[epe-users] high-level summary + system description

M, Simon simon.mille at upf.edu
Sat Aug 12 19:01:27 CEST 2017


Dear Stephan,

please find below the requested high-level information (sorry about the
delay). For DSyntS and PredArg we used rule-based modules so there is no
training data, but since they both include the surface parser as a first
step, I'm not sure if putting a non-applicable or repeat the SSynt
information. About the last column, we have no other publications which
contains more information about the modules as we used them.

upf 0 SSyntS WSJ 02–21 958,167 txt CoNLL'09
upf 1 DSyntS N/A? N/A? txt
upf 2 PredArg N/A? N/A? txt

I also attach the paper with the description of our submissions.
Best,

simon

On Tue, Aug 8, 2017 at 12:20 PM, Stephan Oepen <oe at ifi.uio.no> wrote:

> colleagues,
>
> in preparing our task summary, and to support everyone in trying to
> interpret our empirical results, we would like to gather some
> high-level information about all participating systems, viz. for each
> run:
>
> (0) representation: brief characterization of the type of
> dependencies, e.g. LTH, DM, UD, etc.
> (1) training: source (nature) of your training data, e.g. PTB
> converted by CoreNLP 3.3.1.
> (2) tokens: size of the training data, in tokens (according to your
> tokenization scheme).
> (3) input: whether the parser operated on ‘raw’ texts or our
> pre-processed input files.
> (4) reference: a pointer (web link or bibliographic reference) to more
> background.
>
> —we have added these columns to our master spreadsheet; see:
>
>   http://goo.gl/ZTZxXW
>
> in some cases, i was able to make informed guesses from the README
> files provided with each run.  but there is a lot of information
> missing that i would like to ask each team to fill in, preferably over
> the next few days.
>
> could one representative per team please quickly email
> ‘epe-organizers’ what you would like to appear for your submissions
> for questions (0) to (4) above?
>
> semi-relatedly, thanks to everyone who has filled in our short
> questionnaire about how to best wrap up EPE 2017 and plan for the
> future.  a few teams are still missing, and we would like to share a
> summary with everyone at the end of this week.
>
> if you have not done so already, please take a minute and provide your
> perspective to us:
>
>   http://goo.gl/tpDK3o
>
> best wishes, oe
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nlpl.eu/archives/epe-users/attachments/20170812/8ab7cffc/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TALN-UPF_EPE17_V0.5.pdf
Type: application/pdf
Size: 176184 bytes
Desc: not available
URL: <http://lists.nlpl.eu/archives/epe-users/attachments/20170812/8ab7cffc/attachment-0001.pdf>


More information about the epe-users mailing list