MRS uses underspecified variables (e.g., i5) for things like dropped arguments, but DMRS just drops them. When they carry information in the MRS (e.g., agreement, coindexation, etc.), this is currently lost in DMRS. It should be possible to instantiate Nodes for these when necessary.
See http://lists.delph-in.net/archives/developers/2018/002807.html, and also from Ann in an email with the folks at Cambridge:
The most general solution seems to be to introduce a new node for the
unexpressed argument, making it clear that it is an unexpressed node as
far as the grammar is concerned. It can have all the usual
attributes. In the example above, we would then have an ICONS
relationship between this unexpressed node and the node for dogs. The
ARG1 link from chase would go to the unexpressed node. This means the
DMRS is different from the situation where the grammar managed to
identify the ARG1 as dogs but I think that is the correct approach, at
least until we have a much better understanding of the inventory of
these things.
MRS uses underspecified variables (e.g.,
i5) for things like dropped arguments, but DMRS just drops them. When they carry information in the MRS (e.g., agreement, coindexation, etc.), this is currently lost in DMRS. It should be possible to instantiate Nodes for these when necessary.See http://lists.delph-in.net/archives/developers/2018/002807.html, and also from Ann in an email with the folks at Cambridge: