On Sun, 13 Dec 2020, at 10:03, Andrej Shadura wrote: > I started working on it back then but I ran into a dependency on > opennlp-models, which come with no license. I tried to find a > workaround but I couldn’t find enough time for that.
Just a few informations for the next person interested in it. The opennlp-models have no licencing info because they are models trained on a copyright protected corpus, and it is not clear what happens in such situation. http://mail-archives.apache.org/mod_mbox/opennlp-dev/201912.mbox/browser Github seems to consider that ML is a great IP washer with its copilot, but I'm not sure that this is a good example for the rest of us :) This specific issue is also discussed here: https://github.com/languagetool-org/languagetool/issues/2259 This sounds like a great challenge to me. I'd love to help unlocking this problem, but I have no idea of what it takes to train new models in the first place, not to speak about ensuring that the result has a clean open-source licence. Any hint would be more than welcome here. Bye, Mt. -- Dans le passé, il y avait plus de futur que maintenant. -- Le Chat
signature.asc
Description: PGP signature