[mrp-users] tokenization conventions in morpho-syntactic companion trees

[Stephan Oepen]

dear colleagues,

over the past few days, two teams have inquired about the specifics of
tokenization and anchoring conventions in our companion package.  seeing as
anchoring for most frameworks is part of the target representation (and,
thus, will be considered in scoring), getting these sub-string ranges right
is an important part of the parsing problem.

please consider the following conversation on Microsoft GitHub, where we
have tried to explain more the tokenization conventions in the companion
data:

https://github.com/cfmrp/mtool/issues/64

in unrelated news, there are 18 teams registered for the task on the
CodaLab site as of today, and we have received the first few submissions
... exciting!  if you intend to submit MRP parser outputs for evaluation
and have not yet registered on CodaLab, please do so immediately.  for the
final week of the evaluation phase, we may end up sending feedback to
registered participants through the CodaLab service (rather than through
this much larger mailing list).

finally, please remember that the evaluation period ends on thursday, july
25, 2019, at 12:00 noon (middle of the day) central european summer time.
 no parser outputs can be submitted for the official evaluation after that
date.

kind regards, oe (for the MRP co-organizers)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nlpl.eu/archives/mrp-users/attachments/20190718/c2807e45/attachment.html>