CoNLL 2019 Shared Task: Meaning Representation Parsing --- Training Update

Version 1.0; May 21, 2019


Overview
========

This directory contains an update to the UCCA training data for the MRP 2019
shared task.  The full training data for the task (semantic graphs in five
distinct frameworks: AMR, DM, EDS, PSD, and UCCA) is distributed through the
Linguistic Data Consortium.  Please see the MRP task web site for background
and information on how to access the training data:

  http://mrp.nlpl.eu

This package provides an overlay to the complete MRP training data, i.e. the
contents of this archive are expected to replace the ‘training/ucca/’ folder
from the original release.  By itself, this UCCA update is not sufficient for
participants to engage with the MRP 2019 task.

All graphs are encoded in a uniform abstract representation with a JSON-based
serialization;  The MRP interchange format is documented at:

  http://mrp.nlpl.eu/index.php?page=4#format


Contents
========

The main contents in this release are the following files:

  $ wc -l ucca/*.mrp
  3812 ucca/ewt.mrp
  2673 ucca/wiki.mrp
    87 ucca/wsj.mrp
  6572 total

Here, line counts correspond to the number of graphs available.  In comparison
to the original release of the MRP training data, there are some 500 additional
graphs in the EWT segment, and about one thousand new graphs in the Wikipedia
segment.  Besides adding more training data, the annotations in this re-release
have been revised to better conform to the UCCA 2.x specification, and sentence
and token splitting has been improved.


Release History
===============

[Version 1.0; May 21, 2019]

+ Re-release of the UCCA part of the MRP training data: more and better graphs.


Contact
=======

For questions or comments, please do not hesitate to email the task organizers
at: ‘mrp-organizers@nlpl.eu’.

Omri Abend
Jan Hajič
Daniel Hershcovich
Marco Kuhlmann
Stephan Oepen
Tim O'Gorman
Nianwen Xue