You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@opennlp.apache.org by Michael Poprat <mi...@gmail.com> on 2011/08/16 15:01:34 UTC

OpenNLP-Tools release plan

Dear OpenNLP-team,

we are planning to upgrade our OpenNLP components from 1.4.3 to
1.5.1-incubating. We would like to know in advance if you intend to
release a 1.6.0 version in the very near future (in the next 2 or 3
weeks) or if the most recent version 1.5.1-incubating is stable for
the next few months.

Thanks for your answer!

Best

Michael

Re: OpenNLP-Tools release plan

Posted by Jörn Kottmann <ko...@gmail.com>.
On 8/16/11 3:09 PM, Jörn Kottmann wrote:
> On 8/16/11 3:01 PM, Michael Poprat wrote:
>> Dear OpenNLP-team,
>>
>> we are planning to upgrade our OpenNLP components from 1.4.3 to
>> 1.5.1-incubating. We would like to know in advance if you intend to
>> release a 1.6.0 version in the very near future (in the next 2 or 3
>> weeks) or if the most recent version 1.5.1-incubating is stable for
>> the next few months.
>
> We are planning to release 1.5.2 in the next weeks. After that we haven't
> yet decided if we do one more 1.5 release, or incorporate bigger changes
> and go to 1.6.
>
> There might be a few minor issues when you go from 1.4.3 to 1.5.1,
> which components do you use? Do you have self-trained models?

Michael Poprat:
"We use the tokenizer, the sentence splitter, the POS-tagger and the
chunker and yes, we have self-trained models. I assume that the models
of 1.4.3 aren't valid anymore, but I hope that the training format
hasn't changed too much ..."

Compared to 1.4, we introduced a new model format. All artifacts necessary
to instantiate a component are now packaged in a zip container. To 
instantiate
a component you simply need to hand in an InputStream to this model package.
That is easier than before, because the components can take care of setting
everything up as it has been during training.

We also changed our training API a bit, to make this work, but it should 
not be
difficult to adapt your existing code to it.

As far as I remember the training format stayed completely compatible.

If you are training on publicly available corpora, you could consider to 
sponser your parsing
code to our formats package, where we already have format conversion 
code for several
corpora.

Jörn





Re: OpenNLP-Tools release plan

Posted by Jörn Kottmann <ko...@gmail.com>.
On 8/16/11 3:01 PM, Michael Poprat wrote:
> Dear OpenNLP-team,
>
> we are planning to upgrade our OpenNLP components from 1.4.3 to
> 1.5.1-incubating. We would like to know in advance if you intend to
> release a 1.6.0 version in the very near future (in the next 2 or 3
> weeks) or if the most recent version 1.5.1-incubating is stable for
> the next few months.

We are planning to release 1.5.2 in the next weeks. After that we haven't
yet decided if we do one more 1.5 release, or incorporate bigger changes
and go to 1.6.

There might be a few minor issues when you go from 1.4.3 to 1.5.1,
which components do you use? Do you have self-trained models?

Jörn