCE.md

File metadata and controls

68 lines (50 loc) · 2.75 KB

Governing Licences

This project and all contributions are per se dual-licenced under the:

  • ASLv2 licence (LICENCE-ASL2.txt)

  • GPLv3 (LICENCE-GPL3.txt) or any later version of the GPL licence

You can choose, at your option, to follow the terms of either or produce derivative works licenced under the terms of either, however you are encourged to dual-licence your derivative works in the same way so that code can be merged back to the original project unhampered.

Rationale

We prefer the ASLv2 licence, which is the common licence in the Java NLP ecosystem and we would like to encourage wide usage of our framework; we would appreciate it very much if all users would make their versions of the framework open source (and we believe QA/NLP researchers using the system should do it), but we do not want to scare away company users by making it a legal requirement, with the hope that they will still contribute back as much as they can.

The primary reason for dual-licencing is the fact that some components are licenced under the GPLv2+ licence. Therefore, a binary distribution of this project that is built without modifications includes calls to (third-party, source not included in this project) GPLv2 components and may require to follow the GPLv3 licence!

However, in all cases the GPLv2 components are not essential, are isolated through a generic (non-GPL'd) API, and can be removed without rendering the system useless (just at the cost of some performance).

These components are GPLv2+:

  • Stanford CoreNLP - used through DKpro generic API mainly for producing dependency parse trees of sentences, other alternatives with worse performance exist

  • ClearTK CRFSuiteWrapper - used through ClearTK generic API, used only for training a CRF model (only when YodaQA is invoked as shown in data/ml/biocrf/README.md in particular; not used for CRF-based biotagging during regular YodaQA usage)

Acknowledgements

The software is copyrighted by its respective authors as denoted in the publicly recorded Git history.

Work on YodaQA has been in part kindly sponsored by the Medialab foundation (http://medialab.cz/), a Czech Technical University incubator. The rest of contributions by Petr Baudiš is licenced as open source via Ailao (http://ailao.eu/).

Ailao also provides commercial consulting, customization, deployment and support services.

YodaQA uses many libraries to get things done. It is built on top of the Apache UIMA and DKpro UIMA bindings, used for most of the NLP logic (including the NLP type system); transitively, work like the StanfordParser and Princeton's Wordnet is crucial for us. For machine learning, we use chiefly scikit-learn, but also a little of crfsuite (wrapped by ClearTK, jcrfsuite).