The nominalized clauses analysis in the GM requires a special head-subj and head-opt-subj rule that does not require the head daughter to be an event.
The basic-head-opt-subj-phrase in matrix.tdl specifies event on the head daughter, and it’s not enough to remove that restriction because the INDEX is identified with CLAUSE-KEY, which is also type event. It looks like we have two options:
(1) Get the CLAUSE-KEY to point to the event inside the nominalized clause, or
(2) Decide its sensible to have the nominalized index be the CLAUSE-KEY (this would involve changing CLAUSE-KEY to event-or-ref-index).
Since this is an interaction with the information structure library, I am interested in what others (esp. Sanghoun) think is best.