You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@opennlp.apache.org by Jeff Zemerick <jz...@apache.org> on 2021/03/12 13:39:22 UTC
[VOTE] Release OpenNLP Models trained on UD
All,
I am calling a vote to release the Apache OpenNLP models trained on the
Universal Dependencies corpus. The models were described in a previous
thread you can see at
https://www.mail-archive.com/dev@opennlp.apache.org/msg03054.html.
This vote is to release the models as version 1.0. (The models are still
available in the Dropbox folder at
https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
along with text files showing the training and evaluation results).
Upon a successful vote the models will be made available on the Apache
OpenNLP website on the Models Download page (
https://opennlp.apache.org/models.html). Work will then continue to modify
OpenNLP to be able to automatically download and use these pre-trained
models when none is specifically loaded by the user. (The goal being to
lower the barrier of entry into OpenNLP and make its use more convenient.)
Please vote on releasing the models in the linked DropBox folder as model
version 1.0. The vote is open for at least the next 72 hours.
Only votes from the OpenNLP PMC are binding but everyone is welcome to
check the models and vote. The vote passes if at least three binding +1
votes are cast.
[ ] +1 Release the models as model version 1.0.
[ ] -1 Do not release the models because...
Thanks!
Jeff
Re: [VOTE] Release OpenNLP Models trained on UD
Posted by Tommaso Teofili <to...@gmail.com>.
+1
Tommaso
On Sat, 13 Mar 2021 at 08:25, Bruno P. Kinoshita <ki...@apache.org> wrote:
> [x] +1 Release the models as model version 1.0.
>
>
> On Saturday, 13 March 2021, 2:39:37 am NZDT, Jeff Zemerick <
> jzemerick@apache.org> wrote:
>
> All,
>
> I am calling a vote to release the Apache OpenNLP models trained on the
> Universal Dependencies corpus. The models were described in a previous
> thread you can see at
> https://www.mail-archive.com/dev@opennlp.apache.org/msg03054.html.
>
> This vote is to release the models as version 1.0. (The models are still
> available in the Dropbox folder at
>
> https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
> along with text files showing the training and evaluation results).
>
> Upon a successful vote the models will be made available on the Apache
> OpenNLP website on the Models Download page (
> https://opennlp.apache.org/models.html). Work will then continue to modify
> OpenNLP to be able to automatically download and use these pre-trained
> models when none is specifically loaded by the user. (The goal being to
> lower the barrier of entry into OpenNLP and make its use more convenient.)
>
> Please vote on releasing the models in the linked DropBox folder as model
> version 1.0. The vote is open for at least the next 72 hours.
>
> Only votes from the OpenNLP PMC are binding but everyone is welcome to
> check the models and vote. The vote passes if at least three binding +1
> votes are cast.
>
> [ ] +1 Release the models as model version 1.0.
> [ ] -1 Do not release the models because...
>
> Thanks!
> Jeff
>
Re: [VOTE] Release OpenNLP Models trained on UD
Posted by "Bruno P. Kinoshita" <ki...@apache.org>.
[x] +1 Release the models as model version 1.0.
On Saturday, 13 March 2021, 2:39:37 am NZDT, Jeff Zemerick <jz...@apache.org> wrote:
All,
I am calling a vote to release the Apache OpenNLP models trained on the
Universal Dependencies corpus. The models were described in a previous
thread you can see at
https://www.mail-archive.com/dev@opennlp.apache.org/msg03054.html.
This vote is to release the models as version 1.0. (The models are still
available in the Dropbox folder at
https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
along with text files showing the training and evaluation results).
Upon a successful vote the models will be made available on the Apache
OpenNLP website on the Models Download page (
https://opennlp.apache.org/models.html). Work will then continue to modify
OpenNLP to be able to automatically download and use these pre-trained
models when none is specifically loaded by the user. (The goal being to
lower the barrier of entry into OpenNLP and make its use more convenient.)
Please vote on releasing the models in the linked DropBox folder as model
version 1.0. The vote is open for at least the next 72 hours.
Only votes from the OpenNLP PMC are binding but everyone is welcome to
check the models and vote. The vote passes if at least three binding +1
votes are cast.
[ ] +1 Release the models as model version 1.0.
[ ] -1 Do not release the models because...
Thanks!
Jeff
Re: [VOTE] Release OpenNLP Models trained on UD
Posted by Peter Thygesen <pe...@gmail.com>.
+1
/Peter
On 2021-03-12, 14:39, "Jeff Zemerick" <jz...@apache.org> wrote:
All,
I am calling a vote to release the Apache OpenNLP models trained on the
Universal Dependencies corpus. The models were described in a previous
thread you can see at
https://www.mail-archive.com/dev@opennlp.apache.org/msg03054.html.
This vote is to release the models as version 1.0. (The models are still
available in the Dropbox folder at
https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
along with text files showing the training and evaluation results).
Upon a successful vote the models will be made available on the Apache
OpenNLP website on the Models Download page (
https://opennlp.apache.org/models.html). Work will then continue to modify
OpenNLP to be able to automatically download and use these pre-trained
models when none is specifically loaded by the user. (The goal being to
lower the barrier of entry into OpenNLP and make its use more convenient.)
Please vote on releasing the models in the linked DropBox folder as model
version 1.0. The vote is open for at least the next 72 hours.
Only votes from the OpenNLP PMC are binding but everyone is welcome to
check the models and vote. The vote passes if at least three binding +1
votes are cast.
[ ] +1 Release the models as model version 1.0.
[ ] -1 Do not release the models because...
Thanks!
Jeff
Re: [VOTE] Release OpenNLP Models trained on UD
Posted by Rodrigo Agerri <ro...@ehu.eus>.
+1
Cheers,
R
On Fri, 12 Mar 2021 at 14:39, Jeff Zemerick <jz...@apache.org> wrote:
>
> All,
>
> I am calling a vote to release the Apache OpenNLP models trained on the
> Universal Dependencies corpus. The models were described in a previous
> thread you can see at
> https://www.mail-archive.com/dev@opennlp.apache.org/msg03054.html.
>
> This vote is to release the models as version 1.0. (The models are still
> available in the Dropbox folder at
> https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
> along with text files showing the training and evaluation results).
>
> Upon a successful vote the models will be made available on the Apache
> OpenNLP website on the Models Download page (
> https://opennlp.apache.org/models.html). Work will then continue to modify
> OpenNLP to be able to automatically download and use these pre-trained
> models when none is specifically loaded by the user. (The goal being to
> lower the barrier of entry into OpenNLP and make its use more convenient.)
>
> Please vote on releasing the models in the linked DropBox folder as model
> version 1.0. The vote is open for at least the next 72 hours.
>
> Only votes from the OpenNLP PMC are binding but everyone is welcome to
> check the models and vote. The vote passes if at least three binding +1
> votes are cast.
>
> [ ] +1 Release the models as model version 1.0.
> [ ] -1 Do not release the models because...
>
> Thanks!
> Jeff
Re: [VOTE] Release OpenNLP Models trained on UD
Posted by Jeff Zemerick <jz...@apache.org>.
Joey,
I personally have been thinking of the models as convenience binaries (even
though my vote thread didn't give that impression! :) because users can
always train the models themselves much like they can build OpenNLP
themselves. Instead, as a convenience we are providing pretrained model
files to remove that work from the user. I'm certainly open to the
otherside if that thinking is wrong.
Flyby comments are always welcome!
Thanks,
Jeff
On Mon, Mar 15, 2021 at 6:24 PM Joey Frazee <jo...@icloud.com.invalid>
wrote:
> Sorry for the flyby comment...
>
> What’s the status of the models? Are they being treated like convenience
> binaries or are they a source release artifact?
>
> I ask because it’s relevant to what the vote is about, the process, and
> the options for distribution.
>
> -joey
>
> > On Mar 15, 2021, at 2:22 PM, Jeff Zemerick <jz...@apache.org> wrote:
> >
> > Richard, thanks for your feedback. That is all very important and I
> > apologize for overlooking those things. And as a first release it's
> > important we capture and document those steps to make future releases
> > easier.
> >
> > Thanks to those who voted but we'll cancel this vote and try again soon.
> >
> > Thanks,
> > Jeff
> >
> >
> >> On Sun, Mar 14, 2021 at 3:00 AM Richard Eckart de Castilho <
> rec@apache.org>
> >> wrote:
> >>
> >>> On 12. Mar 2021, at 14:39, Jeff Zemerick <jz...@apache.org> wrote:
> >>>
> >>> This vote is to release the models as version 1.0. (The models are
> still
> >>> available in the Dropbox folder at
> >>>
> >>
> https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
> >>> along with text files showing the training and evaluation results).
> >>
> >> As far as I can see, the files are not signed and cannot be validated
> for
> >> authenticity.
> >> There are no release notes, no information where the files come from, no
> >> license
> >> information, etc. The files are also no on ASF systems (cf.
> >> https://infra.apache.org/release-publishing.html).
> >>
> >> I don't see how in this situation a release vote according to ASF
> >> standards is possible. That is
> >> a vote that:
> >>
> >>> A binding release vote of the PMC is the critical gating step in the
> >> release process. Without such a vote, the release is just a set of files
> >> prepared by an individual. After such a vote, it is a formal offering of
> >> the ASF, backed by the "full faith and credit" of the Foundation.
> >>
> >> However, the voting thread "looks" official - so it looks like you aim
> for
> >> such a proper release.
> >>
> >> I do not think the process should continue as is and should either be
> >> restarted properly or
> >> not take the appearance of being an official release process.
> >>
> >> -1 (non-binding)
> >>
> >> Cheers,
> >>
> >> -- Richard
> >>
> >>
>
Re: [VOTE] Release OpenNLP Models trained on UD
Posted by Joey Frazee <jo...@icloud.com.INVALID>.
Sorry for the flyby comment...
What’s the status of the models? Are they being treated like convenience binaries or are they a source release artifact?
I ask because it’s relevant to what the vote is about, the process, and the options for distribution.
-joey
> On Mar 15, 2021, at 2:22 PM, Jeff Zemerick <jz...@apache.org> wrote:
>
> Richard, thanks for your feedback. That is all very important and I
> apologize for overlooking those things. And as a first release it's
> important we capture and document those steps to make future releases
> easier.
>
> Thanks to those who voted but we'll cancel this vote and try again soon.
>
> Thanks,
> Jeff
>
>
>> On Sun, Mar 14, 2021 at 3:00 AM Richard Eckart de Castilho <re...@apache.org>
>> wrote:
>>
>>> On 12. Mar 2021, at 14:39, Jeff Zemerick <jz...@apache.org> wrote:
>>>
>>> This vote is to release the models as version 1.0. (The models are still
>>> available in the Dropbox folder at
>>>
>> https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
>>> along with text files showing the training and evaluation results).
>>
>> As far as I can see, the files are not signed and cannot be validated for
>> authenticity.
>> There are no release notes, no information where the files come from, no
>> license
>> information, etc. The files are also no on ASF systems (cf.
>> https://infra.apache.org/release-publishing.html).
>>
>> I don't see how in this situation a release vote according to ASF
>> standards is possible. That is
>> a vote that:
>>
>>> A binding release vote of the PMC is the critical gating step in the
>> release process. Without such a vote, the release is just a set of files
>> prepared by an individual. After such a vote, it is a formal offering of
>> the ASF, backed by the "full faith and credit" of the Foundation.
>>
>> However, the voting thread "looks" official - so it looks like you aim for
>> such a proper release.
>>
>> I do not think the process should continue as is and should either be
>> restarted properly or
>> not take the appearance of being an official release process.
>>
>> -1 (non-binding)
>>
>> Cheers,
>>
>> -- Richard
>>
>>
Re: [VOTE] Release OpenNLP Models trained on UD
Posted by Jeff Zemerick <jz...@apache.org>.
Richard, thanks for your feedback. That is all very important and I
apologize for overlooking those things. And as a first release it's
important we capture and document those steps to make future releases
easier.
Thanks to those who voted but we'll cancel this vote and try again soon.
Thanks,
Jeff
On Sun, Mar 14, 2021 at 3:00 AM Richard Eckart de Castilho <re...@apache.org>
wrote:
> On 12. Mar 2021, at 14:39, Jeff Zemerick <jz...@apache.org> wrote:
> >
> > This vote is to release the models as version 1.0. (The models are still
> > available in the Dropbox folder at
> >
> https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
> > along with text files showing the training and evaluation results).
>
> As far as I can see, the files are not signed and cannot be validated for
> authenticity.
> There are no release notes, no information where the files come from, no
> license
> information, etc. The files are also no on ASF systems (cf.
> https://infra.apache.org/release-publishing.html).
>
> I don't see how in this situation a release vote according to ASF
> standards is possible. That is
> a vote that:
>
> > A binding release vote of the PMC is the critical gating step in the
> release process. Without such a vote, the release is just a set of files
> prepared by an individual. After such a vote, it is a formal offering of
> the ASF, backed by the "full faith and credit" of the Foundation.
>
> However, the voting thread "looks" official - so it looks like you aim for
> such a proper release.
>
> I do not think the process should continue as is and should either be
> restarted properly or
> not take the appearance of being an official release process.
>
> -1 (non-binding)
>
> Cheers,
>
> -- Richard
>
>
Re: [VOTE] Release OpenNLP Models trained on UD
Posted by Richard Eckart de Castilho <re...@apache.org>.
On 12. Mar 2021, at 14:39, Jeff Zemerick <jz...@apache.org> wrote:
>
> This vote is to release the models as version 1.0. (The models are still
> available in the Dropbox folder at
> https://www.dropbox.com/sh/p8focuz0qwvw84b/AAC6GqO8mqZn_xkAqHZsVAsoa?dl=0&lst=
> along with text files showing the training and evaluation results).
As far as I can see, the files are not signed and cannot be validated for authenticity.
There are no release notes, no information where the files come from, no license
information, etc. The files are also no on ASF systems (cf. https://infra.apache.org/release-publishing.html).
I don't see how in this situation a release vote according to ASF standards is possible. That is
a vote that:
> A binding release vote of the PMC is the critical gating step in the release process. Without such a vote, the release is just a set of files prepared by an individual. After such a vote, it is a formal offering of the ASF, backed by the "full faith and credit" of the Foundation.
However, the voting thread "looks" official - so it looks like you aim for such a proper release.
I do not think the process should continue as is and should either be restarted properly or
not take the appearance of being an official release process.
-1 (non-binding)
Cheers,
-- Richard