[mrp-users] clarification regarding the training data

[Stephan Oepen]

dear colleagues,

some of you asked about the exact counts of graphs per framework in
the MRP 2019 training data.  this revealed that our statistics on the
task web site were incorrect (if in relatively minor ways).  earlier
today, we have posted updated sentence and token counts for the
training data:

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

more importantly, this exchange revealed that at least one team was
not aware of the update to the UCCA training graphs that was released
on may 21, superseding and extending the earlier UCCA annotations in
the training data package originally distributed via the LDC:

http://svn.nlpl.eu/mrp/2019/public/ucca.tgz

we would kindly like to encourage everyone to regularly review the
announcements of major updates on the task web pages, as well as the
archives of this mailing list:

http://mrp.nlpl.eu/index.php?page=7#news
http://lists.nlpl.eu/archives/mrp-users/2019/date.html

best wishes, oe (for the MRP co-organizers)