[NLPL Task Force (A)] NLPL parsning

Wed Nov 27 09:08:31 UTC 2019

i shall provide a DyNet module by the start of next week on Saga, but we
are behind regarding Puhti.

apologies for slow follow-up, because end of term, and hoting a two-week
boot camp!

oe

On Wed, 27 Nov 2019 at 09:38 Sara Stymne <sara.stymne at lingfil.uu.se> wrote:

> Hi all, and especially Stephan,
>
>
> Is there any news on what the infrastructure task force ended up
> installing? Especialy, did you install Dynet in the end?
>
>
> Best,
>
> Sara
>
>
> ------------------------------
> *Från:* Sara Stymne
> *Skickat:* den 7 november 2019 13:13:17
> *Till:* Stephan Oepen
> *Kopia:* Tiedemann, Jörg; Ali Basirat; Filip Ginter; Joakim Nivre;
> infrastructure
> *Ämne:* Re: NLPL parsning
>
> Hi,
>
> Thanks all for the updates! This sounds good. We will discuss then at
> Uppsala, what we want to add in addition to what you do. We'll aim at
> installing in both places.
>
> Our parser requires Dynet. Lately we have been installing that locally,
> since system versions haven't been in sync, and installation is much easier
> than it used to be. But if the infrastructure task force could consider
> adding an NLPL Dynet module, that would be useful, at least to us. The
> latest version should work for us.
>
> Best,
> Sara
>
>
> 7 nov 2019 kl. 13:04 skrev Stephan Oepen <oe at ifi.uio.no>
> :
>
> dear all,
>
> in general, we are aiming for greatly increased uniformity in this round,
> i.e. the same modules in the same versions on both Puhti and Saga.  thus,
> my recommendation would be fewer modules, if need be, but always think both
> systems.
>
> we are currently working to provision a uniform set of ’base‘ modules,
> e.g. NumPy, SciPy, Gensim, NLTK, spaCy, PyTorch, Keras, TensorFlow,
> CoreNLP, and StanfordNLP.  in case your plans for parsing modules require
> anything in addition to these, please let the infrastructure task force
> know quickly.  we are currently standardizing on Python 3.7; if you
> absolutely need other versions, please get in touch!
>
> even though several of the above are also available as system modules
> (i.e. not maintained by NLPL), i would encourage you to build NLPL modules
> on top of the NLPL versions of these libraries, again for uniformity across
> systems.  the above basic collection is near-complete on Saga, and we
> expect to put the same NLPL modules in place in Puhti this coming week.
>
> please note some updates to the software installation guide, also
> regarding ad-hoc versioning (e.g. ‘201911’) where there are no externally
> defined versions:
>
> http://wiki.nlpl.eu/index.php/Infrastructure/installation/guide
>
> all best, oe
>
>
> On Thu, 7 Nov 2019 at 12:45 Tiedemann, Jörg <jorg.tiedemann at helsinki.fi>
> wrote:
>
>> Hi,
>>
>> I will probably package the old udpipe into a module for some
>> compatibility reasons. If you want to make one for udpipe future then this
>> would be great as well (I actually don’t really know what the status of
>> that package is at the moment).
>>
>> All the best,
>> Jörg
>>
>>
>> ********************************************************************************************
>>
>> Jörg Tiedemann
>> Language Technology https://blogs.helsinki.fi/language-technology/
>> University of Helsinki
>>
>> On 07 Nov 2019, at 12:50, Filip Ginter <figint at utu.fi> wrote:
>>
>> Hi Sara
>>
>> OK. We will look into packaging the current parser pipeline we have as a
>> puhti module. We will also look into packaging Udify into the pipeline, the
>> module structure would be the same then.
>>
>> - Filip
>>
>>
>> On Wed, Nov 6, 2019 at 5:15 PM Sara Stymne <sara.stymne at lingfil.uu.se>
>> wrote:
>>
>>> Hi everyone,
>>>
>>>
>>> It is time to finalize the parsing deliverable for NLPL. We should make
>>> a plan for this. Our main idea is to migrate to Saga (and possibly Puhti)
>>> and mainly update the existing tools. Feel free to add any other relevant
>>> person to this conversation.
>>>
>>>
>>> Jörg, you are not really involved. I just wanted to ask you if you plan
>>> to keep UD-pipe installed as in previous years? Or if we should consider
>>> doing that instead?
>>>
>>>
>>> Filip, do you have any plans for parsing? Maybe you can package your
>>> parser from CoNLL'18, and/or your cool FinBERT+Udify parser? Since they
>>> have really good results.
>>>
>>>
>>> Stephan, will you take care of the semantic parsers and data again? And
>>> if I understood correctly, you also installed some baseline parsers last
>>> year. Is that something you plan to do again on Saga?
>>>
>>>
>>> We plan to update our UUparser installation, and also add models for the
>>> Nordic languages. We will install some other state-of-the-art parser. Maybe
>>> Udify. And also update the data and tutorial. Then it depends a bit on your
>>> plans, if we also will install some more parsers.
>>>
>>>
>>> Also, let me know if you have any other plans or ideas concerning the
>>> parsing part of NLPL!
>>>
>>>
>>> Best,
>>>
>>> Sara
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> När du har kontakt med oss på Uppsala universitet med e-post så innebär
>>> det att vi behandlar dina personuppgifter. För att läsa mer om hur vi gör
>>> det kan du läsa här: http://www.uu.se/om-uu/dataskydd-personuppgifter/
>>>
>>> E-mailing Uppsala University means that we will process your personal
>>> data. For more information on how this is performed, please read here:
>>> http://www.uu.se/en/about-uu/data-protection-policy
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nlpl.eu/archives/infrastructure/attachments/20191127/a4ffca95/attachment.htm>